diff --git a/qoi/.gitignore b/qoi/.gitignore new file mode 100644 index 00000000..daec2ef0 --- /dev/null +++ b/qoi/.gitignore @@ -0,0 +1,124 @@ +genc/genc-qoiconv +genc/genc-qoibench +genc/stainless.c +genc/stainless.h +output/ + +# Stainless +/bin/stainless-* +.stainless-cache/ +.stainless.conf +.stainless-external-tests/ +.stainless-package-standalone/ + +# Vim +*.swp + +# emacs +*~ + +# Ripgrep +.rgignore + +# ensime +.ensime +.ensime_cache + +# sbt +target +/project/build +project/project +/stainless +/.test-history +/out-classes + +# synthesis +derivation*.dot + +# stainless +last.log +*.last +*.table +*.log +smt-sessions/ +tip-sessions/ +pri-sessions/ +stainless.out/ +repairs.dat +tmp +stainless-stack-trace.txt + +#eclipse +.cache +.classpath +.project +.settings/ +.worksheet/ + +#intellij +.idea +Leon.eml +Leon.iml +*.iml +/.idea_modules + +#scripts +out-classes + +#z3 +.z3-trace + +#travis +/travis/builds + +# Windows +/cvc4.exe +/libz3.dll +/libz3.lib +/msvcp100.dll +/msvcr100.dll +/scalaz3.dll +/scalaz3.lib +/vcomp100.dll +/z3.exe +/libz3java.dll +/libz3java.lib + +# Coq files +*# +.#* +*.aux +*.glob +*.vo +*.v.d +verif*.v +.lia.cache +.nia.cache +.coqdeps.d +*.vok +*.vos +slc-lib/.Makefile.d +slc-lib/Makefile +slc-lib/Makefile.conf + +# metals and VSCode +.vscode +.metals +.bsp +project/metals.sbt +null +.bloop/ +project/.bloop/ +project/project/ +.bsp + +# ripgrep +.rgignore + +# C files +*.out + +# LZW GenC benchmarks files +input.txt +decoded.txt +encoded.txt diff --git a/qoi/genc/Makefile b/qoi/genc/Makefile new file mode 100644 index 00000000..dd46f461 --- /dev/null +++ b/qoi/genc/Makefile @@ -0,0 +1,20 @@ +.PHONY: genc + +CC=gcc # Clang seems to not be able to eliminate tail-calls present in decodeLoop and encodeLoop + # For, macOS users, gcc is an alias to clang, make sure to have gcc installed! +CFLAGS=-O3 + +all: genc genc-qoibench genc-qoiconv + +stainless.c: + stainless-dotty ../verified/common.scala ../verified/decoder.scala ../verified/encoder.scala \ + --genc -J-Xms10G -J-Xss20M + +genc: stainless.c + +genc-qoibench: stainless.c genc-qoibench.c genc-bridge.c + +genc-qoiconv: stainless.c genc-qoiconv.c genc-bridge.c + +clean: + rm -f genc-qoibench genc-qoiconv stainless.c stainless.h \ No newline at end of file diff --git a/qoi/genc/genc-bridge.c b/qoi/genc/genc-bridge.c new file mode 100644 index 00000000..7973d5e8 --- /dev/null +++ b/qoi/genc/genc-bridge.c @@ -0,0 +1,78 @@ +#include +#include "stainless.h" +#include "genc-bridge.h" + +int genc_qoi_decode(void *data, size_t length, void **pixels, int *w, int *h, int *channels) { + OptionMut_DecodedResult decoded_res = decode((array_int8) { .data = data, .length = length }, length); + if (decoded_res.tag == tag_SomeMut_DecodedResult) { + *pixels = decoded_res.value.SomeMut_DecodedResult_v.v.pixels.data; + *w = decoded_res.value.SomeMut_DecodedResult_v.v.w; + *h = decoded_res.value.SomeMut_DecodedResult_v.v.h; + *channels = decoded_res.value.SomeMut_DecodedResult_v.v.chan; + return 1; + } else { + return 0; + } +} + +int genc_qoi_encode(void *pixels, int w, int h, int channels, void **data, size_t *length) { + OptionMut_EncodedResult encoded_res = encode((array_int8) { .data = pixels, .length = w * h * channels }, w, h, channels); + if (encoded_res.tag == tag_SomeMut_EncodedResult) { + *data = encoded_res.value.SomeMut_EncodedResult_v.v.encoded.data; + *length = encoded_res.value.SomeMut_EncodedResult_v.v.length; + return 1; + } else { + return 0; + } +} + +int genc_qoi_read(const char *filename, void **pixels, int *w, int *h, int *channels) { + FILE *f = fopen(filename, "rb"); + + if (!f) { + return 0; + } + + fseek(f, 0, SEEK_END); + long size = ftell(f); + if (size <= 0) { + fclose(f); + return 0; + } + fseek(f, 0, SEEK_SET); + + int8_t *data = malloc(size); + if (!data) { + fclose(f); + return 0; + } + + size_t bytes_read = fread(data, 1, size, f); + fclose(f); + if (bytes_read != size) { + return 0; + } + + int res = genc_qoi_decode(data, size, pixels, w, h, channels); + free(data); + return res; +} + +int genc_qoi_write(const char *filename, void *pixels, int w, int h, int channels) { + void *data; + size_t length; + int enc_res = genc_qoi_encode(pixels, w, h, channels, &data, &length); + if (!enc_res) { + return 0; + } + + FILE *f = fopen(filename, "wb"); + if (!f) { + return 0; + } + + size_t written = fwrite(data, 1, length, f); + fclose(f); + free(data); + return written == length; +} diff --git a/qoi/genc/genc-bridge.h b/qoi/genc/genc-bridge.h new file mode 100644 index 00000000..35f1e089 --- /dev/null +++ b/qoi/genc/genc-bridge.h @@ -0,0 +1,9 @@ +#include + +int genc_qoi_decode(void *data, size_t length, void **pixels, int *w, int *h, int *channels); + +int genc_qoi_encode(void *pixels, int w, int h, int channels, void **data, size_t *length); + +int genc_qoi_read(const char *filename, void **pixels, int *w, int *h, int *channels); + +int genc_qoi_write(const char *filename, void *pixels, int w, int h, int channels); \ No newline at end of file diff --git a/qoi/genc/genc-qoibench.c b/qoi/genc/genc-qoibench.c new file mode 100644 index 00000000..2cb9624d --- /dev/null +++ b/qoi/genc/genc-qoibench.c @@ -0,0 +1,489 @@ +// Tweaked version of https://github.com/phoboslab/qoi/blob/master/qoibench.c +// with the following changes: +// - Removed libpng benchmarking +// - Added genc_qoi benchmarking + +/* + +Simple benchmark suite for png, stbi and qoi + +Requires libpng, "stb_image.h" and "stb_image_write.h" +Compile with: + gcc qoibench.c -std=gnu99 -lpng -O3 -o qoibench + +Dominic Szablewski - https://phoboslab.org + + +-- LICENSE: The MIT License(MIT) + +Copyright(c) 2021 Dominic Szablewski + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files(the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and / or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions : +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + +*/ + +#include +#include + +#define STB_IMAGE_IMPLEMENTATION +#define STBI_ONLY_PNG +#define STBI_NO_LINEAR +#include "stb_image.h" + +#define STB_IMAGE_WRITE_IMPLEMENTATION +#include "stb_image_write.h" + +#define QOI_IMPLEMENTATION +#include "qoi.h" + +#include "genc-bridge.h" + +// ----------------------------------------------------------------------------- +// Cross platform high resolution timer +// From https://gist.github.com/ForeverZer0/0a4f80fc02b96e19380ebb7a3debbee5 + +#include +#if defined(__linux) + #define HAVE_POSIX_TIMER + #include + #ifdef CLOCK_MONOTONIC + #define CLOCKID CLOCK_MONOTONIC + #else + #define CLOCKID CLOCK_REALTIME + #endif +#elif defined(__APPLE__) + #define HAVE_MACH_TIMER + #include +#elif defined(_WIN32) + #define WIN32_LEAN_AND_MEAN + #include +#endif + +static uint64_t ns() { + static uint64_t is_init = 0; +#if defined(__APPLE__) + static mach_timebase_info_data_t info; + if (0 == is_init) { + mach_timebase_info(&info); + is_init = 1; + } + uint64_t now; + now = mach_absolute_time(); + now *= info.numer; + now /= info.denom; + return now; +#elif defined(__linux) + static struct timespec linux_rate; + if (0 == is_init) { + clock_getres(CLOCKID, &linux_rate); + is_init = 1; + } + uint64_t now; + struct timespec spec; + clock_gettime(CLOCKID, &spec); + now = spec.tv_sec * 1.0e9 + spec.tv_nsec; + return now; +#elif defined(_WIN32) + static LARGE_INTEGER win_frequency; + if (0 == is_init) { + QueryPerformanceFrequency(&win_frequency); + is_init = 1; + } + LARGE_INTEGER now; + QueryPerformanceCounter(&now); + return (uint64_t) ((1e9 * now.QuadPart) / win_frequency.QuadPart); +#endif +} + +#define STRINGIFY(x) #x +#define TOSTRING(x) STRINGIFY(x) +#define ERROR(...) printf("abort at line " TOSTRING(__LINE__) ": " __VA_ARGS__); printf("\n"); exit(1) + +// ----------------------------------------------------------------------------- +// stb_image encode callback + +void stbi_write_callback(void *context, void *data, int size) { + int *encoded_size = (int *)context; + *encoded_size += size; + // In theory we'd need to do another malloc(), memcpy() and free() here to + // be fair to the other decode functions... +} + +// ----------------------------------------------------------------------------- +// function to load a whole file into memory + +void *fload(const char *path, int *out_size) { + FILE *fh = fopen(path, "rb"); + if (!fh) { + ERROR("Can't open file"); + } + + fseek(fh, 0, SEEK_END); + int size = ftell(fh); + fseek(fh, 0, SEEK_SET); + + void *buffer = malloc(size); + if (!buffer) { + ERROR("Malloc for %d bytes failed", size); + } + + if (!fread(buffer, size, 1, fh)) { + ERROR("Can't read file %s", path); + } + fclose(fh); + + *out_size = size; + return buffer; +} + + +// ----------------------------------------------------------------------------- +// benchmark runner + + +int opt_runs = 1; +int opt_nopng = 0; +int opt_nowarmup = 0; +int opt_noverify = 0; +int opt_nodecode = 0; +int opt_noencode = 0; +int opt_norecurse = 0; +int opt_onlytotals = 0; + + +typedef struct { + uint64_t size; + uint64_t encode_time; + uint64_t decode_time; +} benchmark_lib_result_t; + +typedef struct { + int count; + uint64_t raw_size; + uint64_t px; + int w; + int h; + benchmark_lib_result_t stbi; + benchmark_lib_result_t qoi; + benchmark_lib_result_t genc_qoi; +} benchmark_result_t; + + +void benchmark_print_result(benchmark_result_t res) { + res.px /= res.count; + res.raw_size /= res.count; + res.stbi.encode_time /= res.count; + res.stbi.decode_time /= res.count; + res.stbi.size /= res.count; + res.qoi.encode_time /= res.count; + res.qoi.decode_time /= res.count; + res.qoi.size /= res.count; + res.genc_qoi.encode_time /= res.count; + res.genc_qoi.decode_time /= res.count; + res.genc_qoi.size /= res.count; + + double px = res.px; + printf(" decode ms encode ms decode mpps encode mpps size kb rate\n"); + if (!opt_nopng) { + printf( + "stbi: %8.1f %8.1f %8.2f %8.2f %8ld %4.1f%%\n", + (double)res.stbi.decode_time/1000000.0, + (double)res.stbi.encode_time/1000000.0, + (res.stbi.decode_time > 0 ? px / ((double)res.stbi.decode_time/1000.0) : 0), + (res.stbi.encode_time > 0 ? px / ((double)res.stbi.encode_time/1000.0) : 0), + res.stbi.size/1024, + ((double)res.stbi.size/(double)res.raw_size) * 100.0 + ); + } + printf( + "qoi: %8.1f %8.1f %8.2f %8.2f %8ld %4.1f%%\n", + (double)res.qoi.decode_time/1000000.0, + (double)res.qoi.encode_time/1000000.0, + (res.qoi.decode_time > 0 ? px / ((double)res.qoi.decode_time/1000.0) : 0), + (res.qoi.encode_time > 0 ? px / ((double)res.qoi.encode_time/1000.0) : 0), + res.qoi.size/1024, + ((double)res.qoi.size/(double)res.raw_size) * 100.0 + ); + printf( + "genc_qoi: %8.1f %8.1f %8.2f %8.2f %8ld %4.1f%%\n", + (double)res.genc_qoi.decode_time/1000000.0, + (double)res.genc_qoi.encode_time/1000000.0, + (res.genc_qoi.decode_time > 0 ? px / ((double)res.genc_qoi.decode_time/1000.0) : 0), + (res.genc_qoi.encode_time > 0 ? px / ((double)res.genc_qoi.encode_time/1000.0) : 0), + res.genc_qoi.size/1024, + ((double)res.genc_qoi.size/(double)res.raw_size) * 100.0 + ); + printf("\n"); +} + +// Run __VA_ARGS__ a number of times and measure the time taken. The first +// run is ignored. +#define BENCHMARK_FN(NOWARMUP, RUNS, AVG_TIME, ...) \ + do { \ + uint64_t time = 0; \ + for (int i = NOWARMUP; i <= RUNS; i++) { \ + uint64_t time_start = ns(); \ + __VA_ARGS__ \ + uint64_t time_end = ns(); \ + if (i > 0) { \ + time += time_end - time_start; \ + } \ + } \ + AVG_TIME = time / RUNS; \ + } while (0) + + +benchmark_result_t benchmark_image(const char *path) { + int encoded_png_size; + int encoded_qoi_size; + int w; + int h; + int channels; + + // Load the encoded PNG, encoded QOI and raw pixels into memory + if(!stbi_info(path, &w, &h, &channels)) { + ERROR("Error decoding header %s", path); + } + + if (channels != 3) { + channels = 4; + } + + void *pixels = (void *)stbi_load(path, &w, &h, NULL, channels); + void *encoded_png = fload(path, &encoded_png_size); + void *encoded_qoi = qoi_encode(pixels, &(qoi_desc){ + .width = w, + .height = h, + .channels = channels, + .colorspace = QOI_SRGB + }, &encoded_qoi_size); + + if (!pixels || !encoded_qoi || !encoded_png) { + ERROR("Error decoding %s", path); + } + + // Verify QOI Output + + if (!opt_noverify) { + qoi_desc dc; + void *pixels_qoi = qoi_decode(encoded_qoi, encoded_qoi_size, &dc, channels); + if (memcmp(pixels, pixels_qoi, w * h * channels) != 0) { + ERROR("QOI roundtrip pixel mismatch for %s", path); + } + free(pixels_qoi); + } + + + + benchmark_result_t res = {0}; + res.count = 1; + res.raw_size = w * h * channels; + res.px = w * h; + res.w = w; + res.h = h; + + + // Decoding + + if (!opt_nodecode) { + if (!opt_nopng) { + BENCHMARK_FN(opt_nowarmup, opt_runs, res.stbi.decode_time, { + int dec_w, dec_h, dec_channels; + void *dec_p = stbi_load_from_memory(encoded_png, encoded_png_size, &dec_w, &dec_h, &dec_channels, 4); + free(dec_p); + }); + } + + BENCHMARK_FN(opt_nowarmup, opt_runs, res.qoi.decode_time, { + qoi_desc desc; + void *dec_p = qoi_decode(encoded_qoi, encoded_qoi_size, &desc, 4); + free(dec_p); + }); + BENCHMARK_FN(opt_nowarmup, opt_runs, res.genc_qoi.decode_time, { + void *dec_p = NULL; // In case genc_qoi_decode returns an error and does not write in dec_p. + int w, h, channels; + genc_qoi_decode(encoded_qoi, encoded_qoi_size, &dec_p, &w, &h, &channels); + free(dec_p); + }); + } + + + // Encoding + if (!opt_noencode) { + if (!opt_nopng) { + BENCHMARK_FN(opt_nowarmup, opt_runs, res.stbi.encode_time, { + int enc_size = 0; + stbi_write_png_to_func(stbi_write_callback, &enc_size, w, h, channels, pixels, 0); + res.stbi.size = enc_size; + }); + } + + BENCHMARK_FN(opt_nowarmup, opt_runs, res.qoi.encode_time, { + int enc_size; + void *enc_p = qoi_encode(pixels, &(qoi_desc){ + .width = w, + .height = h, + .channels = channels, + .colorspace = QOI_SRGB + }, &enc_size); + res.qoi.size = enc_size; + free(enc_p); + }); + BENCHMARK_FN(opt_nowarmup, opt_runs, res.genc_qoi.encode_time, { + void *enc_p = NULL; + size_t enc_size = 0; + genc_qoi_encode(pixels, w, h, channels, &enc_p, &enc_size); + res.genc_qoi.size = enc_size; + free(enc_p); + }); + } + + free(pixels); + free(encoded_png); + free(encoded_qoi); + + return res; +} + +void benchmark_directory(const char *path, benchmark_result_t *grand_total) { + DIR *dir = opendir(path); + if (!dir) { + ERROR("Couldn't open directory %s", path); + } + + struct dirent *file; + + if (!opt_norecurse) { + for (int i = 0; (file = readdir(dir)) != NULL; i++) { + if ( + file->d_type & DT_DIR && + strcmp(file->d_name, ".") != 0 && + strcmp(file->d_name, "..") != 0 + ) { + char subpath[1024]; + snprintf(subpath, 1024, "%s/%s", path, file->d_name); + benchmark_directory(subpath, grand_total); + } + } + rewinddir(dir); + } + + benchmark_result_t dir_total = {0}; + + int has_shown_head = 0; + for (int i = 0; (file = readdir(dir)) != NULL; i++) { + if (strcmp(file->d_name + strlen(file->d_name) - 4, ".png") != 0) { + continue; + } + + if (!has_shown_head) { + has_shown_head = 1; + printf("## Benchmarking %s/*.png -- %d runs\n\n", path, opt_runs); + } + + char *file_path = malloc(strlen(file->d_name) + strlen(path)+8); + sprintf(file_path, "%s/%s", path, file->d_name); + + benchmark_result_t res = benchmark_image(file_path); + + if (!opt_onlytotals) { + printf("## %s size: %dx%d\n", file_path, res.w, res.h); + benchmark_print_result(res); + } + + free(file_path); + + dir_total.count++; + dir_total.raw_size += res.raw_size; + dir_total.px += res.px; + dir_total.stbi.encode_time += res.stbi.encode_time; + dir_total.stbi.decode_time += res.stbi.decode_time; + dir_total.stbi.size += res.stbi.size; + dir_total.qoi.encode_time += res.qoi.encode_time; + dir_total.qoi.decode_time += res.qoi.decode_time; + dir_total.qoi.size += res.qoi.size; + dir_total.genc_qoi.encode_time += res.genc_qoi.encode_time; + dir_total.genc_qoi.decode_time += res.genc_qoi.decode_time; + dir_total.genc_qoi.size += res.genc_qoi.size; + + grand_total->count++; + grand_total->raw_size += res.raw_size; + grand_total->px += res.px; + grand_total->stbi.encode_time += res.stbi.encode_time; + grand_total->stbi.decode_time += res.stbi.decode_time; + grand_total->stbi.size += res.stbi.size; + grand_total->qoi.encode_time += res.qoi.encode_time; + grand_total->qoi.decode_time += res.qoi.decode_time; + grand_total->qoi.size += res.qoi.size; + grand_total->genc_qoi.encode_time += res.genc_qoi.encode_time; + grand_total->genc_qoi.decode_time += res.genc_qoi.decode_time; + grand_total->genc_qoi.size += res.genc_qoi.size; + } + closedir(dir); + + if (dir_total.count > 0) { + printf("## Total for %s\n", path); + benchmark_print_result(dir_total); + } +} + +int main(int argc, char **argv) { + if (argc < 3) { + printf("Usage: genc-qoibench [options]\n"); + printf("Options:\n"); + printf(" --nowarmup ... don't perform a warmup run\n"); + printf(" --nopng ...... don't run png encode/decode\n"); + printf(" --noverify ... don't verify qoi roundtrip\n"); + printf(" --noencode ... don't run encoders\n"); + printf(" --nodecode ... don't run decoders\n"); + printf(" --norecurse .. don't descend into directories\n"); + printf(" --onlytotals . don't print individual image results\n"); + printf("Examples\n"); + printf(" qoibench 10 images/textures/\n"); + printf(" qoibench 1 images/textures/ --nopng --nowarmup\n"); + exit(1); + } + + for (int i = 3; i < argc; i++) { + if (strcmp(argv[i], "--nowarmup") == 0) { opt_nowarmup = 1; } + else if (strcmp(argv[i], "--nopng") == 0) { opt_nopng = 1; } + else if (strcmp(argv[i], "--noverify") == 0) { opt_noverify = 1; } + else if (strcmp(argv[i], "--noencode") == 0) { opt_noencode = 1; } + else if (strcmp(argv[i], "--nodecode") == 0) { opt_nodecode = 1; } + else if (strcmp(argv[i], "--norecurse") == 0) { opt_norecurse = 1; } + else if (strcmp(argv[i], "--onlytotals") == 0) { opt_onlytotals = 1; } + else { ERROR("Unknown option %s", argv[i]); } + } + + opt_runs = atoi(argv[1]); + if (opt_runs <=0) { + ERROR("Invalid number of runs %d", opt_runs); + } + + benchmark_result_t grand_total = {0}; + benchmark_directory(argv[2], &grand_total); + + if (grand_total.count > 0) { + printf("# Grand total for %s\n", argv[2]); + benchmark_print_result(grand_total); + } + else { + printf("No images found in %s\n", argv[2]); + } + + return 0; +} diff --git a/qoi/genc/genc-qoiconv.c b/qoi/genc/genc-qoiconv.c new file mode 100644 index 00000000..9e04b932 --- /dev/null +++ b/qoi/genc/genc-qoiconv.c @@ -0,0 +1,99 @@ +// Tweaked version of https://github.com/phoboslab/qoi/blob/master/qoiconv.c +// where we call genc_qoi_read and genc_qoi_write instead of qoi_read and qoi_write + +/* + +Command line tool to convert between png <> qoi format + +Requires "stb_image.h" and "stb_image_write.h" +Compile with: + gcc qoiconv.c -std=c99 -O3 -o qoiconv + +Dominic Szablewski - https://phoboslab.org + + +-- LICENSE: The MIT License(MIT) + +Copyright(c) 2021 Dominic Szablewski + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files(the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and / or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions : +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + +*/ + + +#define STB_IMAGE_IMPLEMENTATION +#define STBI_ONLY_PNG +#define STBI_NO_LINEAR +#include "stb_image.h" + +#define STB_IMAGE_WRITE_IMPLEMENTATION +#include "stb_image_write.h" + +#include "genc-bridge.h" + +#define STR_ENDS_WITH(S, E) (strcmp(S + strlen(S) - (sizeof(E)-1), E) == 0) + + +int main(int argc, char **argv) { + if (argc < 3) { + puts("Usage: genc-qoiconv "); + puts("Examples:"); + puts(" genc-qoiconv input.png output.qoi"); + puts(" genc-qoiconv input.qoi output.png"); + exit(1); + } + + void *pixels = NULL; + int w, h, channels; + if (STR_ENDS_WITH(argv[1], ".png")) { + if(!stbi_info(argv[1], &w, &h, &channels)) { + printf("Couldn't read header %s\n", argv[1]); + exit(1); + } + + // Force all odd encodings to be RGBA + if(channels != 3) { + channels = 4; + } + + pixels = (void *)stbi_load(argv[1], &w, &h, NULL, channels); + } + else if (STR_ENDS_WITH(argv[1], ".qoi")) { + genc_qoi_read(argv[1], &pixels, &w, &h, &channels); + } + + if (pixels == NULL) { + printf("Couldn't load/decode %s\n", argv[1]); + exit(1); + } + + int encoded = 0; + if (STR_ENDS_WITH(argv[2], ".png")) { + encoded = stbi_write_png(argv[2], w, h, channels, pixels, 0); + } + else if (STR_ENDS_WITH(argv[2], ".qoi")) { + encoded = genc_qoi_write(argv[2], pixels, w, h, channels); + } + + if (!encoded) { + printf("Couldn't write/encode %s\n", argv[2]); + exit(1); + } + + free(pixels); + return 0; +} diff --git a/qoi/genc/qoi.h b/qoi/genc/qoi.h new file mode 100644 index 00000000..988f9edc --- /dev/null +++ b/qoi/genc/qoi.h @@ -0,0 +1,671 @@ +/* + +QOI - The "Quite OK Image" format for fast, lossless image compression + +Dominic Szablewski - https://phoboslab.org + + +-- LICENSE: The MIT License(MIT) + +Copyright(c) 2021 Dominic Szablewski + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files(the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and / or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions : +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + + +-- About + +QOI encodes and decodes images in a lossless format. Compared to stb_image and +stb_image_write QOI offers 20x-50x faster encoding, 3x-4x faster decoding and +20% better compression. + + +-- Synopsis + +// Define `QOI_IMPLEMENTATION` in *one* C/C++ file before including this +// library to create the implementation. + +#define QOI_IMPLEMENTATION +#include "qoi.h" + +// Encode and store an RGBA buffer to the file system. The qoi_desc describes +// the input pixel data. +qoi_write("image_new.qoi", rgba_pixels, &(qoi_desc){ + .width = 1920, + .height = 1080, + .channels = 4, + .colorspace = QOI_SRGB +}); + +// Load and decode a QOI image from the file system into a 32bbp RGBA buffer. +// The qoi_desc struct will be filled with the width, height, number of channels +// and colorspace read from the file header. +qoi_desc desc; +void *rgba_pixels = qoi_read("image.qoi", &desc, 4); + + + +-- Documentation + +This library provides the following functions; +- qoi_read -- read and decode a QOI file +- qoi_decode -- decode the raw bytes of a QOI image from memory +- qoi_write -- encode and write a QOI file +- qoi_encode -- encode an rgba buffer into a QOI image in memory + +See the function declaration below for the signature and more information. + +If you don't want/need the qoi_read and qoi_write functions, you can define +QOI_NO_STDIO before including this library. + +This library uses malloc() and free(). To supply your own malloc implementation +you can define QOI_MALLOC and QOI_FREE before including this library. + +This library uses memset() to zero-initialize the index. To supply your own +implementation you can define QOI_ZEROARR before including this library. + + +-- Data Format + +A QOI file has a 14 byte header, followed by any number of data "chunks" and an +8-byte end marker. + +struct qoi_header_t { + char magic[4]; // magic bytes "qoif" + uint32_t width; // image width in pixels (BE) + uint32_t height; // image height in pixels (BE) + uint8_t channels; // 3 = RGB, 4 = RGBA + uint8_t colorspace; // 0 = sRGB with linear alpha, 1 = all channels linear +}; + +Images are encoded row by row, left to right, top to bottom. The decoder and +encoder start with {r: 0, g: 0, b: 0, a: 255} as the previous pixel value. An +image is complete when all pixels specified by width * height have been covered. + +Pixels are encoded as + - a run of the previous pixel + - an index into an array of previously seen pixels + - a difference to the previous pixel value in r,g,b + - full r,g,b or r,g,b,a values + +The color channels are assumed to not be premultiplied with the alpha channel +("un-premultiplied alpha"). + +A running array[64] (zero-initialized) of previously seen pixel values is +maintained by the encoder and decoder. Each pixel that is seen by the encoder +and decoder is put into this array at the position formed by a hash function of +the color value. In the encoder, if the pixel value at the index matches the +current pixel, this index position is written to the stream as QOI_OP_INDEX. +The hash function for the index is: + + index_position = (r * 3 + g * 5 + b * 7 + a * 11) % 64 + +Each chunk starts with a 2- or 8-bit tag, followed by a number of data bits. The +bit length of chunks is divisible by 8 - i.e. all chunks are byte aligned. All +values encoded in these data bits have the most significant bit on the left. + +The 8-bit tags have precedence over the 2-bit tags. A decoder must check for the +presence of an 8-bit tag first. + +The byte stream's end is marked with 7 0x00 bytes followed a single 0x01 byte. + + +The possible chunks are: + + +.- QOI_OP_INDEX ----------. +| Byte[0] | +| 7 6 5 4 3 2 1 0 | +|-------+-----------------| +| 0 0 | index | +`-------------------------` +2-bit tag b00 +6-bit index into the color index array: 0..63 + +A valid encoder must not issue 2 or more consecutive QOI_OP_INDEX chunks to the +same index. QOI_OP_RUN should be used instead. + + +.- QOI_OP_DIFF -----------. +| Byte[0] | +| 7 6 5 4 3 2 1 0 | +|-------+-----+-----+-----| +| 0 1 | dr | dg | db | +`-------------------------` +2-bit tag b01 +2-bit red channel difference from the previous pixel between -2..1 +2-bit green channel difference from the previous pixel between -2..1 +2-bit blue channel difference from the previous pixel between -2..1 + +The difference to the current channel values are using a wraparound operation, +so "1 - 2" will result in 255, while "255 + 1" will result in 0. + +Values are stored as unsigned integers with a bias of 2. E.g. -2 is stored as +0 (b00). 1 is stored as 3 (b11). + +The alpha value remains unchanged from the previous pixel. + + +.- QOI_OP_LUMA -------------------------------------. +| Byte[0] | Byte[1] | +| 7 6 5 4 3 2 1 0 | 7 6 5 4 3 2 1 0 | +|-------+-----------------+-------------+-----------| +| 1 0 | green diff | dr - dg | db - dg | +`---------------------------------------------------` +2-bit tag b10 +6-bit green channel difference from the previous pixel -32..31 +4-bit red channel difference minus green channel difference -8..7 +4-bit blue channel difference minus green channel difference -8..7 + +The green channel is used to indicate the general direction of change and is +encoded in 6 bits. The red and blue channels (dr and db) base their diffs off +of the green channel difference and are encoded in 4 bits. I.e.: + dr_dg = (cur_px.r - prev_px.r) - (cur_px.g - prev_px.g) + db_dg = (cur_px.b - prev_px.b) - (cur_px.g - prev_px.g) + +The difference to the current channel values are using a wraparound operation, +so "10 - 13" will result in 253, while "250 + 7" will result in 1. + +Values are stored as unsigned integers with a bias of 32 for the green channel +and a bias of 8 for the red and blue channel. + +The alpha value remains unchanged from the previous pixel. + + +.- QOI_OP_RUN ------------. +| Byte[0] | +| 7 6 5 4 3 2 1 0 | +|-------+-----------------| +| 1 1 | run | +`-------------------------` +2-bit tag b11 +6-bit run-length repeating the previous pixel: 1..62 + +The run-length is stored with a bias of -1. Note that the run-lengths 63 and 64 +(b111110 and b111111) are illegal as they are occupied by the QOI_OP_RGB and +QOI_OP_RGBA tags. + + +.- QOI_OP_RGB ------------------------------------------. +| Byte[0] | Byte[1] | Byte[2] | Byte[3] | +| 7 6 5 4 3 2 1 0 | 7 .. 0 | 7 .. 0 | 7 .. 0 | +|-------------------------+---------+---------+---------| +| 1 1 1 1 1 1 1 0 | red | green | blue | +`-------------------------------------------------------` +8-bit tag b11111110 +8-bit red channel value +8-bit green channel value +8-bit blue channel value + +The alpha value remains unchanged from the previous pixel. + + +.- QOI_OP_RGBA ---------------------------------------------------. +| Byte[0] | Byte[1] | Byte[2] | Byte[3] | Byte[4] | +| 7 6 5 4 3 2 1 0 | 7 .. 0 | 7 .. 0 | 7 .. 0 | 7 .. 0 | +|-------------------------+---------+---------+---------+---------| +| 1 1 1 1 1 1 1 1 | red | green | blue | alpha | +`-----------------------------------------------------------------` +8-bit tag b11111111 +8-bit red channel value +8-bit green channel value +8-bit blue channel value +8-bit alpha channel value + +*/ + + +/* ----------------------------------------------------------------------------- +Header - Public functions */ + +#ifndef QOI_H +#define QOI_H + +#ifdef __cplusplus +extern "C" { +#endif + +/* A pointer to a qoi_desc struct has to be supplied to all of qoi's functions. +It describes either the input format (for qoi_write and qoi_encode), or is +filled with the description read from the file header (for qoi_read and +qoi_decode). + +The colorspace in this qoi_desc is an enum where + 0 = sRGB, i.e. gamma scaled RGB channels and a linear alpha channel + 1 = all channels are linear +You may use the constants QOI_SRGB or QOI_LINEAR. The colorspace is purely +informative. It will be saved to the file header, but does not affect +how chunks are en-/decoded. */ + +#define QOI_SRGB 0 +#define QOI_LINEAR 1 + +typedef struct { + unsigned int width; + unsigned int height; + unsigned char channels; + unsigned char colorspace; +} qoi_desc; + +#ifndef QOI_NO_STDIO + +/* Encode raw RGB or RGBA pixels into a QOI image and write it to the file +system. The qoi_desc struct must be filled with the image width, height, +number of channels (3 = RGB, 4 = RGBA) and the colorspace. + +The function returns 0 on failure (invalid parameters, or fopen or malloc +failed) or the number of bytes written on success. */ + +int qoi_write(const char *filename, const void *data, const qoi_desc *desc); + + +/* Read and decode a QOI image from the file system. If channels is 0, the +number of channels from the file header is used. If channels is 3 or 4 the +output format will be forced into this number of channels. + +The function either returns NULL on failure (invalid data, or malloc or fopen +failed) or a pointer to the decoded pixels. On success, the qoi_desc struct +will be filled with the description from the file header. + +The returned pixel data should be free()d after use. */ + +void *qoi_read(const char *filename, qoi_desc *desc, int channels); + +#endif /* QOI_NO_STDIO */ + + +/* Encode raw RGB or RGBA pixels into a QOI image in memory. + +The function either returns NULL on failure (invalid parameters or malloc +failed) or a pointer to the encoded data on success. On success the out_len +is set to the size in bytes of the encoded data. + +The returned qoi data should be free()d after use. */ + +void *qoi_encode(const void *data, const qoi_desc *desc, int *out_len); + + +/* Decode a QOI image from memory. + +The function either returns NULL on failure (invalid parameters or malloc +failed) or a pointer to the decoded pixels. On success, the qoi_desc struct +is filled with the description from the file header. + +The returned pixel data should be free()d after use. */ + +void *qoi_decode(const void *data, int size, qoi_desc *desc, int channels); + + +#ifdef __cplusplus +} +#endif +#endif /* QOI_H */ + + +/* ----------------------------------------------------------------------------- +Implementation */ + +#ifdef QOI_IMPLEMENTATION +#include +#include + +#ifndef QOI_MALLOC + #define QOI_MALLOC(sz) malloc(sz) + #define QOI_FREE(p) free(p) +#endif +#ifndef QOI_ZEROARR + #define QOI_ZEROARR(a) memset((a),0,sizeof(a)) +#endif + +#define QOI_OP_INDEX 0x00 /* 00xxxxxx */ +#define QOI_OP_DIFF 0x40 /* 01xxxxxx */ +#define QOI_OP_LUMA 0x80 /* 10xxxxxx */ +#define QOI_OP_RUN 0xc0 /* 11xxxxxx */ +#define QOI_OP_RGB 0xfe /* 11111110 */ +#define QOI_OP_RGBA 0xff /* 11111111 */ + +#define QOI_MASK_2 0xc0 /* 11000000 */ + +#define QOI_COLOR_HASH(C) (C.rgba.r*3 + C.rgba.g*5 + C.rgba.b*7 + C.rgba.a*11) +#define QOI_MAGIC \ + (((unsigned int)'q') << 24 | ((unsigned int)'o') << 16 | \ + ((unsigned int)'i') << 8 | ((unsigned int)'f')) +#define QOI_HEADER_SIZE 14 + +/* 2GB is the max file size that this implementation can safely handle. We guard +against anything larger than that, assuming the worst case with 5 bytes per +pixel, rounded down to a nice clean value. 400 million pixels ought to be +enough for anybody. */ +#define QOI_PIXELS_MAX ((unsigned int)400000000) + +typedef union { + struct { unsigned char r, g, b, a; } rgba; + unsigned int v; +} qoi_rgba_t; + +static const unsigned char qoi_padding[8] = {0,0,0,0,0,0,0,1}; + +static void qoi_write_32(unsigned char *bytes, int *p, unsigned int v) { + bytes[(*p)++] = (0xff000000 & v) >> 24; + bytes[(*p)++] = (0x00ff0000 & v) >> 16; + bytes[(*p)++] = (0x0000ff00 & v) >> 8; + bytes[(*p)++] = (0x000000ff & v); +} + +static unsigned int qoi_read_32(const unsigned char *bytes, int *p) { + unsigned int a = bytes[(*p)++]; + unsigned int b = bytes[(*p)++]; + unsigned int c = bytes[(*p)++]; + unsigned int d = bytes[(*p)++]; + return a << 24 | b << 16 | c << 8 | d; +} + +void *qoi_encode(const void *data, const qoi_desc *desc, int *out_len) { + int i, max_size, p, run; + int px_len, px_end, px_pos, channels; + unsigned char *bytes; + const unsigned char *pixels; + qoi_rgba_t index[64]; + qoi_rgba_t px, px_prev; + + if ( + data == NULL || out_len == NULL || desc == NULL || + desc->width == 0 || desc->height == 0 || + desc->channels < 3 || desc->channels > 4 || + desc->colorspace > 1 || + desc->height >= QOI_PIXELS_MAX / desc->width + ) { + return NULL; + } + + max_size = + desc->width * desc->height * (desc->channels + 1) + + QOI_HEADER_SIZE + sizeof(qoi_padding); + + p = 0; + bytes = (unsigned char *) QOI_MALLOC(max_size); + if (!bytes) { + return NULL; + } + + qoi_write_32(bytes, &p, QOI_MAGIC); + qoi_write_32(bytes, &p, desc->width); + qoi_write_32(bytes, &p, desc->height); + bytes[p++] = desc->channels; + bytes[p++] = desc->colorspace; + + + pixels = (const unsigned char *)data; + + QOI_ZEROARR(index); + + run = 0; + px_prev.rgba.r = 0; + px_prev.rgba.g = 0; + px_prev.rgba.b = 0; + px_prev.rgba.a = 255; + px = px_prev; + + px_len = desc->width * desc->height * desc->channels; + px_end = px_len - desc->channels; + channels = desc->channels; + + for (px_pos = 0; px_pos < px_len; px_pos += channels) { + if (channels == 4) { + px = *(qoi_rgba_t *)(pixels + px_pos); + } + else { + px.rgba.r = pixels[px_pos + 0]; + px.rgba.g = pixels[px_pos + 1]; + px.rgba.b = pixels[px_pos + 2]; + } + + if (px.v == px_prev.v) { + run++; + if (run == 62 || px_pos == px_end) { + bytes[p++] = QOI_OP_RUN | (run - 1); + run = 0; + } + } + else { + int index_pos; + + if (run > 0) { + bytes[p++] = QOI_OP_RUN | (run - 1); + run = 0; + } + + index_pos = QOI_COLOR_HASH(px) % 64; + + if (index[index_pos].v == px.v) { + bytes[p++] = QOI_OP_INDEX | index_pos; + } + else { + index[index_pos] = px; + + if (px.rgba.a == px_prev.rgba.a) { + signed char vr = px.rgba.r - px_prev.rgba.r; + signed char vg = px.rgba.g - px_prev.rgba.g; + signed char vb = px.rgba.b - px_prev.rgba.b; + + signed char vg_r = vr - vg; + signed char vg_b = vb - vg; + + if ( + vr > -3 && vr < 2 && + vg > -3 && vg < 2 && + vb > -3 && vb < 2 + ) { + bytes[p++] = QOI_OP_DIFF | (vr + 2) << 4 | (vg + 2) << 2 | (vb + 2); + } + else if ( + vg_r > -9 && vg_r < 8 && + vg > -33 && vg < 32 && + vg_b > -9 && vg_b < 8 + ) { + bytes[p++] = QOI_OP_LUMA | (vg + 32); + bytes[p++] = (vg_r + 8) << 4 | (vg_b + 8); + } + else { + bytes[p++] = QOI_OP_RGB; + bytes[p++] = px.rgba.r; + bytes[p++] = px.rgba.g; + bytes[p++] = px.rgba.b; + } + } + else { + bytes[p++] = QOI_OP_RGBA; + bytes[p++] = px.rgba.r; + bytes[p++] = px.rgba.g; + bytes[p++] = px.rgba.b; + bytes[p++] = px.rgba.a; + } + } + } + px_prev = px; + } + + for (i = 0; i < (int)sizeof(qoi_padding); i++) { + bytes[p++] = qoi_padding[i]; + } + + *out_len = p; + return bytes; +} + +void *qoi_decode(const void *data, int size, qoi_desc *desc, int channels) { + const unsigned char *bytes; + unsigned int header_magic; + unsigned char *pixels; + qoi_rgba_t index[64]; + qoi_rgba_t px; + int px_len, chunks_len, px_pos; + int p = 0, run = 0; + + if ( + data == NULL || desc == NULL || + (channels != 0 && channels != 3 && channels != 4) || + size < QOI_HEADER_SIZE + (int)sizeof(qoi_padding) + ) { + return NULL; + } + + bytes = (const unsigned char *)data; + + header_magic = qoi_read_32(bytes, &p); + desc->width = qoi_read_32(bytes, &p); + desc->height = qoi_read_32(bytes, &p); + desc->channels = bytes[p++]; + desc->colorspace = bytes[p++]; + + if ( + desc->width == 0 || desc->height == 0 || + desc->channels < 3 || desc->channels > 4 || + desc->colorspace > 1 || + header_magic != QOI_MAGIC || + desc->height >= QOI_PIXELS_MAX / desc->width + ) { + return NULL; + } + + if (channels == 0) { + channels = desc->channels; + } + + px_len = desc->width * desc->height * channels; + pixels = (unsigned char *) QOI_MALLOC(px_len); + if (!pixels) { + return NULL; + } + + QOI_ZEROARR(index); + px.rgba.r = 0; + px.rgba.g = 0; + px.rgba.b = 0; + px.rgba.a = 255; + + chunks_len = size - (int)sizeof(qoi_padding); + for (px_pos = 0; px_pos < px_len; px_pos += channels) { + if (run > 0) { + run--; + } + else if (p < chunks_len) { + int b1 = bytes[p++]; + + if (b1 == QOI_OP_RGB) { + px.rgba.r = bytes[p++]; + px.rgba.g = bytes[p++]; + px.rgba.b = bytes[p++]; + } + else if (b1 == QOI_OP_RGBA) { + px.rgba.r = bytes[p++]; + px.rgba.g = bytes[p++]; + px.rgba.b = bytes[p++]; + px.rgba.a = bytes[p++]; + } + else if ((b1 & QOI_MASK_2) == QOI_OP_INDEX) { + px = index[b1]; + } + else if ((b1 & QOI_MASK_2) == QOI_OP_DIFF) { + px.rgba.r += ((b1 >> 4) & 0x03) - 2; + px.rgba.g += ((b1 >> 2) & 0x03) - 2; + px.rgba.b += ( b1 & 0x03) - 2; + } + else if ((b1 & QOI_MASK_2) == QOI_OP_LUMA) { + int b2 = bytes[p++]; + int vg = (b1 & 0x3f) - 32; + px.rgba.r += vg - 8 + ((b2 >> 4) & 0x0f); + px.rgba.g += vg; + px.rgba.b += vg - 8 + (b2 & 0x0f); + } + else if ((b1 & QOI_MASK_2) == QOI_OP_RUN) { + run = (b1 & 0x3f); + } + + index[QOI_COLOR_HASH(px) % 64] = px; + } + + if (channels == 4) { + *(qoi_rgba_t*)(pixels + px_pos) = px; + } + else { + pixels[px_pos + 0] = px.rgba.r; + pixels[px_pos + 1] = px.rgba.g; + pixels[px_pos + 2] = px.rgba.b; + } + } + + return pixels; +} + +#ifndef QOI_NO_STDIO +#include + +int qoi_write(const char *filename, const void *data, const qoi_desc *desc) { + FILE *f = fopen(filename, "wb"); + int size; + void *encoded; + + if (!f) { + return 0; + } + + encoded = qoi_encode(data, desc, &size); + if (!encoded) { + fclose(f); + return 0; + } + + fwrite(encoded, 1, size, f); + fclose(f); + + QOI_FREE(encoded); + return size; +} + +void *qoi_read(const char *filename, qoi_desc *desc, int channels) { + FILE *f = fopen(filename, "rb"); + int size, bytes_read; + void *pixels, *data; + + if (!f) { + return NULL; + } + + fseek(f, 0, SEEK_END); + size = ftell(f); + if (size <= 0) { + fclose(f); + return NULL; + } + fseek(f, 0, SEEK_SET); + + data = QOI_MALLOC(size); + if (!data) { + fclose(f); + return NULL; + } + + bytes_read = fread(data, 1, size, f); + fclose(f); + + pixels = qoi_decode(data, bytes_read, desc, channels); + QOI_FREE(data); + return pixels; +} + +#endif /* QOI_NO_STDIO */ +#endif /* QOI_IMPLEMENTATION */ diff --git a/qoi/genc/run-bench.sh b/qoi/genc/run-bench.sh new file mode 100755 index 00000000..de7812f9 --- /dev/null +++ b/qoi/genc/run-bench.sh @@ -0,0 +1,4 @@ +#!/bin/bash + +make genc-qoibench +./genc-qoibench 3 ../images \ No newline at end of file diff --git a/qoi/genc/stb_image.h b/qoi/genc/stb_image.h new file mode 100644 index 00000000..d60371b9 --- /dev/null +++ b/qoi/genc/stb_image.h @@ -0,0 +1,7897 @@ +/* stb_image - v2.27 - public domain image loader - http://nothings.org/stb + 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 1/2/4/8/16-bit-per-channel + + 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. + + +LICENSE + + See end of file for license information. + +RECENT REVISION HISTORY: + + 2.27 (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes + 2.26 (2020-07-13) many minor fixes + 2.25 (2020-02-02) fix warnings + 2.24 (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically + 2.23 (2019-08-11) fix clang static analysis warning + 2.22 (2019-03-04) gif fixes, fix warnings + 2.21 (2019-02-25) fix typo in comment + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings + 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes + 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64 + RGB-format JPEG; remove white matting in PSD; + allocate large structures on the stack; + correct channel count for PNG & BMP + 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 + + 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) + github:urraka (animated gif) Junggon Kim (PNM comments) + Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA) + socks-the-fox (16-bit PNG) + Jeremy Sawicki (handle all ImageNet JPGs) + Optimizations & bugfixes Mikhail Morozov (1-bit BMP) + Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query) + Arseny Kapoulkine Simon Breuss (16-bit PNM) + John-Mark Allen + Carmelo J Fdez-Aguera + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Jerry Jansson Joseph Thomson Blazej Dariusz Roszkowski + Phil Jordan Dave Moore Roy Eltham + Hayaki Saito Nathan Reed Won Chun + Luke Graham Johan Duparc Nick Verigakis the Horde3D community + Thomas Ruf Ronny Chevalier github:rlyeh + Janez Zemva John Bartholomew Michal Cichon github:romigrou + Jonathan Blow Ken Hamada Tero Hanninen github:svdijk + Eugene Golushkov Laurent Gomila Cort Stratton github:snagar + Aruelien Pocheville Sergio Gonzalez Thibault Reuille github:Zelex + Cass Everitt Ryamond Barbiero github:grim210 + Paul Du Bois Engin Manap Aldo Culquicondor github:sammyhw + Philipp Wiesemann Dale Weiler Oriol Ferrer Mesia github:phprus + Josh Tobin Matthew Gregan github:poppolopoppo + Julian Raschke Gregory Mullen Christian Floisand github:darealshinji + Baldur Karlsson Kevin Schmidt JR Smith github:Michaelangel007 + Brad Weinberger Matvey Cherevko github:mosra + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + Jacko Dirks + + To add your name to the credits, pick a random blank space in the middle and fill it. + 80% of merge conflicts on stb PRs are due to people adding their name at the end + of the credits. +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - 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 *channels_in_file -- outputs # of image components in image file +// int desired_channels -- 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 'desired_channels' if desired_channels is non-zero, or +// *channels_in_file otherwise. If desired_channels is non-zero, +// *channels_in_file has the number of components that _would_ have been +// output otherwise. E.g. if you set desired_channels to 4, you will always +// get RGBA output, but you can check *channels_in_file 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, *channels_in_file 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. +// +// To query the width, height and component count of an image without having to +// decode the full file, you can use the stbi_info family of functions: +// +// int x,y,n,ok; +// ok = stbi_info(filename, &x, &y, &n); +// // returns ok=1 and sets x, y, n if image is a supported format, +// // 0 otherwise. +// +// Note that stb_image pervasively uses ints in its public API for sizes, +// including sizes of memory buffers. This is now part of the API and thus +// hard to change without causing breakage. As a result, the various image +// loaders all have certain limits on image size; these differ somewhat +// by format but generally boil down to either just under 2GB or just under +// 1GB. When the decoded image would be larger than this, stb_image decoding +// will fail. +// +// Additionally, stb_image will reject image files that have any of their +// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS, +// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit, +// the only way to have an image with such dimensions load correctly +// is for it to have a rather extreme aspect ratio. Either way, the +// assumption here is that such larger images are likely to be malformed +// or malicious. If you do need to load an image with individual dimensions +// larger than that, and it still fits in the overall size limit, you can +// #define STBI_MAX_DIMENSIONS on your own to be something larger. +// +// =========================================================================== +// +// UNICODE: +// +// If compiling for Windows and you wish to use Unicode filenames, compile +// with +// #define STBI_WINDOWS_UTF8 +// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert +// Windows wchar_t filenames to utf8. +// +// =========================================================================== +// +// 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 +// provide more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small source code 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. +// +// 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 supports loading HDR images in general, and currently the Radiance +// .HDR file format specifically. 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: +// +// We optionally support converting iPhone-formatted PNGs (which store +// premultiplied BGRA) back to RGB, even though they're internally encoded +// differently. To enable this conversion, call +// stbi_convert_iphone_png_to_rgb(1). +// +// 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). +// +// =========================================================================== +// +// ADDITIONAL CONFIGURATION +// +// - 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) +// +// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still +// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB +// +// - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater +// than that size (in either width or height) without further processing. +// This is to let programs in the wild set an upper bound to prevent +// denial-of-service attacks on untrusted data, as one could generate a +// valid image of gigantic dimensions and force stb_image to allocate a +// huge block of memory and spend disproportionate time decoding it. By +// default this is set to (1 << 24), which is 16777216, but that's still +// very big. + +#ifndef STBI_NO_STDIO +#include +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for desired_channels + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +#include +typedef unsigned char stbi_uc; +typedef unsigned short stbi_us; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef STBIDEF +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif +#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; + +//////////////////////////////////// +// +// 8-bits-per-channel interface +// + +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +#endif + +#ifdef STBI_WINDOWS_UTF8 +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif + +//////////////////////////////////// +// +// 16-bits-per-channel interface +// + +STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +#endif + +//////////////////////////////////// +// +// float-per-channel interface +// +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); + #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 +// on most compilers (and ALL modern mainstream compilers) this is 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); +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len); +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user); + +#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); +STBIDEF int stbi_is_16_bit (char const *filename); +STBIDEF int stbi_is_16_bit_from_file(FILE *f); +#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); + +// as above, but only applies to images loaded on the thread that calls the function +// this function is only available if your compiler supports thread-local variables; +// calling it will fail to link if your compiler doesn't +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply); +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert); +STBIDEF void stbi_set_flip_vertically_on_load_thread(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 +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include // ldexp, pow +#endif + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_ASSERT +#include +#define STBI_ASSERT(x) assert(x) +#endif + +#ifdef __cplusplus +#define STBI_EXTERN extern "C" +#else +#define STBI_EXTERN extern +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + +#ifndef STBI_NO_THREAD_LOCALS + #if defined(__cplusplus) && __cplusplus >= 201103L + #define STBI_THREAD_LOCAL thread_local + #elif defined(__GNUC__) && __GNUC__ < 5 + #define STBI_THREAD_LOCAL __thread + #elif defined(_MSC_VER) + #define STBI_THREAD_LOCAL __declspec(thread) + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__) + #define STBI_THREAD_LOCAL _Thread_local + #endif + + #ifndef STBI_THREAD_LOCAL + #if defined(__GNUC__) + #define STBI_THREAD_LOCAL __thread + #endif + #endif +#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) >> (-(y) & 31))) +#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(__SSE2__) && !defined(STBI_NO_SIMD) +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// which in turn means it gets to use SSE2 everywhere. This is unfortunate, +// but previous attempts to provide the SSE2 functions with runtime +// detection caused numerous issues. The way architecture extensions are +// exposed in GCC/Clang is, sadly, not really suited for one-file libs. +// New behavior: if compiled with -msse2, we use SSE2 without any +// detection; if not, we don't use it at all. +#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) || defined(STBI__X64_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 + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#endif + +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + // If we're even attempting to compile this on GCC/Clang, that means + // -msse2 is on, which means the compiler is allowed to use SSE2 + // instructions at will, and so are we. + return 1; +} +#endif + +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +#ifdef _MSC_VER +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name +#else +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#ifndef STBI_MAX_DIMENSIONS +#define STBI_MAX_DIMENSIONS (1 << 24) +#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]; + int callback_already_read; + + 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->callback_already_read = 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->callback_already_read = 0; + s->img_buffer = 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) +{ + int ch; + fseek((FILE*) user, n, SEEK_CUR); + ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */ + if (ch != EOF) { + ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */ + } +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user) || ferror((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; +} + +enum +{ + STBI_ORDER_RGB, + STBI_ORDER_BGR +}; + +typedef struct +{ + int bits_per_channel; + int num_channels; + int channel_order; +} stbi__result_info; + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +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 void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__png_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +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 void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +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 void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__psd_is16(stbi__context *s); +#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, stbi__result_info *ri); +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 void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +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 void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, 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 void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__pnm_is16(stbi__context *s); +#endif + +static +#ifdef STBI_THREAD_LOCAL +STBI_THREAD_LOCAL +#endif +const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +#ifndef STBI_NO_FAILURE_STRINGS +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} +#endif + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stb_image uses ints pervasively, including for offset calculations. +// therefore the largest decoded image size we can support with the +// current code, even on 64-bit targets, is INT_MAX. this is not a +// significant limitation for the intended use case. +// +// we do, however, need to make sure our size calculations don't +// overflow. hence a few helper functions for size calculations that +// multiply integers together, making sure that they're non-negative +// and no overflow occurs. + +// return 1 if the sum is valid, 0 on overflow. +// negative terms are considered invalid. +static int stbi__addsizes_valid(int a, int b) +{ + if (b < 0) return 0; + // now 0 <= b <= INT_MAX, hence also + // 0 <= INT_MAX - b <= INTMAX. + // And "a + b <= INT_MAX" (which might overflow) is the + // same as a <= INT_MAX - b (no overflow) + return a <= INT_MAX - b; +} + +// returns 1 if the product is valid, 0 on overflow. +// negative factors are considered invalid. +static int stbi__mul2sizes_valid(int a, int b) +{ + if (a < 0 || b < 0) return 0; + if (b == 0) return 1; // mul-by-0 is always safe + // portable way to check for no overflows in a*b + return a <= INT_MAX/b; +} + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow +static int stbi__mad2sizes_valid(int a, int b, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add); +} +#endif + +// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow +static int stbi__mad3sizes_valid(int a, int b, int c, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__addsizes_valid(a*b*c, add); +} + +// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add); +} +#endif + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// mallocs with size overflow checking +static void *stbi__malloc_mad2(int a, int b, int add) +{ + if (!stbi__mad2sizes_valid(a, b, add)) return NULL; + return stbi__malloc(a*b + add); +} +#endif + +static void *stbi__malloc_mad3(int a, int b, int c, int add) +{ + if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL; + return stbi__malloc(a*b*c + add); +} + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static void *stbi__malloc_mad4(int a, int b, int c, int d, int add) +{ + if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL; + return stbi__malloc(a*b*c*d + add); +} +#endif + +// 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_global = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_global = flag_true_if_should_flip; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__vertically_flip_on_load stbi__vertically_flip_on_load_global +#else +static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set; + +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_local = flag_true_if_should_flip; + stbi__vertically_flip_on_load_set = 1; +} + +#define stbi__vertically_flip_on_load (stbi__vertically_flip_on_load_set \ + ? stbi__vertically_flip_on_load_local \ + : stbi__vertically_flip_on_load_global) +#endif // STBI_THREAD_LOCAL + +static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields + ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed + ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order + ri->num_channels = 0; + + // test the formats with a very explicit header first (at least a FOURCC + // or distinctive magic number first) + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc); + #else + STBI_NOTUSED(bpc); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri); + #endif + + // then the formats that can end up attempting to load with just 1 or 2 + // bytes matching expectations; these are prone to false positives, so + // try them later + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri); + 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, ri); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi_uc *reduced; + + reduced = (stbi_uc *) stbi__malloc(img_len); + if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling + + STBI_FREE(orig); + return reduced; +} + +static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi__uint16 *enlarged; + + enlarged = (stbi__uint16 *) stbi__malloc(img_len*2); + if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff + + STBI_FREE(orig); + return enlarged; +} + +static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel) +{ + int row; + size_t bytes_per_row = (size_t)w * bytes_per_pixel; + stbi_uc temp[2048]; + stbi_uc *bytes = (stbi_uc *)image; + + for (row = 0; row < (h>>1); row++) { + stbi_uc *row0 = bytes + row*bytes_per_row; + stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row; + // swap row0 with row1 + size_t bytes_left = bytes_per_row; + while (bytes_left) { + size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp); + memcpy(temp, row0, bytes_copy); + memcpy(row0, row1, bytes_copy); + memcpy(row1, temp, bytes_copy); + row0 += bytes_copy; + row1 += bytes_copy; + bytes_left -= bytes_copy; + } + } +} + +#ifndef STBI_NO_GIF +static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel) +{ + int slice; + int slice_size = w * h * bytes_per_pixel; + + stbi_uc *bytes = (stbi_uc *)image; + for (slice = 0; slice < z; ++slice) { + stbi__vertical_flip(bytes, w, h, bytes_per_pixel); + bytes += slice_size; + } +} +#endif + +static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 8) { + result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 8; + } + + // @TODO: move stbi__convert_format to here + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc)); + } + + return (unsigned char *) result; +} + +static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 16) { + result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 16; + } + + // @TODO: move stbi__convert_format16 to here + // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16)); + } + + return (stbi__uint16 *) result; +} + +#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR) +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 channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(float)); + } +} +#endif + +#ifndef STBI_NO_STDIO + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); +#endif + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif 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_and_postprocess_8bit(&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; +} + +STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__uint16 *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_16bit(&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; +} + +STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + stbi__uint16 *result; + if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file_16(f,x,y,comp,req_comp); + fclose(f); + return result; +} + + +#endif //!STBI_NO_STDIO + +STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +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_and_postprocess_8bit(&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_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_mem(&s,buffer,len); + + result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp); + if (stbi__vertically_flip_on_load) { + stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); + } + + return result; +} +#endif + +#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)) { + stbi__result_info ri; + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_and_postprocess_8bit(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 + long pos = ftell(f); + int res; + stbi__context s; + stbi__start_file(&s,f); + res = stbi__hdr_test(&s); + fseek(f, pos, SEEK_SET); + return res; + #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); + s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original); + 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; +} + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +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; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) +// nothing +#else +static void stbi__skip(stbi__context *s, int n) +{ + if (n == 0) return; // already there! + 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; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM) +// nothing +#else +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; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} +#endif + +#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); + z += (stbi__uint32)stbi__get16le(s) << 16; + return z; +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +////////////////////////////////////////////////////////////////////////////// +// +// 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); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +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_mad3(req_comp, x, y, 0); + 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 STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__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 (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 stbi__compute_y_16(int r, int g, int b) +{ + return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + stbi__uint16 *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2); + if (good == NULL) { + STBI_FREE(data); + return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + stbi__uint16 *src = data + j * x * img_n ; + stbi__uint16 *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__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 (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#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; + if (!data) return NULL; + output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0); + 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 (n < comp) { + for (i=0; i < x*y; ++i) { + output[i*comp + n] = data[i*comp + n]/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; + if (!data) return NULL; + output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0); + 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__uint16 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 jfif; + int app14_color_transform; // Adobe APP14 tag + int rgb; + + 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; + unsigned int 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 >= (1u << 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 += (~0U << 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 * 256) + (run * 16) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + unsigned int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes + 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 const 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 = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative) + 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__jbias[n] & (sgn - 1)); +} + +// 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 const 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__uint16 *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 || t > 15) 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); + if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + 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 * (1 << 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) * (1 << 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) * (1 << 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) * 4096) + +// 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]*4; + 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); // consume repeated 0xff fill bytes + 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 = j->img_comp[3].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__uint16 *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, sixteen = (p != 0); + int t = q & 15,i; + if (p != 0 && p != 1) 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__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s)); + L -= (sixteen ? 129 : 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) { + L = stbi__get16be(z->s); + if (L < 2) { + if (m == 0xFE) + return stbi__err("bad COM len","Corrupt JPEG"); + else + return stbi__err("bad APP len","Corrupt JPEG"); + } + L -= 2; + + if (m == 0xE0 && L >= 5) { // JFIF APP0 segment + static const unsigned char tag[5] = {'J','F','I','F','\0'}; + int ok = 1; + int i; + for (i=0; i < 5; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 5; + if (ok) + z->jfif = 1; + } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment + static const unsigned char tag[6] = {'A','d','o','b','e','\0'}; + int ok = 1; + int i; + for (i=0; i < 6; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 6; + if (ok) { + stbi__get8(z->s); // version + stbi__get16be(z->s); // flags0 + stbi__get16be(z->s); // flags1 + z->app14_color_transform = stbi__get8(z->s); // color transform + L -= 6; + } + } + + stbi__skip(z->s, L); + return 1; + } + + return stbi__err("unknown marker","Corrupt JPEG"); +} + +// 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__free_jpeg_components(stbi__jpeg *z, int ncomp, int why) +{ + int i; + for (i=0; i < ncomp; ++i) { + if (z->img_comp[i].raw_data) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + z->img_comp[i].data = NULL; + } + if (z->img_comp[i].raw_coeff) { + STBI_FREE(z->img_comp[i].raw_coeff); + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].coeff = 0; + } + if (z->img_comp[i].linebuf) { + STBI_FREE(z->img_comp[i].linebuf); + z->img_comp[i].linebuf = NULL; + } + } + return why; +} + +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 + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + c = stbi__get8(s); + if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG"); + 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"); + + z->rgb = 0; + for (i=0; i < s->img_n; ++i) { + static const unsigned char rgb[3] = { 'R', 'G', 'B' }; + z->img_comp[i].id = stbi__get8(s); + if (s->img_n == 3 && z->img_comp[i].id == rgb[i]) + ++z->rgb; + 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 (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) 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; + } + + // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios + // and I've never seen a non-corrupted JPEG file actually use them + for (i=0; i < s->img_n; ++i) { + if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG"); + if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG"); + } + + // 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; + // these sizes can't be more than 17 bits + 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 + // + // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier) + // so these muls can't overflow with 32-bit ints (which we require) + 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].coeff = 0; + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].linebuf = NULL; + z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15); + if (z->img_comp[i].raw_data == NULL) + return stbi__free_jpeg_components(z, i+1, 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); + if (z->progressive) { + // w2, h2 are multiples of 8 (see above) + z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8; + z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8; + z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15); + if (z->img_comp[i].raw_coeff == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } + } + + 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->jfif = 0; + z->app14_color_transform = -1; // valid values are 0,1,2 + 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; + } + } + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + } else if (stbi__DNL(m)) { + int Ld = stbi__get16be(j->s); + stbi__uint32 NL = stbi__get16be(j->s); + if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG"); + if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG"); + } 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; +} + +// 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 stbi__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* stbi__float2fixed(1.40200f); + g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__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; + } +} + +#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* stbi__float2fixed(1.40200f); + g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__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; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + 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) +{ + stbi__free_jpeg_components(j, j->s->img_n, 0); +} + +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; + +// fast 0..255 * 0..255 => 0..255 rounded multiplication +static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y) +{ + unsigned int t = x*y + 128; + return (stbi_uc) ((t + (t >>8)) >> 8); +} + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n, is_rgb; + 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 >= 3 ? 3 : 1; + + is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif)); + + if (z->s->img_n == 3 && n < 3 && !is_rgb) + decode_n = 1; + else + decode_n = z->s->img_n; + + // nothing to do if no components requested; check this now to avoid + // accessing uninitialized coutput[0] later + if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; } + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL }; + + 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_mad3(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) { + if (is_rgb) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = y[i]; + out[1] = coutput[1][i]; + out[2] = coutput[2][i]; + out[3] = 255; + out += n; + } + } else { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else if (z->s->img_n == 4) { + if (z->app14_color_transform == 0) { // CMYK + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(coutput[0][i], m); + out[1] = stbi__blinn_8x8(coutput[1][i], m); + out[2] = stbi__blinn_8x8(coutput[2][i], m); + out[3] = 255; + out += n; + } + } else if (z->app14_color_transform == 2) { // YCCK + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(255 - out[0], m); + out[1] = stbi__blinn_8x8(255 - out[1], m); + out[2] = stbi__blinn_8x8(255 - out[2], m); + out += n; + } + } else { // YCbCr + alpha? Ignore the fourth channel for now + 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 { + if (is_rgb) { + if (n == 1) + for (i=0; i < z->s->img_x; ++i) + *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + else { + for (i=0; i < z->s->img_x; ++i, out += 2) { + out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + out[1] = 255; + } + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 0) { + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + stbi_uc r = stbi__blinn_8x8(coutput[0][i], m); + stbi_uc g = stbi__blinn_8x8(coutput[1][i], m); + stbi_uc b = stbi__blinn_8x8(coutput[2][i], m); + out[0] = stbi__compute_y(r, g, b); + out[1] = 255; + out += n; + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 2) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]); + out[1] = 255; + 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 >= 3 ? 3 : 1; // report original components, not output + return output; + } +} + +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + unsigned char* result; + stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__errpuc("outofmem", "Out of memory"); + STBI_NOTUSED(ri); + j->s = s; + stbi__setup_jpeg(j); + result = load_jpeg_image(j, x,y,comp,req_comp); + STBI_FREE(j); + return result; +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__err("outofmem", "Out of memory"); + j->s = s; + stbi__setup_jpeg(j); + r = stbi__decode_jpeg_header(j, STBI__SCAN_type); + stbi__rewind(s); + STBI_FREE(j); + 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 >= 3 ? 3 : 1; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + int result; + stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg))); + if (!j) return stbi__err("outofmem", "Out of memory"); + j->s = s; + result = stbi__jpeg_info_raw(j, x, y, comp); + STBI_FREE(j); + return result; +} +#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) +#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet + +// 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[STBI__ZNSYMS]; + stbi__uint16 value[STBI__ZNSYMS]; +} 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, const 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 int stbi__zeof(stbi__zbuf *z) +{ + return (z->zbuffer >= z->zbuffer_end); +} + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + return stbi__zeof(z) ? 0 : *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + if (z->code_buffer >= (1U << z->num_bits)) { + z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */ + return; + } + 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]; + if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere! + if (z->size[b] != s) return -1; // was originally an assert, but report failure instead. + 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) { + if (stbi__zeof(a)) { + return -1; /* report error for unexpected end of data. */ + } + 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; + unsigned int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (unsigned int) (z->zout - z->zout_start); + limit = old_limit = (unsigned) (z->zout_end - z->zout_start); + if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory"); + while (cur + n > limit) { + if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory"); + 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 const 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 const 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 const 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 const 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 const 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; + int ntot = hlit + hdist; + + 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 < ntot) { + 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 { + stbi_uc fill = 0; + if (c == 16) { + c = stbi__zreceive(a,2)+3; + if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG"); + fill = lencodes[n-1]; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + } else if (c == 18) { + c = stbi__zreceive(a,7)+11; + } else { + return stbi__err("bad codelengths", "Corrupt PNG"); + } + if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG"); + memset(lencodes+n, fill, c); + n += c; + } + } + if (n != ntot) 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_uncompressed_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; + } + if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG"); + // 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 (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + 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; +} + +static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] = +{ + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8 +}; +static const stbi_uc stbi__zdefault_distance[32] = +{ + 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5 +}; +/* +Init algorithm: +{ + 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_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) 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 const 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; + int depth; +} 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 const 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) +{ + int bytes = (depth == 16? 2 : 1); + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n*bytes; + stbi__uint32 img_len, img_width_bytes; + int k; + int img_n = s->img_n; // copy it into a local for later + + int output_bytes = out_n*bytes; + int filter_bytes = img_n*bytes; + int width = x; + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG"); + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + img_len = (img_width_bytes + 1) * y; + + // we used to check for exact match between raw_len and img_len on non-interlaced PNGs, + // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros), + // so just check for raw_len < img_len always. + 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; + int filter = *raw++; + + if (filter > 4) + return stbi__err("invalid filter","Corrupt PNG"); + + if (depth < 8) { + if (img_width_bytes > x) return stbi__err("invalid width","Corrupt PNG"); + 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; + } + prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above + + // 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 if (depth == 16) { + if (img_n != out_n) { + cur[filter_bytes] = 255; // first pixel top byte + cur[filter_bytes+1] = 255; // first pixel bottom byte + } + raw += filter_bytes; + cur += output_bytes; + prior += output_bytes; + } 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)*filter_bytes; + #define STBI__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; + STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break; + STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break; + STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break; + STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break; + STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break; + STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break; + } + #undef STBI__CASE + raw += nk; + } else { + STBI_ASSERT(img_n+1 == out_n); + #define STBI__CASE(f) \ + case f: \ + for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \ + for (k=0; k < filter_bytes; ++k) + switch (filter) { + STBI__CASE(STBI__F_none) { cur[k] = raw[k]; } break; + STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break; + STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break; + STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break; + STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break; + STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break; + STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break; + } + #undef STBI__CASE + + // the loop above sets the high byte of the pixels' alpha, but for + // 16 bit png files we also need the low byte set. we'll do that here. + if (depth == 16) { + cur = a->out + stride*j; // start at the beginning of the row again + for (i=0; i < x; ++i,cur+=output_bytes) { + cur[filter_bytes+1] = 255; + } + } + } + } + + // 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]; + } + } + } + } + } else if (depth == 16) { + // force the image data from big-endian to platform-native. + // this is done in a separate pass due to the decoding relying + // on the data being untouched, but could probably be done + // per-line during decode if care is taken. + stbi_uc *cur = a->out; + stbi__uint16 *cur16 = (stbi__uint16*)cur; + + for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) { + *cur16 = (cur[0] << 8) | cur[1]; + } + } + + 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) +{ + int bytes = (depth == 16 ? 2 : 1); + int out_bytes = out_n * bytes; + 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_mad3(a->s->img_x, a->s->img_y, out_bytes, 0); + if (!final) return stbi__err("outofmem", "Out of memory"); + 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_bytes + out_x*out_bytes, + a->out + (j*x+i)*out_bytes, out_bytes); + } + } + 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__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi__uint16 *p = (stbi__uint16*) z->out; + + // compute color-based transparency, assuming we've + // already got 65535 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 : 65535); + 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_mad2(pixel_count, pal_img_n, 0); + 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_global = 0; +static int stbi__de_iphone_flag_global = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_global = flag_true_if_should_convert; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__unpremultiply_on_load stbi__unpremultiply_on_load_global +#define stbi__de_iphone_flag stbi__de_iphone_flag_global +#else +static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set; +static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set; + +STBIDEF void stbi__unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply; + stbi__unpremultiply_on_load_set = 1; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_local = flag_true_if_should_convert; + stbi__de_iphone_flag_set = 1; +} + +#define stbi__unpremultiply_on_load (stbi__unpremultiply_on_load_set \ + ? stbi__unpremultiply_on_load_local \ + : stbi__unpremultiply_on_load_global) +#define stbi__de_iphone_flag (stbi__de_iphone_flag_set \ + ? stbi__de_iphone_flag_local \ + : stbi__de_iphone_flag_global) +#endif // STBI_THREAD_LOCAL + +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) { + stbi_uc half = a / 2; + p[0] = (p[2] * 255 + half) / a; + p[1] = (p[1] * 255 + half) / a; + p[2] = ( t * 255 + half) / 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) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (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]={0}; + stbi__uint16 tc16[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=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); + s->img_y = stbi__get32be(s); + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3 && z->depth == 16) 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; + if (z->depth == 16) { + for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is + } else { + for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->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 * z->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, z->depth, color, interlace)) return 0; + if (has_trans) { + if (z->depth == 16) { + if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0; + } else { + 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; + } else if (has_trans) { + // non-paletted image with tRNS -> source image has (constant) alpha + ++s->img_n; + } + STBI_FREE(z->expanded); z->expanded = NULL; + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + 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 void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri) +{ + void *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)) { + if (p->depth <= 8) + ri->bits_per_channel = 8; + else if (p->depth == 16) + ri->bits_per_channel = 16; + else + return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth"); + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + if (ri->bits_per_channel == 8) + result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + else + result = stbi__convert_format16((stbi__uint16 *) 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_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 void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp, ri); +} + +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); +} + +static int stbi__png_is16(stbi__context *s) +{ + stbi__png p; + p.s = s; + if (!stbi__png_info_raw(&p, NULL, NULL, NULL)) + return 0; + if (p.depth != 16) { + stbi__rewind(p.s); + return 0; + } + return 1; +} +#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;/* >>= 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; +} + +// extract an arbitrarily-aligned N-bit value (N=bits) +// from v, and then make it 8-bits long and fractionally +// extend it to full full range. +static int stbi__shiftsigned(unsigned int v, int shift, int bits) +{ + static unsigned int mul_table[9] = { + 0, + 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/, + 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/, + }; + static unsigned int shift_table[9] = { + 0, 0,0,1,0,2,4,6,0, + }; + if (shift < 0) + v <<= -shift; + else + v >>= shift; + STBI_ASSERT(v < 256); + v >>= (8-bits); + STBI_ASSERT(bits >= 0 && bits <= 8); + return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits]; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; + int extra_read; +} stbi__bmp_data; + +static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress) +{ + // BI_BITFIELDS specifies masks explicitly, don't override + if (compress == 3) + return 1; + + if (compress == 0) { + if (info->bpp == 16) { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } else 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 { + // otherwise, use defaults, which is all-0 + info->mr = info->mg = info->mb = info->ma = 0; + } + return 1; + } + return 0; // error +} + +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); + info->mr = info->mg = info->mb = info->ma = 0; + info->extra_read = 14; + + if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP"); + + 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 (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes + if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel + 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) { + if (compress == 0) { + stbi__bmp_set_mask_defaults(info, compress); + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->extra_read += 12; + // 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 { + // V4/V5 header + 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); + if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs + stbi__bmp_set_mask_defaults(info, compress); + 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 void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + 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; + STBI_NOTUSED(ri); + + 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); + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + 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 - info.extra_read - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - info.extra_read - info.hsz) >> 2; + } + if (psize == 0) { + if (info.offset != s->callback_already_read + (s->img_buffer - s->img_buffer_original)) { + return stbi__errpuc("bad offset", "Corrupt BMP"); + } + } + + if (info.bpp == 24 && ma == 0xff000000) + s->img_n = 3; + else + 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 + + // sanity-check size + if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "Corrupt BMP"); + + out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0); + 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 - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 1) width = (s->img_x + 7) >> 3; + else 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; + if (info.bpp == 1) { + for (j=0; j < (int) s->img_y; ++j) { + int bit_offset = 7, v = stbi__get8(s); + for (i=0; i < (int) s->img_x; ++i) { + int color = (v>>bit_offset)&0x1; + out[z++] = pal[color][0]; + out[z++] = pal[color][1]; + out[z++] = pal[color][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + if((--bit_offset) < 0) { + bit_offset = 7; + v = stbi__get8(s); + } + } + stbi__skip(s, pad); + } + } else { + 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 - info.extra_read - 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); + if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + } + 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)); + unsigned 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; + // fallthrough + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fallthrough + 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 +static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = (stbi__uint16)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] = (stbi_uc)((r * 255)/31); + out[1] = (stbi_uc)((g * 255)/31); + out[2] = (stbi_uc)((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 void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + // 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] = {0}; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + STBI_NOTUSED(ri); + STBI_NOTUSED(tga_x_origin); // @TODO + STBI_NOTUSED(tga_y_origin); // @TODO + + if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // 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; + + if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0)) + return stbi__errpuc("too large", "Corrupt TGA"); + + tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0); + 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) + { + if (tga_palette_len == 0) { /* you have to have at least one entry! */ + STBI_FREE(tga_data); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0); + 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; + STBI_NOTUSED(tga_palette_start); + // 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 int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount) +{ + int count, nleft, len; + + count = 0; + while ((nleft = pixelCount - count) > 0) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + if (len > nleft) return 0; // corrupt data + 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 = 257 - len; + if (len > nleft) return 0; // corrupt data + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + + return 1; +} + +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + int pixelCount; + int channelCount, compression; + int channel, i; + int bitdepth; + int w,h; + stbi_uc *out; + STBI_NOTUSED(ri); + + // 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); + + if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // 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"); + + // Check size + if (!stbi__mad3sizes_valid(4, w, h, 0)) + return stbi__errpuc("too large", "Corrupt PSD"); + + // Create the destination image. + + if (!compression && bitdepth == 16 && bpc == 16) { + out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0); + ri->bits_per_channel = 16; + } else + 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 preceded 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. + if (!stbi__psd_decode_rle(s, p, pixelCount)) { + STBI_FREE(out); + return stbi__errpuc("corrupt", "bad RLE data"); + } + } + } + + } 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 (or 16-bit) value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + if (channel >= channelCount) { + // Fill this channel with default data. + if (bitdepth == 16 && bpc == 16) { + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + stbi__uint16 val = channel == 3 ? 65535 : 0; + for (i = 0; i < pixelCount; i++, q += 4) + *q = val; + } else { + stbi_uc *p = out+channel; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } + } else { + if (ri->bits_per_channel == 16) { // output bpc + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + for (i = 0; i < pixelCount; i++, q += 4) + *q = (stbi__uint16) stbi__get16be(s); + } else { + stbi_uc *p = out+channel; + if (bitdepth == 16) { // input bpc + 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); + } + } + } + } + } + + // remove weird white matte from PSD + if (channelCount >= 4) { + if (ri->bits_per_channel == 16) { + for (i=0; i < w*h; ++i) { + stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i; + if (pixel[3] != 0 && pixel[3] != 65535) { + float a = pixel[3] / 65535.0f; + float ra = 1.0f / a; + float inv_a = 65535.0f * (1 - ra); + pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a); + pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a); + pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a); + } + } + } else { + for (i=0; i < w*h; ++i) { + unsigned char *pixel = out + 4*i; + if (pixel[3] != 0 && pixel[3] != 255) { + float a = pixel[3] / 255.0f; + float ra = 1.0f / a; + float inv_a = 255.0f * (1 - ra); + pixel[0] = (unsigned char) (pixel[0]*ra + inv_a); + pixel[1] = (unsigned char) (pixel[1]*ra + inv_a); + pixel[2] = (unsigned char) (pixel[2]*ra + inv_a); + } + } + } + } + + // convert to desired output format + if (req_comp && req_comp != 4) { + if (ri->bits_per_channel == 16) + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h); + else + 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 void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri) +{ + stbi_uc *result; + int i, x,y, internal_comp; + STBI_NOTUSED(ri); + + if (!comp) comp = &internal_comp; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + + if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC 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_mad3(x, y, 4, 0); + if (!result) return stbi__errpuc("outofmem", "Out of memory"); + 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; // output buffer (always 4 components) + stbi_uc *background; // The current "background" as far as a gif is concerned + stbi_uc *history; + int flags, bgindex, ratio, transparent, eflags; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[8192]; + 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; + int delay; +} 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 (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + + 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 = (stbi__gif*) stbi__malloc(sizeof(stbi__gif)); + if (!g) return stbi__err("outofmem", "Out of memory"); + if (!stbi__gif_header(s, g, comp, 1)) { + STBI_FREE(g); + stbi__rewind( s ); + return 0; + } + if (x) *x = g->w; + if (y) *y = g->h; + STBI_FREE(g); + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + int idx; + + // 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; + + idx = g->cur_x + g->cur_y; + p = &g->out[idx]; + g->history[idx / 4] = 1; + + c = &g->color_table[g->codes[code].suffix * 4]; + if (c[3] > 128) { // don't render transparent pixels; + 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 > 8192) { + 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"); + } + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +// two back is the image from two frames ago, used for a very specific disposal format +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back) +{ + int dispose; + int first_frame; + int pi; + int pcount; + STBI_NOTUSED(req_comp); + + // on first frame, any non-written pixels get the background colour (non-transparent) + first_frame = 0; + if (g->out == 0) { + if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header + if (!stbi__mad3sizes_valid(4, g->w, g->h, 0)) + return stbi__errpuc("too large", "GIF image is too large"); + pcount = g->w * g->h; + g->out = (stbi_uc *) stbi__malloc(4 * pcount); + g->background = (stbi_uc *) stbi__malloc(4 * pcount); + g->history = (stbi_uc *) stbi__malloc(pcount); + if (!g->out || !g->background || !g->history) + return stbi__errpuc("outofmem", "Out of memory"); + + // image is treated as "transparent" at the start - ie, nothing overwrites the current background; + // background colour is only used for pixels that are not rendered first frame, after that "background" + // color refers to the color that was there the previous frame. + memset(g->out, 0x00, 4 * pcount); + memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent) + memset(g->history, 0x00, pcount); // pixels that were affected previous frame + first_frame = 1; + } else { + // second frame - how do we dispose of the previous one? + dispose = (g->eflags & 0x1C) >> 2; + pcount = g->w * g->h; + + if ((dispose == 3) && (two_back == 0)) { + dispose = 2; // if I don't have an image to revert back to, default to the old background + } + + if (dispose == 3) { // use previous graphic + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); + } + } + } else if (dispose == 2) { + // restore what was changed last frame to background before that frame; + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); + } + } + } else { + // This is a non-disposal case eithe way, so just + // leave the pixels as is, and they will become the new background + // 1: do not dispose + // 0: not specified. + } + + // background is what out is after the undoing of the previou frame; + memcpy( g->background, g->out, 4 * g->w * g->h ); + } + + // clear my history; + memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame + + for (;;) { + int tag = stbi__get8(s); + switch (tag) { + case 0x2C: /* Image Descriptor */ + { + 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; + + // if the width of the specified rectangle is 0, that means + // we may not see *any* pixels or the image is malformed; + // to make sure this is caught, move the current y down to + // max_y (which is what out_gif_code checks). + if (w == 0) + g->cur_y = g->max_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) { + 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) return NULL; + + // if this was the first frame, + pcount = g->w * g->h; + if (first_frame && (g->bgindex > 0)) { + // if first frame, any pixel not drawn to gets the background color + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi] == 0) { + g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; + memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); + } + } + } + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + int ext = stbi__get8(s); + if (ext == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths. + + // unset old transparent + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 255; + } + if (g->eflags & 0x01) { + g->transparent = stbi__get8(s); + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 0; + } + } else { + // don't need transparent + stbi__skip(s, 1); + g->transparent = -1; + } + } 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"); + } + } +} + +static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays) +{ + STBI_FREE(g->out); + STBI_FREE(g->history); + STBI_FREE(g->background); + + if (out) STBI_FREE(out); + if (delays && *delays) STBI_FREE(*delays); + return stbi__errpuc("outofmem", "Out of memory"); +} + +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + if (stbi__gif_test(s)) { + int layers = 0; + stbi_uc *u = 0; + stbi_uc *out = 0; + stbi_uc *two_back = 0; + stbi__gif g; + int stride; + int out_size = 0; + int delays_size = 0; + + STBI_NOTUSED(out_size); + STBI_NOTUSED(delays_size); + + memset(&g, 0, sizeof(g)); + if (delays) { + *delays = 0; + } + + do { + u = stbi__gif_load_next(s, &g, comp, req_comp, two_back); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + + if (u) { + *x = g.w; + *y = g.h; + ++layers; + stride = g.w * g.h * 4; + + if (out) { + void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride ); + if (!tmp) + return stbi__load_gif_main_outofmem(&g, out, delays); + else { + out = (stbi_uc*) tmp; + out_size = layers * stride; + } + + if (delays) { + int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers ); + if (!new_delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + *delays = new_delays; + delays_size = layers * sizeof(int); + } + } else { + out = (stbi_uc*)stbi__malloc( layers * stride ); + if (!out) + return stbi__load_gif_main_outofmem(&g, out, delays); + out_size = layers * stride; + if (delays) { + *delays = (int*) stbi__malloc( layers * sizeof(int) ); + if (!*delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + delays_size = layers * sizeof(int); + } + } + memcpy( out + ((layers - 1) * stride), u, stride ); + if (layers >= 2) { + two_back = out - 2 * stride; + } + + if (delays) { + (*delays)[layers - 1U] = g.delay; + } + } + } while (u != 0); + + // free temp buffer; + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + + // do the final conversion after loading everything; + if (req_comp && req_comp != 4) + out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h); + + *z = layers; + return out; + } else { + return stbi__errpuc("not GIF", "Image was not as a gif type."); + } +} + +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + STBI_NOTUSED(ri); + + u = stbi__gif_load_next(s, &g, comp, req_comp, 0); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + + // moved conversion to after successful load so that the same + // can be done for multiple frames. + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } else if (g.out) { + // if there was an error and we allocated an image buffer, free it! + STBI_FREE(g.out); + } + + // free buffers needed for multiple frame loading; + STBI_FREE(g.history); + STBI_FREE(g.background); + + 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) +{ + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + stbi__rewind(s); + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s, "#?RADIANCE\n"); + stbi__rewind(s); + if(!r) { + r = stbi__hdr_test_core(s, "#?RGBE\n"); + 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, stbi__result_info *ri) +{ + 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; + const char *headerToken; + STBI_NOTUSED(ri); + + // Check identifier + headerToken = stbi__hdr_gettoken(s,buffer); + if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 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); + + if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0)) + return stbi__errpf("too large", "HDR image is too large"); + + // Read data + hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0); + if (!hdr_data) + return stbi__errpf("outofmem", "Out of memory"); + + // 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_mad2(width, 4, 0); + if (!scanline) { + STBI_FREE(hdr_data); + return stbi__errpf("outofmem", "Out of memory"); + } + } + + for (k = 0; k < 4; ++k) { + int nleft; + i = 0; + while ((nleft = width - i) > 0) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + 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); + } + if (scanline) + 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; + int dummy; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + 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); + if (p == NULL) { + stbi__rewind( s ); + return 0; + } + if (x) *x = s->img_x; + if (y) *y = s->img_y; + if (comp) { + if (info.bpp == 24 && info.ma == 0xff000000) + *comp = 3; + else + *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, dummy, depth; + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + 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); + depth = stbi__get16be(s); + if (depth != 8 && depth != 16) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi__psd_is16(stbi__context *s) +{ + int channelCount, depth; + 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; + } + STBI_NOTUSED(stbi__get32be(s)); + STBI_NOTUSED(stbi__get32be(s)); + depth = stbi__get16be(s); + if (depth != 16) { + stbi__rewind( s ); + return 0; + } + 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,dummy; + stbi__pic_packet packets[10]; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + 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) + +#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 void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + STBI_NOTUSED(ri); + + ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n); + if (ri->bits_per_channel == 0) + return 0; + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + + if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8)); + + 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, dummy; + char c, p, t; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + 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 > 65535) + return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images"); + else if (maxv > 255) + return 16; + else + return 8; +} + +static int stbi__pnm_is16(stbi__context *s) +{ + if (stbi__pnm_info(s, NULL, NULL, NULL) == 16) + return 1; + return 0; +} +#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"); +} + +static int stbi__is_16_main(stbi__context *s) +{ + #ifndef STBI_NO_PNG + if (stbi__png_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_is16(s)) return 1; + #endif + return 0; +} + +#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; +} + +STBIDEF int stbi_is_16_bit(char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_is_16_bit_from_file(f); + fclose(f); + return result; +} + +STBIDEF int stbi_is_16_bit_from_file(FILE *f) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__is_16_main(&s); + 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); +} + +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__is_16_main(&s); +} + +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__is_16_main(&s); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug + 1-bit BMP + *_is_16_bit api + avoid warnings + 2.16 (2017-07-23) all functions have 16-bit variants; + STBI_NO_STDIO works again; + compilation fixes; + fix rounding in unpremultiply; + optimize vertical flip; + disable raw_len validation; + documentation fixes + 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode; + warning fixes; disable run-time SSE detection on gcc; + uniform handling of optional "return" values; + thread-safe initialization of zlib tables + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) allocate large structures on the stack + remove white matting for transparent PSD + fix reported channel count for PNG & BMP + re-enable SSE2 in non-gcc 64-bit + support RGB-formatted JPEG + read 16-bit PNGs (only as 8-bit) + 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 +*/ + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/qoi/genc/stb_image_write.h b/qoi/genc/stb_image_write.h new file mode 100644 index 00000000..e4b32ed1 --- /dev/null +++ b/qoi/genc/stb_image_write.h @@ -0,0 +1,1724 @@ +/* stb_image_write - v1.16 - public domain - http://nothings.org/stb + writes out PNG/BMP/TGA/JPEG/HDR images to C stdio - Sean Barrett 2010-2015 + no warranty implied; use at your own risk + + Before #including, + + #define STB_IMAGE_WRITE_IMPLEMENTATION + + in the file that you want to have the implementation. + + Will probably not work correctly with strict-aliasing optimizations. + +ABOUT: + + This header file is a library for writing images to C stdio or a callback. + + The PNG output is not optimal; it is 20-50% larger than the file + written by a decent optimizing implementation; though providing a custom + zlib compress function (see STBIW_ZLIB_COMPRESS) can mitigate that. + This library is designed for source code compactness and simplicity, + not optimal image file size or run-time performance. + +BUILDING: + + You can #define STBIW_ASSERT(x) before the #include to avoid using assert.h. + You can #define STBIW_MALLOC(), STBIW_REALLOC(), and STBIW_FREE() to replace + malloc,realloc,free. + You can #define STBIW_MEMMOVE() to replace memmove() + You can #define STBIW_ZLIB_COMPRESS to use a custom zlib-style compress function + for PNG compression (instead of the builtin one), it must have the following signature: + unsigned char * my_compress(unsigned char *data, int data_len, int *out_len, int quality); + The returned data will be freed with STBIW_FREE() (free() by default), + so it must be heap allocated with STBIW_MALLOC() (malloc() by default), + +UNICODE: + + If compiling for Windows and you wish to use Unicode filenames, compile + with + #define STBIW_WINDOWS_UTF8 + and pass utf8-encoded filenames. Call stbiw_convert_wchar_to_utf8 to convert + Windows wchar_t filenames to utf8. + +USAGE: + + There are five functions, one for each image file format: + + int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes); + int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data); + int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data); + int stbi_write_jpg(char const *filename, int w, int h, int comp, const void *data, int quality); + int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data); + + void stbi_flip_vertically_on_write(int flag); // flag is non-zero to flip data vertically + + There are also five equivalent functions that use an arbitrary write function. You are + expected to open/close your file-equivalent before and after calling these: + + int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes); + int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); + int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); + int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data); + int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality); + + where the callback is: + void stbi_write_func(void *context, void *data, int size); + + You can configure it with these global variables: + int stbi_write_tga_with_rle; // defaults to true; set to 0 to disable RLE + int stbi_write_png_compression_level; // defaults to 8; set to higher for more compression + int stbi_write_force_png_filter; // defaults to -1; set to 0..5 to force a filter mode + + + You can define STBI_WRITE_NO_STDIO to disable the file variant of these + functions, so the library will not use stdio.h at all. However, this will + also disable HDR writing, because it requires stdio for formatted output. + + Each function returns 0 on failure and non-0 on success. + + The functions create an image file defined by the parameters. The image + is a rectangle of pixels stored from left-to-right, top-to-bottom. + Each pixel contains 'comp' channels of data stored interleaved with 8-bits + per channel, in the following order: 1=Y, 2=YA, 3=RGB, 4=RGBA. (Y is + monochrome color.) The rectangle is 'w' pixels wide and 'h' pixels tall. + The *data pointer points to the first byte of the top-left-most pixel. + For PNG, "stride_in_bytes" is the distance in bytes from the first byte of + a row of pixels to the first byte of the next row of pixels. + + PNG creates output files with the same number of components as the input. + The BMP format expands Y to RGB in the file format and does not + output alpha. + + PNG supports writing rectangles of data even when the bytes storing rows of + data are not consecutive in memory (e.g. sub-rectangles of a larger image), + by supplying the stride between the beginning of adjacent rows. The other + formats do not. (Thus you cannot write a native-format BMP through the BMP + writer, both because it is in BGR order and because it may have padding + at the end of the line.) + + PNG allows you to set the deflate compression level by setting the global + variable 'stbi_write_png_compression_level' (it defaults to 8). + + HDR expects linear float data. Since the format is always 32-bit rgb(e) + data, alpha (if provided) is discarded, and for monochrome data it is + replicated across all three channels. + + TGA supports RLE or non-RLE compressed data. To use non-RLE-compressed + data, set the global variable 'stbi_write_tga_with_rle' to 0. + + JPEG does ignore alpha channels in input data; quality is between 1 and 100. + Higher quality looks better but results in a bigger image. + JPEG baseline (no JPEG progressive). + +CREDITS: + + + Sean Barrett - PNG/BMP/TGA + Baldur Karlsson - HDR + Jean-Sebastien Guay - TGA monochrome + Tim Kelsey - misc enhancements + Alan Hickman - TGA RLE + Emmanuel Julien - initial file IO callback implementation + Jon Olick - original jo_jpeg.cpp code + Daniel Gibson - integrate JPEG, allow external zlib + Aarni Koskela - allow choosing PNG filter + + bugfixes: + github:Chribba + Guillaume Chereau + github:jry2 + github:romigrou + Sergio Gonzalez + Jonas Karlsson + Filip Wasil + Thatcher Ulrich + github:poppolopoppo + Patrick Boettcher + github:xeekworx + Cap Petschulat + Simon Rodriguez + Ivan Tikhonov + github:ignotion + Adam Schackart + Andrew Kensler + +LICENSE + + See end of file for license information. + +*/ + +#ifndef INCLUDE_STB_IMAGE_WRITE_H +#define INCLUDE_STB_IMAGE_WRITE_H + +#include + +// if STB_IMAGE_WRITE_STATIC causes problems, try defining STBIWDEF to 'inline' or 'static inline' +#ifndef STBIWDEF +#ifdef STB_IMAGE_WRITE_STATIC +#define STBIWDEF static +#else +#ifdef __cplusplus +#define STBIWDEF extern "C" +#else +#define STBIWDEF extern +#endif +#endif +#endif + +#ifndef STB_IMAGE_WRITE_STATIC // C++ forbids static forward declarations +STBIWDEF int stbi_write_tga_with_rle; +STBIWDEF int stbi_write_png_compression_level; +STBIWDEF int stbi_write_force_png_filter; +#endif + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes); +STBIWDEF int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data); +STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality); + +#ifdef STBIW_WINDOWS_UTF8 +STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif +#endif + +typedef void stbi_write_func(void *context, void *data, int size); + +STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes); +STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data); +STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality); + +STBIWDEF void stbi_flip_vertically_on_write(int flip_boolean); + +#endif//INCLUDE_STB_IMAGE_WRITE_H + +#ifdef STB_IMAGE_WRITE_IMPLEMENTATION + +#ifdef _WIN32 + #ifndef _CRT_SECURE_NO_WARNINGS + #define _CRT_SECURE_NO_WARNINGS + #endif + #ifndef _CRT_NONSTDC_NO_DEPRECATE + #define _CRT_NONSTDC_NO_DEPRECATE + #endif +#endif + +#ifndef STBI_WRITE_NO_STDIO +#include +#endif // STBI_WRITE_NO_STDIO + +#include +#include +#include +#include + +#if defined(STBIW_MALLOC) && defined(STBIW_FREE) && (defined(STBIW_REALLOC) || defined(STBIW_REALLOC_SIZED)) +// ok +#elif !defined(STBIW_MALLOC) && !defined(STBIW_FREE) && !defined(STBIW_REALLOC) && !defined(STBIW_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBIW_MALLOC, STBIW_FREE, and STBIW_REALLOC (or STBIW_REALLOC_SIZED)." +#endif + +#ifndef STBIW_MALLOC +#define STBIW_MALLOC(sz) malloc(sz) +#define STBIW_REALLOC(p,newsz) realloc(p,newsz) +#define STBIW_FREE(p) free(p) +#endif + +#ifndef STBIW_REALLOC_SIZED +#define STBIW_REALLOC_SIZED(p,oldsz,newsz) STBIW_REALLOC(p,newsz) +#endif + + +#ifndef STBIW_MEMMOVE +#define STBIW_MEMMOVE(a,b,sz) memmove(a,b,sz) +#endif + + +#ifndef STBIW_ASSERT +#include +#define STBIW_ASSERT(x) assert(x) +#endif + +#define STBIW_UCHAR(x) (unsigned char) ((x) & 0xff) + +#ifdef STB_IMAGE_WRITE_STATIC +static int stbi_write_png_compression_level = 8; +static int stbi_write_tga_with_rle = 1; +static int stbi_write_force_png_filter = -1; +#else +int stbi_write_png_compression_level = 8; +int stbi_write_tga_with_rle = 1; +int stbi_write_force_png_filter = -1; +#endif + +static int stbi__flip_vertically_on_write = 0; + +STBIWDEF void stbi_flip_vertically_on_write(int flag) +{ + stbi__flip_vertically_on_write = flag; +} + +typedef struct +{ + stbi_write_func *func; + void *context; + unsigned char buffer[64]; + int buf_used; +} stbi__write_context; + +// initialize a callback-based context +static void stbi__start_write_callbacks(stbi__write_context *s, stbi_write_func *c, void *context) +{ + s->func = c; + s->context = context; +} + +#ifndef STBI_WRITE_NO_STDIO + +static void stbi__stdio_write(void *context, void *data, int size) +{ + fwrite(data,1,size,(FILE*) context); +} + +#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8) +#ifdef __cplusplus +#define STBIW_EXTERN extern "C" +#else +#define STBIW_EXTERN extern +#endif +STBIW_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBIW_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); + +STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbiw__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + +static int stbi__start_write_file(stbi__write_context *s, const char *filename) +{ + FILE *f = stbiw__fopen(filename, "wb"); + stbi__start_write_callbacks(s, stbi__stdio_write, (void *) f); + return f != NULL; +} + +static void stbi__end_write_file(stbi__write_context *s) +{ + fclose((FILE *)s->context); +} + +#endif // !STBI_WRITE_NO_STDIO + +typedef unsigned int stbiw_uint32; +typedef int stb_image_write_test[sizeof(stbiw_uint32)==4 ? 1 : -1]; + +static void stbiw__writefv(stbi__write_context *s, const char *fmt, va_list v) +{ + while (*fmt) { + switch (*fmt++) { + case ' ': break; + case '1': { unsigned char x = STBIW_UCHAR(va_arg(v, int)); + s->func(s->context,&x,1); + break; } + case '2': { int x = va_arg(v,int); + unsigned char b[2]; + b[0] = STBIW_UCHAR(x); + b[1] = STBIW_UCHAR(x>>8); + s->func(s->context,b,2); + break; } + case '4': { stbiw_uint32 x = va_arg(v,int); + unsigned char b[4]; + b[0]=STBIW_UCHAR(x); + b[1]=STBIW_UCHAR(x>>8); + b[2]=STBIW_UCHAR(x>>16); + b[3]=STBIW_UCHAR(x>>24); + s->func(s->context,b,4); + break; } + default: + STBIW_ASSERT(0); + return; + } + } +} + +static void stbiw__writef(stbi__write_context *s, const char *fmt, ...) +{ + va_list v; + va_start(v, fmt); + stbiw__writefv(s, fmt, v); + va_end(v); +} + +static void stbiw__write_flush(stbi__write_context *s) +{ + if (s->buf_used) { + s->func(s->context, &s->buffer, s->buf_used); + s->buf_used = 0; + } +} + +static void stbiw__putc(stbi__write_context *s, unsigned char c) +{ + s->func(s->context, &c, 1); +} + +static void stbiw__write1(stbi__write_context *s, unsigned char a) +{ + if ((size_t)s->buf_used + 1 > sizeof(s->buffer)) + stbiw__write_flush(s); + s->buffer[s->buf_used++] = a; +} + +static void stbiw__write3(stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c) +{ + int n; + if ((size_t)s->buf_used + 3 > sizeof(s->buffer)) + stbiw__write_flush(s); + n = s->buf_used; + s->buf_used = n+3; + s->buffer[n+0] = a; + s->buffer[n+1] = b; + s->buffer[n+2] = c; +} + +static void stbiw__write_pixel(stbi__write_context *s, int rgb_dir, int comp, int write_alpha, int expand_mono, unsigned char *d) +{ + unsigned char bg[3] = { 255, 0, 255}, px[3]; + int k; + + if (write_alpha < 0) + stbiw__write1(s, d[comp - 1]); + + switch (comp) { + case 2: // 2 pixels = mono + alpha, alpha is written separately, so same as 1-channel case + case 1: + if (expand_mono) + stbiw__write3(s, d[0], d[0], d[0]); // monochrome bmp + else + stbiw__write1(s, d[0]); // monochrome TGA + break; + case 4: + if (!write_alpha) { + // composite against pink background + for (k = 0; k < 3; ++k) + px[k] = bg[k] + ((d[k] - bg[k]) * d[3]) / 255; + stbiw__write3(s, px[1 - rgb_dir], px[1], px[1 + rgb_dir]); + break; + } + /* FALLTHROUGH */ + case 3: + stbiw__write3(s, d[1 - rgb_dir], d[1], d[1 + rgb_dir]); + break; + } + if (write_alpha > 0) + stbiw__write1(s, d[comp - 1]); +} + +static void stbiw__write_pixels(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad, int expand_mono) +{ + stbiw_uint32 zero = 0; + int i,j, j_end; + + if (y <= 0) + return; + + if (stbi__flip_vertically_on_write) + vdir *= -1; + + if (vdir < 0) { + j_end = -1; j = y-1; + } else { + j_end = y; j = 0; + } + + for (; j != j_end; j += vdir) { + for (i=0; i < x; ++i) { + unsigned char *d = (unsigned char *) data + (j*x+i)*comp; + stbiw__write_pixel(s, rgb_dir, comp, write_alpha, expand_mono, d); + } + stbiw__write_flush(s); + s->func(s->context, &zero, scanline_pad); + } +} + +static int stbiw__outfile(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, int expand_mono, void *data, int alpha, int pad, const char *fmt, ...) +{ + if (y < 0 || x < 0) { + return 0; + } else { + va_list v; + va_start(v, fmt); + stbiw__writefv(s, fmt, v); + va_end(v); + stbiw__write_pixels(s,rgb_dir,vdir,x,y,comp,data,alpha,pad, expand_mono); + return 1; + } +} + +static int stbi_write_bmp_core(stbi__write_context *s, int x, int y, int comp, const void *data) +{ + if (comp != 4) { + // write RGB bitmap + int pad = (-x*3) & 3; + return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *) data,0,pad, + "11 4 22 4" "4 44 22 444444", + 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header + 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header + } else { + // RGBA bitmaps need a v4 header + // use BI_BITFIELDS mode with 32bpp and alpha mask + // (straight BI_RGB with alpha mask doesn't work in most readers) + return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *)data,1,0, + "11 4 22 4" "4 44 22 444444 4444 4 444 444 444 444", + 'B', 'M', 14+108+x*y*4, 0, 0, 14+108, // file header + 108, x,y, 1,32, 3,0,0,0,0,0, 0xff0000,0xff00,0xff,0xff000000u, 0, 0,0,0, 0,0,0, 0,0,0, 0,0,0); // bitmap V4 header + } +} + +STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_bmp_core(&s, x, y, comp, data); +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_bmp(char const *filename, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_bmp_core(&s, x, y, comp, data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif //!STBI_WRITE_NO_STDIO + +static int stbi_write_tga_core(stbi__write_context *s, int x, int y, int comp, void *data) +{ + int has_alpha = (comp == 2 || comp == 4); + int colorbytes = has_alpha ? comp-1 : comp; + int format = colorbytes < 2 ? 3 : 2; // 3 color channels (RGB/RGBA) = 2, 1 color channel (Y/YA) = 3 + + if (y < 0 || x < 0) + return 0; + + if (!stbi_write_tga_with_rle) { + return stbiw__outfile(s, -1, -1, x, y, comp, 0, (void *) data, has_alpha, 0, + "111 221 2222 11", 0, 0, format, 0, 0, 0, 0, 0, x, y, (colorbytes + has_alpha) * 8, has_alpha * 8); + } else { + int i,j,k; + int jend, jdir; + + stbiw__writef(s, "111 221 2222 11", 0,0,format+8, 0,0,0, 0,0,x,y, (colorbytes + has_alpha) * 8, has_alpha * 8); + + if (stbi__flip_vertically_on_write) { + j = 0; + jend = y; + jdir = 1; + } else { + j = y-1; + jend = -1; + jdir = -1; + } + for (; j != jend; j += jdir) { + unsigned char *row = (unsigned char *) data + j * x * comp; + int len; + + for (i = 0; i < x; i += len) { + unsigned char *begin = row + i * comp; + int diff = 1; + len = 1; + + if (i < x - 1) { + ++len; + diff = memcmp(begin, row + (i + 1) * comp, comp); + if (diff) { + const unsigned char *prev = begin; + for (k = i + 2; k < x && len < 128; ++k) { + if (memcmp(prev, row + k * comp, comp)) { + prev += comp; + ++len; + } else { + --len; + break; + } + } + } else { + for (k = i + 2; k < x && len < 128; ++k) { + if (!memcmp(begin, row + k * comp, comp)) { + ++len; + } else { + break; + } + } + } + } + + if (diff) { + unsigned char header = STBIW_UCHAR(len - 1); + stbiw__write1(s, header); + for (k = 0; k < len; ++k) { + stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin + k * comp); + } + } else { + unsigned char header = STBIW_UCHAR(len - 129); + stbiw__write1(s, header); + stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin); + } + } + } + stbiw__write_flush(s); + } + return 1; +} + +STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_tga_core(&s, x, y, comp, (void *) data); +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_tga(char const *filename, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_tga_core(&s, x, y, comp, (void *) data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR writer +// by Baldur Karlsson + +#define stbiw__max(a, b) ((a) > (b) ? (a) : (b)) + +#ifndef STBI_WRITE_NO_STDIO + +static void stbiw__linear_to_rgbe(unsigned char *rgbe, float *linear) +{ + int exponent; + float maxcomp = stbiw__max(linear[0], stbiw__max(linear[1], linear[2])); + + if (maxcomp < 1e-32f) { + rgbe[0] = rgbe[1] = rgbe[2] = rgbe[3] = 0; + } else { + float normalize = (float) frexp(maxcomp, &exponent) * 256.0f/maxcomp; + + rgbe[0] = (unsigned char)(linear[0] * normalize); + rgbe[1] = (unsigned char)(linear[1] * normalize); + rgbe[2] = (unsigned char)(linear[2] * normalize); + rgbe[3] = (unsigned char)(exponent + 128); + } +} + +static void stbiw__write_run_data(stbi__write_context *s, int length, unsigned char databyte) +{ + unsigned char lengthbyte = STBIW_UCHAR(length+128); + STBIW_ASSERT(length+128 <= 255); + s->func(s->context, &lengthbyte, 1); + s->func(s->context, &databyte, 1); +} + +static void stbiw__write_dump_data(stbi__write_context *s, int length, unsigned char *data) +{ + unsigned char lengthbyte = STBIW_UCHAR(length); + STBIW_ASSERT(length <= 128); // inconsistent with spec but consistent with official code + s->func(s->context, &lengthbyte, 1); + s->func(s->context, data, length); +} + +static void stbiw__write_hdr_scanline(stbi__write_context *s, int width, int ncomp, unsigned char *scratch, float *scanline) +{ + unsigned char scanlineheader[4] = { 2, 2, 0, 0 }; + unsigned char rgbe[4]; + float linear[3]; + int x; + + scanlineheader[2] = (width&0xff00)>>8; + scanlineheader[3] = (width&0x00ff); + + /* skip RLE for images too small or large */ + if (width < 8 || width >= 32768) { + for (x=0; x < width; x++) { + switch (ncomp) { + case 4: /* fallthrough */ + case 3: linear[2] = scanline[x*ncomp + 2]; + linear[1] = scanline[x*ncomp + 1]; + linear[0] = scanline[x*ncomp + 0]; + break; + default: + linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0]; + break; + } + stbiw__linear_to_rgbe(rgbe, linear); + s->func(s->context, rgbe, 4); + } + } else { + int c,r; + /* encode into scratch buffer */ + for (x=0; x < width; x++) { + switch(ncomp) { + case 4: /* fallthrough */ + case 3: linear[2] = scanline[x*ncomp + 2]; + linear[1] = scanline[x*ncomp + 1]; + linear[0] = scanline[x*ncomp + 0]; + break; + default: + linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0]; + break; + } + stbiw__linear_to_rgbe(rgbe, linear); + scratch[x + width*0] = rgbe[0]; + scratch[x + width*1] = rgbe[1]; + scratch[x + width*2] = rgbe[2]; + scratch[x + width*3] = rgbe[3]; + } + + s->func(s->context, scanlineheader, 4); + + /* RLE each component separately */ + for (c=0; c < 4; c++) { + unsigned char *comp = &scratch[width*c]; + + x = 0; + while (x < width) { + // find first run + r = x; + while (r+2 < width) { + if (comp[r] == comp[r+1] && comp[r] == comp[r+2]) + break; + ++r; + } + if (r+2 >= width) + r = width; + // dump up to first run + while (x < r) { + int len = r-x; + if (len > 128) len = 128; + stbiw__write_dump_data(s, len, &comp[x]); + x += len; + } + // if there's a run, output it + if (r+2 < width) { // same test as what we break out of in search loop, so only true if we break'd + // find next byte after run + while (r < width && comp[r] == comp[x]) + ++r; + // output run up to r + while (x < r) { + int len = r-x; + if (len > 127) len = 127; + stbiw__write_run_data(s, len, comp[x]); + x += len; + } + } + } + } + } +} + +static int stbi_write_hdr_core(stbi__write_context *s, int x, int y, int comp, float *data) +{ + if (y <= 0 || x <= 0 || data == NULL) + return 0; + else { + // Each component is stored separately. Allocate scratch space for full output scanline. + unsigned char *scratch = (unsigned char *) STBIW_MALLOC(x*4); + int i, len; + char buffer[128]; + char header[] = "#?RADIANCE\n# Written by stb_image_write.h\nFORMAT=32-bit_rle_rgbe\n"; + s->func(s->context, header, sizeof(header)-1); + +#ifdef __STDC_LIB_EXT1__ + len = sprintf_s(buffer, sizeof(buffer), "EXPOSURE= 1.0000000000000\n\n-Y %d +X %d\n", y, x); +#else + len = sprintf(buffer, "EXPOSURE= 1.0000000000000\n\n-Y %d +X %d\n", y, x); +#endif + s->func(s->context, buffer, len); + + for(i=0; i < y; i++) + stbiw__write_hdr_scanline(s, x, comp, scratch, data + comp*x*(stbi__flip_vertically_on_write ? y-1-i : i)); + STBIW_FREE(scratch); + return 1; + } +} + +STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_hdr_core(&s, x, y, comp, (float *) data); +} + +STBIWDEF int stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_hdr_core(&s, x, y, comp, (float *) data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif // STBI_WRITE_NO_STDIO + + +////////////////////////////////////////////////////////////////////////////// +// +// PNG writer +// + +#ifndef STBIW_ZLIB_COMPRESS +// stretchy buffer; stbiw__sbpush() == vector<>::push_back() -- stbiw__sbcount() == vector<>::size() +#define stbiw__sbraw(a) ((int *) (void *) (a) - 2) +#define stbiw__sbm(a) stbiw__sbraw(a)[0] +#define stbiw__sbn(a) stbiw__sbraw(a)[1] + +#define stbiw__sbneedgrow(a,n) ((a)==0 || stbiw__sbn(a)+n >= stbiw__sbm(a)) +#define stbiw__sbmaybegrow(a,n) (stbiw__sbneedgrow(a,(n)) ? stbiw__sbgrow(a,n) : 0) +#define stbiw__sbgrow(a,n) stbiw__sbgrowf((void **) &(a), (n), sizeof(*(a))) + +#define stbiw__sbpush(a, v) (stbiw__sbmaybegrow(a,1), (a)[stbiw__sbn(a)++] = (v)) +#define stbiw__sbcount(a) ((a) ? stbiw__sbn(a) : 0) +#define stbiw__sbfree(a) ((a) ? STBIW_FREE(stbiw__sbraw(a)),0 : 0) + +static void *stbiw__sbgrowf(void **arr, int increment, int itemsize) +{ + int m = *arr ? 2*stbiw__sbm(*arr)+increment : increment+1; + void *p = STBIW_REALLOC_SIZED(*arr ? stbiw__sbraw(*arr) : 0, *arr ? (stbiw__sbm(*arr)*itemsize + sizeof(int)*2) : 0, itemsize * m + sizeof(int)*2); + STBIW_ASSERT(p); + if (p) { + if (!*arr) ((int *) p)[1] = 0; + *arr = (void *) ((int *) p + 2); + stbiw__sbm(*arr) = m; + } + return *arr; +} + +static unsigned char *stbiw__zlib_flushf(unsigned char *data, unsigned int *bitbuffer, int *bitcount) +{ + while (*bitcount >= 8) { + stbiw__sbpush(data, STBIW_UCHAR(*bitbuffer)); + *bitbuffer >>= 8; + *bitcount -= 8; + } + return data; +} + +static int stbiw__zlib_bitrev(int code, int codebits) +{ + int res=0; + while (codebits--) { + res = (res << 1) | (code & 1); + code >>= 1; + } + return res; +} + +static unsigned int stbiw__zlib_countm(unsigned char *a, unsigned char *b, int limit) +{ + int i; + for (i=0; i < limit && i < 258; ++i) + if (a[i] != b[i]) break; + return i; +} + +static unsigned int stbiw__zhash(unsigned char *data) +{ + stbiw_uint32 hash = data[0] + (data[1] << 8) + (data[2] << 16); + hash ^= hash << 3; + hash += hash >> 5; + hash ^= hash << 4; + hash += hash >> 17; + hash ^= hash << 25; + hash += hash >> 6; + return hash; +} + +#define stbiw__zlib_flush() (out = stbiw__zlib_flushf(out, &bitbuf, &bitcount)) +#define stbiw__zlib_add(code,codebits) \ + (bitbuf |= (code) << bitcount, bitcount += (codebits), stbiw__zlib_flush()) +#define stbiw__zlib_huffa(b,c) stbiw__zlib_add(stbiw__zlib_bitrev(b,c),c) +// default huffman tables +#define stbiw__zlib_huff1(n) stbiw__zlib_huffa(0x30 + (n), 8) +#define stbiw__zlib_huff2(n) stbiw__zlib_huffa(0x190 + (n)-144, 9) +#define stbiw__zlib_huff3(n) stbiw__zlib_huffa(0 + (n)-256,7) +#define stbiw__zlib_huff4(n) stbiw__zlib_huffa(0xc0 + (n)-280,8) +#define stbiw__zlib_huff(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : (n) <= 255 ? stbiw__zlib_huff2(n) : (n) <= 279 ? stbiw__zlib_huff3(n) : stbiw__zlib_huff4(n)) +#define stbiw__zlib_huffb(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : stbiw__zlib_huff2(n)) + +#define stbiw__ZHASH 16384 + +#endif // STBIW_ZLIB_COMPRESS + +STBIWDEF unsigned char * stbi_zlib_compress(unsigned char *data, int data_len, int *out_len, int quality) +{ +#ifdef STBIW_ZLIB_COMPRESS + // user provided a zlib compress implementation, use that + return STBIW_ZLIB_COMPRESS(data, data_len, out_len, quality); +#else // use builtin + static unsigned short lengthc[] = { 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, 259 }; + static unsigned char lengtheb[]= { 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 }; + static unsigned short distc[] = { 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, 32768 }; + static unsigned char disteb[] = { 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 }; + unsigned int bitbuf=0; + int i,j, bitcount=0; + unsigned char *out = NULL; + unsigned char ***hash_table = (unsigned char***) STBIW_MALLOC(stbiw__ZHASH * sizeof(unsigned char**)); + if (hash_table == NULL) + return NULL; + if (quality < 5) quality = 5; + + stbiw__sbpush(out, 0x78); // DEFLATE 32K window + stbiw__sbpush(out, 0x5e); // FLEVEL = 1 + stbiw__zlib_add(1,1); // BFINAL = 1 + stbiw__zlib_add(1,2); // BTYPE = 1 -- fixed huffman + + for (i=0; i < stbiw__ZHASH; ++i) + hash_table[i] = NULL; + + i=0; + while (i < data_len-3) { + // hash next 3 bytes of data to be compressed + int h = stbiw__zhash(data+i)&(stbiw__ZHASH-1), best=3; + unsigned char *bestloc = 0; + unsigned char **hlist = hash_table[h]; + int n = stbiw__sbcount(hlist); + for (j=0; j < n; ++j) { + if (hlist[j]-data > i-32768) { // if entry lies within window + int d = stbiw__zlib_countm(hlist[j], data+i, data_len-i); + if (d >= best) { best=d; bestloc=hlist[j]; } + } + } + // when hash table entry is too long, delete half the entries + if (hash_table[h] && stbiw__sbn(hash_table[h]) == 2*quality) { + STBIW_MEMMOVE(hash_table[h], hash_table[h]+quality, sizeof(hash_table[h][0])*quality); + stbiw__sbn(hash_table[h]) = quality; + } + stbiw__sbpush(hash_table[h],data+i); + + if (bestloc) { + // "lazy matching" - check match at *next* byte, and if it's better, do cur byte as literal + h = stbiw__zhash(data+i+1)&(stbiw__ZHASH-1); + hlist = hash_table[h]; + n = stbiw__sbcount(hlist); + for (j=0; j < n; ++j) { + if (hlist[j]-data > i-32767) { + int e = stbiw__zlib_countm(hlist[j], data+i+1, data_len-i-1); + if (e > best) { // if next match is better, bail on current match + bestloc = NULL; + break; + } + } + } + } + + if (bestloc) { + int d = (int) (data+i - bestloc); // distance back + STBIW_ASSERT(d <= 32767 && best <= 258); + for (j=0; best > lengthc[j+1]-1; ++j); + stbiw__zlib_huff(j+257); + if (lengtheb[j]) stbiw__zlib_add(best - lengthc[j], lengtheb[j]); + for (j=0; d > distc[j+1]-1; ++j); + stbiw__zlib_add(stbiw__zlib_bitrev(j,5),5); + if (disteb[j]) stbiw__zlib_add(d - distc[j], disteb[j]); + i += best; + } else { + stbiw__zlib_huffb(data[i]); + ++i; + } + } + // write out final bytes + for (;i < data_len; ++i) + stbiw__zlib_huffb(data[i]); + stbiw__zlib_huff(256); // end of block + // pad with 0 bits to byte boundary + while (bitcount) + stbiw__zlib_add(0,1); + + for (i=0; i < stbiw__ZHASH; ++i) + (void) stbiw__sbfree(hash_table[i]); + STBIW_FREE(hash_table); + + // store uncompressed instead if compression was worse + if (stbiw__sbn(out) > data_len + 2 + ((data_len+32766)/32767)*5) { + stbiw__sbn(out) = 2; // truncate to DEFLATE 32K window and FLEVEL = 1 + for (j = 0; j < data_len;) { + int blocklen = data_len - j; + if (blocklen > 32767) blocklen = 32767; + stbiw__sbpush(out, data_len - j == blocklen); // BFINAL = ?, BTYPE = 0 -- no compression + stbiw__sbpush(out, STBIW_UCHAR(blocklen)); // LEN + stbiw__sbpush(out, STBIW_UCHAR(blocklen >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(~blocklen)); // NLEN + stbiw__sbpush(out, STBIW_UCHAR(~blocklen >> 8)); + memcpy(out+stbiw__sbn(out), data+j, blocklen); + stbiw__sbn(out) += blocklen; + j += blocklen; + } + } + + { + // compute adler32 on input + unsigned int s1=1, s2=0; + int blocklen = (int) (data_len % 5552); + j=0; + while (j < data_len) { + for (i=0; i < blocklen; ++i) { s1 += data[j+i]; s2 += s1; } + s1 %= 65521; s2 %= 65521; + j += blocklen; + blocklen = 5552; + } + stbiw__sbpush(out, STBIW_UCHAR(s2 >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(s2)); + stbiw__sbpush(out, STBIW_UCHAR(s1 >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(s1)); + } + *out_len = stbiw__sbn(out); + // make returned pointer freeable + STBIW_MEMMOVE(stbiw__sbraw(out), out, *out_len); + return (unsigned char *) stbiw__sbraw(out); +#endif // STBIW_ZLIB_COMPRESS +} + +static unsigned int stbiw__crc32(unsigned char *buffer, int len) +{ +#ifdef STBIW_CRC32 + return STBIW_CRC32(buffer, len); +#else + static unsigned int crc_table[256] = + { + 0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, 0xE963A535, 0x9E6495A3, + 0x0eDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91, + 0x1DB71064, 0x6AB020F2, 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7, + 0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, 0xFA0F3D63, 0x8D080DF5, + 0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B, + 0x35B5A8FA, 0x42B2986C, 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59, + 0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, 0xCFBA9599, 0xB8BDA50F, + 0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D, + 0x76DC4190, 0x01DB7106, 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433, + 0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, 0x91646C97, 0xE6635C01, + 0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457, + 0x65B0D9C6, 0x12B7E950, 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65, + 0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, 0xA4D1C46D, 0xD3D6F4FB, + 0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9, + 0x5005713C, 0x270241AA, 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F, + 0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, 0xB7BD5C3B, 0xC0BA6CAD, + 0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683, + 0xE3630B12, 0x94643B84, 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1, + 0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, 0x196C3671, 0x6E6B06E7, + 0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5, + 0xD6D6A3E8, 0xA1D1937E, 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B, + 0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, 0x316E8EEF, 0x4669BE79, + 0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F, + 0xC5BA3BBE, 0xB2BD0B28, 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D, + 0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, 0x72076785, 0x05005713, + 0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21, + 0x86D3D2D4, 0xF1D4E242, 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777, + 0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, 0x616BFFD3, 0x166CCF45, + 0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB, + 0xAED16A4A, 0xD9D65ADC, 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9, + 0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, 0x54DE5729, 0x23D967BF, + 0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D + }; + + unsigned int crc = ~0u; + int i; + for (i=0; i < len; ++i) + crc = (crc >> 8) ^ crc_table[buffer[i] ^ (crc & 0xff)]; + return ~crc; +#endif +} + +#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=STBIW_UCHAR(a),(o)[1]=STBIW_UCHAR(b),(o)[2]=STBIW_UCHAR(c),(o)[3]=STBIW_UCHAR(d),(o)+=4) +#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v)); +#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3]) + +static void stbiw__wpcrc(unsigned char **data, int len) +{ + unsigned int crc = stbiw__crc32(*data - len - 4, len+4); + stbiw__wp32(*data, crc); +} + +static unsigned char stbiw__paeth(int a, int b, int c) +{ + int p = a + b - c, pa = abs(p-a), pb = abs(p-b), pc = abs(p-c); + if (pa <= pb && pa <= pc) return STBIW_UCHAR(a); + if (pb <= pc) return STBIW_UCHAR(b); + return STBIW_UCHAR(c); +} + +// @OPTIMIZE: provide an option that always forces left-predict or paeth predict +static void stbiw__encode_png_line(unsigned char *pixels, int stride_bytes, int width, int height, int y, int n, int filter_type, signed char *line_buffer) +{ + static int mapping[] = { 0,1,2,3,4 }; + static int firstmap[] = { 0,1,0,5,6 }; + int *mymap = (y != 0) ? mapping : firstmap; + int i; + int type = mymap[filter_type]; + unsigned char *z = pixels + stride_bytes * (stbi__flip_vertically_on_write ? height-1-y : y); + int signed_stride = stbi__flip_vertically_on_write ? -stride_bytes : stride_bytes; + + if (type==0) { + memcpy(line_buffer, z, width*n); + return; + } + + // first loop isn't optimized since it's just one pixel + for (i = 0; i < n; ++i) { + switch (type) { + case 1: line_buffer[i] = z[i]; break; + case 2: line_buffer[i] = z[i] - z[i-signed_stride]; break; + case 3: line_buffer[i] = z[i] - (z[i-signed_stride]>>1); break; + case 4: line_buffer[i] = (signed char) (z[i] - stbiw__paeth(0,z[i-signed_stride],0)); break; + case 5: line_buffer[i] = z[i]; break; + case 6: line_buffer[i] = z[i]; break; + } + } + switch (type) { + case 1: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-n]; break; + case 2: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-signed_stride]; break; + case 3: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - ((z[i-n] + z[i-signed_stride])>>1); break; + case 4: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], z[i-signed_stride], z[i-signed_stride-n]); break; + case 5: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - (z[i-n]>>1); break; + case 6: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], 0,0); break; + } +} + +STBIWDEF unsigned char *stbi_write_png_to_mem(const unsigned char *pixels, int stride_bytes, int x, int y, int n, int *out_len) +{ + int force_filter = stbi_write_force_png_filter; + int ctype[5] = { -1, 0, 4, 2, 6 }; + unsigned char sig[8] = { 137,80,78,71,13,10,26,10 }; + unsigned char *out,*o, *filt, *zlib; + signed char *line_buffer; + int j,zlen; + + if (stride_bytes == 0) + stride_bytes = x * n; + + if (force_filter >= 5) { + force_filter = -1; + } + + filt = (unsigned char *) STBIW_MALLOC((x*n+1) * y); if (!filt) return 0; + line_buffer = (signed char *) STBIW_MALLOC(x * n); if (!line_buffer) { STBIW_FREE(filt); return 0; } + for (j=0; j < y; ++j) { + int filter_type; + if (force_filter > -1) { + filter_type = force_filter; + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, force_filter, line_buffer); + } else { // Estimate the best filter by running through all of them: + int best_filter = 0, best_filter_val = 0x7fffffff, est, i; + for (filter_type = 0; filter_type < 5; filter_type++) { + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, filter_type, line_buffer); + + // Estimate the entropy of the line using this filter; the less, the better. + est = 0; + for (i = 0; i < x*n; ++i) { + est += abs((signed char) line_buffer[i]); + } + if (est < best_filter_val) { + best_filter_val = est; + best_filter = filter_type; + } + } + if (filter_type != best_filter) { // If the last iteration already got us the best filter, don't redo it + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, best_filter, line_buffer); + filter_type = best_filter; + } + } + // when we get here, filter_type contains the filter type, and line_buffer contains the data + filt[j*(x*n+1)] = (unsigned char) filter_type; + STBIW_MEMMOVE(filt+j*(x*n+1)+1, line_buffer, x*n); + } + STBIW_FREE(line_buffer); + zlib = stbi_zlib_compress(filt, y*( x*n+1), &zlen, stbi_write_png_compression_level); + STBIW_FREE(filt); + if (!zlib) return 0; + + // each tag requires 12 bytes of overhead + out = (unsigned char *) STBIW_MALLOC(8 + 12+13 + 12+zlen + 12); + if (!out) return 0; + *out_len = 8 + 12+13 + 12+zlen + 12; + + o=out; + STBIW_MEMMOVE(o,sig,8); o+= 8; + stbiw__wp32(o, 13); // header length + stbiw__wptag(o, "IHDR"); + stbiw__wp32(o, x); + stbiw__wp32(o, y); + *o++ = 8; + *o++ = STBIW_UCHAR(ctype[n]); + *o++ = 0; + *o++ = 0; + *o++ = 0; + stbiw__wpcrc(&o,13); + + stbiw__wp32(o, zlen); + stbiw__wptag(o, "IDAT"); + STBIW_MEMMOVE(o, zlib, zlen); + o += zlen; + STBIW_FREE(zlib); + stbiw__wpcrc(&o, zlen); + + stbiw__wp32(o,0); + stbiw__wptag(o, "IEND"); + stbiw__wpcrc(&o,0); + + STBIW_ASSERT(o == out + *out_len); + + return out; +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_png(char const *filename, int x, int y, int comp, const void *data, int stride_bytes) +{ + FILE *f; + int len; + unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len); + if (png == NULL) return 0; + + f = stbiw__fopen(filename, "wb"); + if (!f) { STBIW_FREE(png); return 0; } + fwrite(png, 1, len, f); + fclose(f); + STBIW_FREE(png); + return 1; +} +#endif + +STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int stride_bytes) +{ + int len; + unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len); + if (png == NULL) return 0; + func(context, png, len); + STBIW_FREE(png); + return 1; +} + + +/* *************************************************************************** + * + * JPEG writer + * + * This is based on Jon Olick's jo_jpeg.cpp: + * public domain Simple, Minimalistic JPEG writer - http://www.jonolick.com/code.html + */ + +static const unsigned char stbiw__jpg_ZigZag[] = { 0,1,5,6,14,15,27,28,2,4,7,13,16,26,29,42,3,8,12,17,25,30,41,43,9,11,18, + 24,31,40,44,53,10,19,23,32,39,45,52,54,20,22,33,38,46,51,55,60,21,34,37,47,50,56,59,61,35,36,48,49,57,58,62,63 }; + +static void stbiw__jpg_writeBits(stbi__write_context *s, int *bitBufP, int *bitCntP, const unsigned short *bs) { + int bitBuf = *bitBufP, bitCnt = *bitCntP; + bitCnt += bs[1]; + bitBuf |= bs[0] << (24 - bitCnt); + while(bitCnt >= 8) { + unsigned char c = (bitBuf >> 16) & 255; + stbiw__putc(s, c); + if(c == 255) { + stbiw__putc(s, 0); + } + bitBuf <<= 8; + bitCnt -= 8; + } + *bitBufP = bitBuf; + *bitCntP = bitCnt; +} + +static void stbiw__jpg_DCT(float *d0p, float *d1p, float *d2p, float *d3p, float *d4p, float *d5p, float *d6p, float *d7p) { + float d0 = *d0p, d1 = *d1p, d2 = *d2p, d3 = *d3p, d4 = *d4p, d5 = *d5p, d6 = *d6p, d7 = *d7p; + float z1, z2, z3, z4, z5, z11, z13; + + float tmp0 = d0 + d7; + float tmp7 = d0 - d7; + float tmp1 = d1 + d6; + float tmp6 = d1 - d6; + float tmp2 = d2 + d5; + float tmp5 = d2 - d5; + float tmp3 = d3 + d4; + float tmp4 = d3 - d4; + + // Even part + float tmp10 = tmp0 + tmp3; // phase 2 + float tmp13 = tmp0 - tmp3; + float tmp11 = tmp1 + tmp2; + float tmp12 = tmp1 - tmp2; + + d0 = tmp10 + tmp11; // phase 3 + d4 = tmp10 - tmp11; + + z1 = (tmp12 + tmp13) * 0.707106781f; // c4 + d2 = tmp13 + z1; // phase 5 + d6 = tmp13 - z1; + + // Odd part + tmp10 = tmp4 + tmp5; // phase 2 + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + // The rotator is modified from fig 4-8 to avoid extra negations. + z5 = (tmp10 - tmp12) * 0.382683433f; // c6 + z2 = tmp10 * 0.541196100f + z5; // c2-c6 + z4 = tmp12 * 1.306562965f + z5; // c2+c6 + z3 = tmp11 * 0.707106781f; // c4 + + z11 = tmp7 + z3; // phase 5 + z13 = tmp7 - z3; + + *d5p = z13 + z2; // phase 6 + *d3p = z13 - z2; + *d1p = z11 + z4; + *d7p = z11 - z4; + + *d0p = d0; *d2p = d2; *d4p = d4; *d6p = d6; +} + +static void stbiw__jpg_calcBits(int val, unsigned short bits[2]) { + int tmp1 = val < 0 ? -val : val; + val = val < 0 ? val-1 : val; + bits[1] = 1; + while(tmp1 >>= 1) { + ++bits[1]; + } + bits[0] = val & ((1<0)&&(DU[end0pos]==0); --end0pos) { + } + // end0pos = first element in reverse order !=0 + if(end0pos == 0) { + stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB); + return DU[0]; + } + for(i = 1; i <= end0pos; ++i) { + int startpos = i; + int nrzeroes; + unsigned short bits[2]; + for (; DU[i]==0 && i<=end0pos; ++i) { + } + nrzeroes = i-startpos; + if ( nrzeroes >= 16 ) { + int lng = nrzeroes>>4; + int nrmarker; + for (nrmarker=1; nrmarker <= lng; ++nrmarker) + stbiw__jpg_writeBits(s, bitBuf, bitCnt, M16zeroes); + nrzeroes &= 15; + } + stbiw__jpg_calcBits(DU[i], bits); + stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTAC[(nrzeroes<<4)+bits[1]]); + stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits); + } + if(end0pos != 63) { + stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB); + } + return DU[0]; +} + +static int stbi_write_jpg_core(stbi__write_context *s, int width, int height, int comp, const void* data, int quality) { + // Constants that don't pollute global namespace + static const unsigned char std_dc_luminance_nrcodes[] = {0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0}; + static const unsigned char std_dc_luminance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11}; + static const unsigned char std_ac_luminance_nrcodes[] = {0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d}; + static const unsigned char std_ac_luminance_values[] = { + 0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08, + 0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0,0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28, + 0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59, + 0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89, + 0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6, + 0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2, + 0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa + }; + static const unsigned char std_dc_chrominance_nrcodes[] = {0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0}; + static const unsigned char std_dc_chrominance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11}; + static const unsigned char std_ac_chrominance_nrcodes[] = {0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77}; + static const unsigned char std_ac_chrominance_values[] = { + 0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91, + 0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0,0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26, + 0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58, + 0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87, + 0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4, + 0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda, + 0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa + }; + // Huffman tables + static const unsigned short YDC_HT[256][2] = { {0,2},{2,3},{3,3},{4,3},{5,3},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9}}; + static const unsigned short UVDC_HT[256][2] = { {0,2},{1,2},{2,2},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9},{1022,10},{2046,11}}; + static const unsigned short YAC_HT[256][2] = { + {10,4},{0,2},{1,2},{4,3},{11,4},{26,5},{120,7},{248,8},{1014,10},{65410,16},{65411,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {12,4},{27,5},{121,7},{502,9},{2038,11},{65412,16},{65413,16},{65414,16},{65415,16},{65416,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {28,5},{249,8},{1015,10},{4084,12},{65417,16},{65418,16},{65419,16},{65420,16},{65421,16},{65422,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {58,6},{503,9},{4085,12},{65423,16},{65424,16},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {59,6},{1016,10},{65430,16},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {122,7},{2039,11},{65438,16},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {123,7},{4086,12},{65446,16},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {250,8},{4087,12},{65454,16},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {504,9},{32704,15},{65462,16},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {505,9},{65470,16},{65471,16},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {506,9},{65479,16},{65480,16},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1017,10},{65488,16},{65489,16},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1018,10},{65497,16},{65498,16},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2040,11},{65506,16},{65507,16},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {65515,16},{65516,16},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2041,11},{65525,16},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0} + }; + static const unsigned short UVAC_HT[256][2] = { + {0,2},{1,2},{4,3},{10,4},{24,5},{25,5},{56,6},{120,7},{500,9},{1014,10},{4084,12},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {11,4},{57,6},{246,8},{501,9},{2038,11},{4085,12},{65416,16},{65417,16},{65418,16},{65419,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {26,5},{247,8},{1015,10},{4086,12},{32706,15},{65420,16},{65421,16},{65422,16},{65423,16},{65424,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {27,5},{248,8},{1016,10},{4087,12},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{65430,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {58,6},{502,9},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{65438,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {59,6},{1017,10},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{65446,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {121,7},{2039,11},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{65454,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {122,7},{2040,11},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{65462,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {249,8},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{65470,16},{65471,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {503,9},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{65479,16},{65480,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {504,9},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{65488,16},{65489,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {505,9},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{65497,16},{65498,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {506,9},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{65506,16},{65507,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2041,11},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{65515,16},{65516,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {16352,14},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{65525,16},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1018,10},{32707,15},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0} + }; + static const int YQT[] = {16,11,10,16,24,40,51,61,12,12,14,19,26,58,60,55,14,13,16,24,40,57,69,56,14,17,22,29,51,87,80,62,18,22, + 37,56,68,109,103,77,24,35,55,64,81,104,113,92,49,64,78,87,103,121,120,101,72,92,95,98,112,100,103,99}; + static const int UVQT[] = {17,18,24,47,99,99,99,99,18,21,26,66,99,99,99,99,24,26,56,99,99,99,99,99,47,66,99,99,99,99,99,99, + 99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99}; + static const float aasf[] = { 1.0f * 2.828427125f, 1.387039845f * 2.828427125f, 1.306562965f * 2.828427125f, 1.175875602f * 2.828427125f, + 1.0f * 2.828427125f, 0.785694958f * 2.828427125f, 0.541196100f * 2.828427125f, 0.275899379f * 2.828427125f }; + + int row, col, i, k, subsample; + float fdtbl_Y[64], fdtbl_UV[64]; + unsigned char YTable[64], UVTable[64]; + + if(!data || !width || !height || comp > 4 || comp < 1) { + return 0; + } + + quality = quality ? quality : 90; + subsample = quality <= 90 ? 1 : 0; + quality = quality < 1 ? 1 : quality > 100 ? 100 : quality; + quality = quality < 50 ? 5000 / quality : 200 - quality * 2; + + for(i = 0; i < 64; ++i) { + int uvti, yti = (YQT[i]*quality+50)/100; + YTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (yti < 1 ? 1 : yti > 255 ? 255 : yti); + uvti = (UVQT[i]*quality+50)/100; + UVTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (uvti < 1 ? 1 : uvti > 255 ? 255 : uvti); + } + + for(row = 0, k = 0; row < 8; ++row) { + for(col = 0; col < 8; ++col, ++k) { + fdtbl_Y[k] = 1 / (YTable [stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]); + fdtbl_UV[k] = 1 / (UVTable[stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]); + } + } + + // Write Headers + { + static const unsigned char head0[] = { 0xFF,0xD8,0xFF,0xE0,0,0x10,'J','F','I','F',0,1,1,0,0,1,0,1,0,0,0xFF,0xDB,0,0x84,0 }; + static const unsigned char head2[] = { 0xFF,0xDA,0,0xC,3,1,0,2,0x11,3,0x11,0,0x3F,0 }; + const unsigned char head1[] = { 0xFF,0xC0,0,0x11,8,(unsigned char)(height>>8),STBIW_UCHAR(height),(unsigned char)(width>>8),STBIW_UCHAR(width), + 3,1,(unsigned char)(subsample?0x22:0x11),0,2,0x11,1,3,0x11,1,0xFF,0xC4,0x01,0xA2,0 }; + s->func(s->context, (void*)head0, sizeof(head0)); + s->func(s->context, (void*)YTable, sizeof(YTable)); + stbiw__putc(s, 1); + s->func(s->context, UVTable, sizeof(UVTable)); + s->func(s->context, (void*)head1, sizeof(head1)); + s->func(s->context, (void*)(std_dc_luminance_nrcodes+1), sizeof(std_dc_luminance_nrcodes)-1); + s->func(s->context, (void*)std_dc_luminance_values, sizeof(std_dc_luminance_values)); + stbiw__putc(s, 0x10); // HTYACinfo + s->func(s->context, (void*)(std_ac_luminance_nrcodes+1), sizeof(std_ac_luminance_nrcodes)-1); + s->func(s->context, (void*)std_ac_luminance_values, sizeof(std_ac_luminance_values)); + stbiw__putc(s, 1); // HTUDCinfo + s->func(s->context, (void*)(std_dc_chrominance_nrcodes+1), sizeof(std_dc_chrominance_nrcodes)-1); + s->func(s->context, (void*)std_dc_chrominance_values, sizeof(std_dc_chrominance_values)); + stbiw__putc(s, 0x11); // HTUACinfo + s->func(s->context, (void*)(std_ac_chrominance_nrcodes+1), sizeof(std_ac_chrominance_nrcodes)-1); + s->func(s->context, (void*)std_ac_chrominance_values, sizeof(std_ac_chrominance_values)); + s->func(s->context, (void*)head2, sizeof(head2)); + } + + // Encode 8x8 macroblocks + { + static const unsigned short fillBits[] = {0x7F, 7}; + int DCY=0, DCU=0, DCV=0; + int bitBuf=0, bitCnt=0; + // comp == 2 is grey+alpha (alpha is ignored) + int ofsG = comp > 2 ? 1 : 0, ofsB = comp > 2 ? 2 : 0; + const unsigned char *dataR = (const unsigned char *)data; + const unsigned char *dataG = dataR + ofsG; + const unsigned char *dataB = dataR + ofsB; + int x, y, pos; + if(subsample) { + for(y = 0; y < height; y += 16) { + for(x = 0; x < width; x += 16) { + float Y[256], U[256], V[256]; + for(row = y, pos = 0; row < y+16; ++row) { + // row >= height => use last input row + int clamped_row = (row < height) ? row : height - 1; + int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp; + for(col = x; col < x+16; ++col, ++pos) { + // if col >= width => use pixel from last input column + int p = base_p + ((col < width) ? col : (width-1))*comp; + float r = dataR[p], g = dataG[p], b = dataB[p]; + Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128; + U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b; + V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b; + } + } + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+0, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+8, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+128, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+136, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + + // subsample U,V + { + float subU[64], subV[64]; + int yy, xx; + for(yy = 0, pos = 0; yy < 8; ++yy) { + for(xx = 0; xx < 8; ++xx, ++pos) { + int j = yy*32+xx*2; + subU[pos] = (U[j+0] + U[j+1] + U[j+16] + U[j+17]) * 0.25f; + subV[pos] = (V[j+0] + V[j+1] + V[j+16] + V[j+17]) * 0.25f; + } + } + DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subU, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT); + DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subV, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT); + } + } + } + } else { + for(y = 0; y < height; y += 8) { + for(x = 0; x < width; x += 8) { + float Y[64], U[64], V[64]; + for(row = y, pos = 0; row < y+8; ++row) { + // row >= height => use last input row + int clamped_row = (row < height) ? row : height - 1; + int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp; + for(col = x; col < x+8; ++col, ++pos) { + // if col >= width => use pixel from last input column + int p = base_p + ((col < width) ? col : (width-1))*comp; + float r = dataR[p], g = dataG[p], b = dataB[p]; + Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128; + U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b; + V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b; + } + } + + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y, 8, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, U, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT); + DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, V, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT); + } + } + } + + // Do the bit alignment of the EOI marker + stbiw__jpg_writeBits(s, &bitBuf, &bitCnt, fillBits); + } + + // EOI + stbiw__putc(s, 0xFF); + stbiw__putc(s, 0xD9); + + return 1; +} + +STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_jpg_core(&s, x, y, comp, (void *) data, quality); +} + + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_jpg_core(&s, x, y, comp, data, quality); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif + +#endif // STB_IMAGE_WRITE_IMPLEMENTATION + +/* Revision history + 1.16 (2021-07-11) + make Deflate code emit uncompressed blocks when it would otherwise expand + support writing BMPs with alpha channel + 1.15 (2020-07-13) unknown + 1.14 (2020-02-02) updated JPEG writer to downsample chroma channels + 1.13 + 1.12 + 1.11 (2019-08-11) + + 1.10 (2019-02-07) + support utf8 filenames in Windows; fix warnings and platform ifdefs + 1.09 (2018-02-11) + fix typo in zlib quality API, improve STB_I_W_STATIC in C++ + 1.08 (2018-01-29) + add stbi__flip_vertically_on_write, external zlib, zlib quality, choose PNG filter + 1.07 (2017-07-24) + doc fix + 1.06 (2017-07-23) + writing JPEG (using Jon Olick's code) + 1.05 ??? + 1.04 (2017-03-03) + monochrome BMP expansion + 1.03 ??? + 1.02 (2016-04-02) + avoid allocating large structures on the stack + 1.01 (2016-01-16) + STBIW_REALLOC_SIZED: support allocators with no realloc support + avoid race-condition in crc initialization + minor compile issues + 1.00 (2015-09-14) + installable file IO function + 0.99 (2015-09-13) + warning fixes; TGA rle support + 0.98 (2015-04-08) + added STBIW_MALLOC, STBIW_ASSERT etc + 0.97 (2015-01-18) + fixed HDR asserts, rewrote HDR rle logic + 0.96 (2015-01-17) + add HDR output + fix monochrome BMP + 0.95 (2014-08-17) + add monochrome TGA output + 0.94 (2014-05-31) + rename private functions to avoid conflicts with stb_image.h + 0.93 (2014-05-27) + warning fixes + 0.92 (2010-08-01) + casts to unsigned char to fix warnings + 0.91 (2010-07-17) + first public release + 0.90 first internal release +*/ + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/qoi/genc/test-genc.sh b/qoi/genc/test-genc.sh new file mode 100755 index 00000000..cbf3d5c5 --- /dev/null +++ b/qoi/genc/test-genc.sh @@ -0,0 +1,27 @@ +#!/bin/bash + +make genc-qoiconv +mkdir -p ../output + +# Encoding (decoding .png into raw pixels and encoding these raw pixels to .qoi) +for i in ../images/*.png; do + [ -f "$i" ] || break + filename=$(basename -- "$i") + filename="${filename%.*}" + + ./genc-qoiconv "$i" "../output/$filename.qoi" + cmp -l "../images/$filename.qoi" "../output/$filename.qoi" + if [[ "$?" = 0 ]]; then + echo "../images/$filename.qoi and ../output/$filename.qoi are the same" + fi + # else: cmp -l will print a message saying otherwise +done + +# Decoding (decoding .qoi into raw pixels and encoding these raw pixels into .png) +for i in ../images/*.qoi; do + [ -f "$i" ] || break + filename=$(basename -- "$i") + filename="${filename%.*}" + + ./genc-qoiconv "../images/$filename.qoi" "../output/$filename.png" +done \ No newline at end of file diff --git a/qoi/images/Central_Bern_from_north.png b/qoi/images/Central_Bern_from_north.png new file mode 100644 index 00000000..d763f53b Binary files /dev/null and b/qoi/images/Central_Bern_from_north.png differ diff --git a/qoi/images/Central_Bern_from_north.qoi b/qoi/images/Central_Bern_from_north.qoi new file mode 100644 index 00000000..6dc0dacf Binary files /dev/null and b/qoi/images/Central_Bern_from_north.qoi differ diff --git a/qoi/images/Chocolate_Hills_overview.png b/qoi/images/Chocolate_Hills_overview.png new file mode 100644 index 00000000..322abf0c Binary files /dev/null and b/qoi/images/Chocolate_Hills_overview.png differ diff --git a/qoi/images/Chocolate_Hills_overview.qoi b/qoi/images/Chocolate_Hills_overview.qoi new file mode 100644 index 00000000..006f3ef2 Binary files /dev/null and b/qoi/images/Chocolate_Hills_overview.qoi differ diff --git a/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.png b/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.png new file mode 100644 index 00000000..c6d30989 Binary files /dev/null and b/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.png differ diff --git a/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.qoi b/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.qoi new file mode 100644 index 00000000..05bf9ed1 Binary files /dev/null and b/qoi/images/Eyjafjallajokull_sous_les_aurores_boreales.qoi differ diff --git a/qoi/images/README.md b/qoi/images/README.md new file mode 100644 index 00000000..9244fd2c --- /dev/null +++ b/qoi/images/README.md @@ -0,0 +1,28 @@ +# Images accreditation + +(listed in alphabetical order) + +## `Central_Bern_from_north` +* Retrieved from https://commons.wikimedia.org/wiki/File:Central_Bern_from_north.jpg +* Author: [CucombreLibre](https://www.flickr.com/people/33200530@N04) +* License: [Creative Commons](https://en.wikipedia.org/wiki/en:Creative_Commons) [Attribution 2.0 Generic license](https://creativecommons.org/licenses/by/2.0/deed.en) +* Modifications: + * downscaled and exported from `.jpg` to `.png` through GIMP 2.10 + * exported from `.png` to `.qoi` through [`qoiconv`](https://github.com/phoboslab/qoi/blob/master/qoiconv.c) + +## `Chocolate_Hills_overview` +* Retrieved from https://commons.wikimedia.org/wiki/File:Chocolate_Hills_overview.JPG +* Author: [P199](https://commons.wikimedia.org/wiki/User:P199) +* License: [Creative Commons](https://en.wikipedia.org/wiki/en:Creative_Commons) [Attribution-Share Alike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/deed.en) +* Modifications: + * downscaled and exported from `.jpg` to `.png` through GIMP 2.10 + * exported from `.png` to `.qoi` through [`qoiconv`](https://github.com/phoboslab/qoi/blob/master/qoiconv.c) + +## `Eyjafjallajokull_sous_les_aurores_boreales` +* Retrieved from https://commons.wikimedia.org/wiki/File:L%27Eyjafjallaj%C3%B6kull_sous_les_aurores_bor%C3%A9ales.jpg +* Author: [Sébastien Giguère](https://commons.wikimedia.org/wiki/User:Thaumazein1) +* License: [Creative Commons](https://en.wikipedia.org/wiki/en:Creative_Commons) [Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/deed.en) +* Modifications: + * renamed from `L'Eyjafjallajökull_sous_les_aurores_boréales` to `Eyjafjallajokull_sous_les_aurores_boreales` + * downscaled and exported from `.jpg` to `.png` through GIMP 2.10 + * exported from `.png` to `.qoi` through [`qoiconv`](https://github.com/phoboslab/qoi/blob/master/qoiconv.c) diff --git a/qoi/stainless.conf b/qoi/stainless.conf new file mode 100644 index 00000000..1173bde7 --- /dev/null +++ b/qoi/stainless.conf @@ -0,0 +1,6 @@ +vc-cache = true +timeout = 180 +strict-arithmetic = false +batched = true +solvers = "smt-cvc4,smt-z3,no-inc:smt-z3:z3 tactic.default_tactic=smt sat.euf=true" +infer-measures = false \ No newline at end of file diff --git a/qoi/unannotated/common.scala b/qoi/unannotated/common.scala new file mode 100644 index 00000000..12c6a664 --- /dev/null +++ b/qoi/unannotated/common.scala @@ -0,0 +1,54 @@ +object common { + val OpIndex = 0x00 + val OpDiff = 0x40 + val OpLuma = 0x80 + val OpRun = 0xc0 + val OpRgb = 0xfe + val OpRgba = 0xff + + val Mask2 = 0xc0 + + val MagicNumber = 1903126886 + val HeaderSize = 14 + val Padding = 8 + + val MaxWidth = 8192 + val MaxHeight = 8192 + + object Pixel { + def r(px: Int): Byte = ((px >>> 24) & 0xff).toByte + def g(px: Int): Byte = ((px >>> 16) & 0xff).toByte + def b(px: Int): Byte = ((px >>> 8) & 0xff).toByte + def a(px: Int): Byte = (px & 0xff).toByte + + def fromRgba(r: Byte, g: Byte, b: Byte, a: Byte): Int = + (r << 24) | ((g << 16) & 0xffffff) | ((b << 8) & 0xffff) | (a.toInt & 0xff) + + def incremented(px: Int)(r: Byte = 0, g: Byte = 0, b: Byte = 0, a: Byte = 0): Int = + fromRgba(((Pixel.r(px) + r) & 0xff).toByte, ((Pixel.g(px) + g) & 0xff).toByte, ((Pixel.b(px) + b) & 0xff).toByte, ((Pixel.a(px) + a) & 0xff).toByte) + + def withRgba(px: Int)(r: Byte = Pixel.r(px), g: Byte = Pixel.g(px), b: Byte = Pixel.b(px), a: Byte = Pixel.a(px)): Int = + fromRgba(r, g, b, a) + } + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + def colorPos(px: Int): Int = + ((Pixel.r(px) & 0xff) * 3 + (Pixel.g(px) & 0xff) * 5 + (Pixel.b(px) & 0xff) * 7 + (Pixel.a(px) & 0xff) * 11) % 64 + + def write16(data: Array[Byte], i: Int, value: Short): Unit = { + data(i) = ((0xff00 & value) >>> 8).toByte + data(i + 1) = (0xff & value).toByte + } + + def write32(data: Array[Byte], i: Int, value: Int): Unit = { + write16(data, i, (value >>> 16).toShort) + write16(data, i + 2, value.toShort) + } + + def read16(data: Array[Byte], i: Int): Short = + (((((data(i) & 0xff) << 8) & 0xffff) | (data(i + 1) & 0xff)) & 0xffff).toShort + + def read32(data: Array[Byte], i: Int): Int = + (read16(data, i) << 16) | (read16(data, i + 2) & 0xffff) +} diff --git a/qoi/unannotated/decoder.scala b/qoi/unannotated/decoder.scala new file mode 100644 index 00000000..fbae46c2 --- /dev/null +++ b/qoi/unannotated/decoder.scala @@ -0,0 +1,169 @@ +import common.* + +object decoder { + + case class Ctx(bytes: Array[Byte], w: Long, h: Long, chan: Long) { + def pixelsLen: Long = w * h * chan + + def chunksLen: Long = bytes.length - Padding + } + + case class DecodedResult(pixels: Array[Byte], w: Long, h: Long, chan: Long) + + case class WriteRunPixelsResult(remainingRun: Long, pxPos: Long) + + sealed trait DecodedNext + case class Run(run: Long) extends DecodedNext + case class DiffOrIndexOrColor(px: Int) extends DecodedNext + + case class DecodingIteration(px: Int, inPos: Long, pxPos: Long, remainingRun: Long) + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + def decode(bytes: Array[Byte], to: Long): Option[DecodedResult] = { + if (!(bytes.length > HeaderSize + Padding && HeaderSize + Padding < to && to <= bytes.length)) { + None + } else { + decodeHeader(bytes).map { + case (w, h, chan) => + val index = Array.fill(64)(0) + val pixels = Array.fill(w.toInt * h.toInt * chan.toInt)(0: Byte) + val px = Pixel.fromRgba(0, 0, 0, 255.toByte) + val decIter = decodeLoop(index, pixels, px, HeaderSize, to - Padding, 0)(using Ctx(bytes, w, h, chan)) + if (decIter.pxPos != pixels.length) { + writeRemainingPixels(pixels, decIter.px, decIter.pxPos)(using Ctx(bytes, w, h, chan)) + } + DecodedResult(pixels, w, h, chan) + } + } + } + + def writeRemainingPixels(pixels: Array[Byte], pxPrev: Int, pxPos: Long)(using Ctx): Unit = { + writePixel(pixels, pxPrev, pxPos) + if (pxPos + chan < pixels.length) { + writeRemainingPixels(pixels, pxPrev, pxPos + chan) + } + } + + def decodeLoop(index: Array[Int], pixels: Array[Byte], pxPrev: Int, + inPos0: Long, untilInPos: Long, pxPos0: Long)(using Ctx): DecodingIteration = { + val (res, decIter) = decodeNext(index, pixels, pxPrev, inPos0, pxPos0) + + if (decIter.inPos < untilInPos && decIter.pxPos + chan <= pixels.length) { + decodeLoop(index, pixels, decIter.px, decIter.inPos, untilInPos, decIter.pxPos) + } else { + decIter + } + } + + def decodeNext(index: Array[Int], pixels: Array[Byte], pxPrev: Int, inPos0: Long, pxPos0: Long)(using Ctx): (DecodedNext, DecodingIteration) = { + val (decRes, inPos) = doDecodeNext(index, pxPrev, inPos0) + + decRes match { + case Run(run) => + val WriteRunPixelsResult(resRun, resPxPos) = writeRunPixels(pixels, pxPrev, run, pxPos0) + (decRes, DecodingIteration(pxPrev, inPos, resPxPos, resRun)) + + case DiffOrIndexOrColor(px) => + writePixel(pixels, px, pxPos0) + index(colorPos(px)) = px + (decRes, DecodingIteration(px, inPos, pxPos0 + chan, 0)) + } + } + + def doDecodeNext(index: Array[Int], pxPrev: Int, inPos0: Long)(using Ctx): (DecodedNext, Long) = { + var px = pxPrev + var inPos = inPos0 + var run = 0L + + val b1 = bytes(inPos.toInt).toInt & 0xff + inPos += 1 + val res: DecodedNext = if (b1 == OpRgb) { + val px = Pixel.withRgba(pxPrev)(r = bytes(inPos.toInt), g = bytes(inPos.toInt + 1), b = bytes(inPos.toInt + 2)) + inPos += 3 + DiffOrIndexOrColor(px) + } else if (b1 == OpRgba) { + val px = Pixel.withRgba(pxPrev)(r = bytes(inPos.toInt), g = bytes(inPos.toInt + 1), b = bytes(inPos.toInt + 2), a = bytes(inPos.toInt + 3)) + inPos += 4 + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpIndex) { + val px = index(b1) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpDiff) { + val px = decodeDiff(pxPrev, b1) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpLuma) { + val b2 = bytes(inPos.toInt).toInt & 0xff + inPos += 1 + val px = decodeLuma(pxPrev, b1, b2) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpRun) { + val run = decodeRun(b1) + Run(run) + } else { + DiffOrIndexOrColor(pxPrev) + } + (res, inPos) + } + + def decodeHeader(bytes: Array[Byte]): Option[(Long, Long, Long)] = { + val magic = read32(bytes, 0) + val w = read32(bytes, 4) + val h = read32(bytes, 8) + val chan = bytes(12) + + if (0 < w && w <= MaxWidth && 0 < h && h <= MaxHeight && magic == MagicNumber && 3 <= chan && chan <= 4) + Some((w.toLong, h.toLong, chan.toLong)) + else + None + } + + def writeRunPixels(pixels: Array[Byte], px: Int, run0: Long, pxPos0: Long)(using Ctx): WriteRunPixelsResult = { + writePixel(pixels, px, pxPos0) + if (run0 > 0 && pxPos0 + chan < pixels.length) { + writeRunPixels(pixels, px, run0 - 1, pxPos0 + chan) + } else { + WriteRunPixelsResult(run0, pxPos0 + chan) + } + } + + def writePixel(pixels: Array[Byte], px: Int, pxPos: Long)(using Ctx): Unit = { + pixels(pxPos.toInt) = Pixel.r(px) + pixels(pxPos.toInt + 1) = Pixel.g(px) + pixels(pxPos.toInt + 2) = Pixel.b(px) + if (chan == 4) { + pixels(pxPos.toInt + 3) = Pixel.a(px) + } + } + + def decodeDiff(pxPrev: Int, b1: Int): Int = { + Pixel.incremented(pxPrev)( + (((b1 >>> 4) & 0x03) - 2).toByte, + (((b1 >>> 2) & 0x03) - 2).toByte, + ((b1 & 0x03) - 2).toByte) + } + + def decodeLuma(pxPrev: Int, b1: Int, b2: Int): Int = { + val vg = (b1 & 0x3f) - 32 + Pixel.incremented(pxPrev)( + (vg - 8 + ((b2 >>> 4) & 0x0f)).toByte, + vg.toByte, + (vg - 8 + (b2 & 0x0f)).toByte) + } + + def decodeRun(b1: Int): Int = b1 & 0x3f + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + def w(using ctx: Ctx): Long = ctx.w + + def h(using ctx: Ctx): Long = ctx.h + + def chan(using ctx: Ctx): Long = ctx.chan + + def pixelsLen(using ctx: Ctx): Long = ctx.pixelsLen + + def chunksLen(using ctx: Ctx): Long = ctx.chunksLen + + def bytes(using ctx: Ctx): Array[Byte] = ctx.bytes +} diff --git a/qoi/unannotated/encoder.scala b/qoi/unannotated/encoder.scala new file mode 100644 index 00000000..b4b9f7b1 --- /dev/null +++ b/qoi/unannotated/encoder.scala @@ -0,0 +1,180 @@ +import common.* + +object encoder { + + case class Ctx(pixels: Array[Byte], w: Long, h: Long, chan: Long) { + def maxSize: Long = w * h * (chan + 1) + HeaderSize + Padding + + def pxEnd: Long = pixels.length - chan + } + + case class EncodedResult(encoded: Array[Byte], length: Long) + + case class EncodeSingleStepResult(px: Int, outPos: Long, run: Long) + + case class EncodingIteration(px: Int, outPos: Long) + + case class RunUpdate(reset: Boolean, run: Long, outPos: Long) + + case class LoopIter(px: Int, pxPrev: Int, pxPos: Long) + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + def encode(pixels: Array[Byte], w: Long, h: Long, chan: Long): Option[EncodedResult] = { + if (!(0 < w && w <= MaxWidth && + 0 < h && h <= MaxHeight && + 3 <= chan && chan <= 4 && + w * h * chan == pixels.length)) + None + else doEncode(using Ctx(pixels, w, h, chan)) + } + + def doEncode(using Ctx): Option[EncodedResult] = { + val bytes = Array.fill(maxSize.toInt)(0: Byte) + writeHeader(bytes) + val index = Array.fill(64)(0) + val pxPrev = Pixel.fromRgba(0, 0, 0, 255.toByte) + val EncodingIteration(pxRes, outPos) = encodeLoop(index, bytes, pxPrev, 0, HeaderSize, 0) + Some(EncodedResult(bytes, outPos + Padding)) + } + + def encodeLoop(index: Array[Int], bytes: Array[Byte], pxPrev: Int, run0: Long, outPos0: Long, pxPos: Long)(using Ctx): EncodingIteration = { + val EncodeSingleStepResult(px, outPos2, run1) = encodeSingleStep(index, bytes, pxPrev, run0, outPos0, pxPos) + + if (pxPos + chan < pixels.length) { + encodeLoop(index, bytes, px, run1, outPos2, pxPos + chan) + } else { + bytes(outPos2.toInt + Padding - 1) = 1 + EncodingIteration(px, outPos2) + } + } + + + def writeHeader(bytes: Array[Byte])(using Ctx): Unit = { + write32(bytes, 0, MagicNumber) + assert(read32(bytes, 0) == MagicNumber) + + write32(bytes, 4, w.toInt) + assert(read32(bytes, 4) == w.toInt) + + write32(bytes, 8, h.toInt) + assert(read32(bytes, 8) == h.toInt) + + bytes(12) = chan.toByte + assert(bytes(12) == chan.toByte) + + bytes(13) = 0 // Color-space (unused) + } + + def encodeSingleStep(index: Array[Int], bytes: Array[Byte], pxPrev: Int, run0: Long, outPos0: Long, pxPos: Long)(using Ctx): EncodeSingleStepResult = { + val px = + if (chan == 4) read32(pixels, pxPos.toInt) + else Pixel.fromRgba(pixels(pxPos.toInt), pixels(pxPos.toInt + 1), pixels(pxPos.toInt + 2), Pixel.a(pxPrev)) + given li: LoopIter = LoopIter(px, pxPrev, pxPos) + + val runUpd = updateRun(bytes, run0, outPos0) + val run1 = runUpd.run + val outPos1 = runUpd.outPos + + val outPos2 = if (px != pxPrev) { + encodeNoRun(index, bytes, outPos1) + } else { + outPos1 + } + EncodeSingleStepResult(px, outPos2, run1) + } + + def updateRun(bytes: Array[Byte], run0: Long, outPos0: Long)(using Ctx, LoopIter): RunUpdate = { + var run = run0 + var outPos = outPos0 + + if (px == pxPrev) + run += 1 + + var runReset = false + if (run > 0 && (run == 62 || px != pxPrev || pxPos == pxEnd)) { + val b1 = (OpRun | (run - 1)).toByte + bytes(outPos.toInt) = b1 + outPos += 1 + run = 0 + runReset = true + } + RunUpdate(runReset, run, outPos) + } + + def encodeNoRun(index: Array[Int], bytes: Array[Byte], outPos1: Long)(using Ctx, LoopIter): Long = { + val indexPos = colorPos(px) + var newOutPos = outPos1 + + if (index(indexPos) == px) { + val b1 = (OpIndex | indexPos) & 0xff + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + } else { + index(indexPos) = px + + if (Pixel.a(px) == Pixel.a(pxPrev)) { + // Note: these 5 variables are declared as signed char in the reference implementation + val vr = ((Pixel.r(px).toInt & 0xff) - (Pixel.r(pxPrev).toInt & 0xff)).toByte + val vg = ((Pixel.g(px).toInt & 0xff) - (Pixel.g(pxPrev).toInt & 0xff)).toByte + val vb = ((Pixel.b(px).toInt & 0xff) - (Pixel.b(pxPrev).toInt & 0xff)).toByte + val vgR = (vr - vg).toByte + val vgB = (vb - vg).toByte + + if (vr > -3 && vr < 2 && vg > -3 && vg < 2 && vb > -3 && vb < 2) { + val b1 = OpDiff | (((vr + 2) << 4) & 0xff) | (((vg + 2) << 2) & 0xff) | ((vb + 2) & 0xff) + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + } else if (vgR > -9 && vgR < 8 && vg > -33 && vg < 32 && vgB > -9 && vgB < 8) { + val b1 = OpLuma | ((vg + 32) & 0xff) + val b2 = (((vgR + 8) << 4) & 0xff) | ((vgB + 8) & 0xff) + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = b2.toByte + newOutPos += 1 + } else { + bytes(newOutPos.toInt) = OpRgb.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.r(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.g(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.b(px) + newOutPos += 1 + } + } else { + bytes(newOutPos.toInt) = OpRgba.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.r(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.g(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.b(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.a(px) + newOutPos += 1 + } + } + + newOutPos + } + + def maxSize(using ctx: Ctx): Long = ctx.maxSize + + def pxEnd(using ctx: Ctx): Long = ctx.pxEnd + + def w(using ctx: Ctx): Long = ctx.w + + def h(using ctx: Ctx): Long = ctx.h + + def chan(using ctx: Ctx): Long = ctx.chan + + def pixels(using ctx: Ctx): Array[Byte] = ctx.pixels + + def px(using li: LoopIter): Int = li.px + + def pxPrev(using li: LoopIter): Int = li.pxPrev + + def pxPos(using li: LoopIter): Long = li.pxPos +} diff --git a/qoi/verified/common.scala b/qoi/verified/common.scala new file mode 100644 index 00000000..db64f43a --- /dev/null +++ b/qoi/verified/common.scala @@ -0,0 +1,658 @@ +import stainless.* +import stainless.lang.* +import stainless.collection.* +import stainless.annotation.{cCode, extern, ghost, ignore, inlineOnce, mutable, opaque, pure} +import stainless.proof.* +import StaticChecks.* + +object common { + val OpIndex = 0x00 + val OpDiff = 0x40 + val OpLuma = 0x80 + val OpRun = 0xc0 + val OpRgb = 0xfe + val OpRgba = 0xff + + val Mask2 = 0xc0 + + val MagicNumber = 1903126886 + val HeaderSize = 14 + val Padding = 8 + + val MaxWidth = 8192 + val MaxHeight = 8192 + + object Pixel { + def r(px: Int): Byte = ((px >>> 24) & 0xff).toByte + def g(px: Int): Byte = ((px >>> 16) & 0xff).toByte + def b(px: Int): Byte = ((px >>> 8) & 0xff).toByte + def a(px: Int): Byte = (px & 0xff).toByte + + def fromRgba(r: Byte, g: Byte, b: Byte, a: Byte): Int = + (r << 24) | ((g << 16) & 0xffffff) | ((b << 8) & 0xffff) | (a.toInt & 0xff) + + def incremented(px: Int)(r: Byte = 0, g: Byte = 0, b: Byte = 0, a: Byte = 0): Int = + fromRgba(((Pixel.r(px) + r) & 0xff).toByte, ((Pixel.g(px) + g) & 0xff).toByte, ((Pixel.b(px) + b) & 0xff).toByte, ((Pixel.a(px) + a) & 0xff).toByte) + + def withRgba(px: Int)(r: Byte = Pixel.r(px), g: Byte = Pixel.g(px), b: Byte = Pixel.b(px), a: Byte = Pixel.a(px)): Int = + fromRgba(r, g, b, a) + } + + @cCode.noMangling + sealed trait OptionMut[@mutable T] + @cCode.noMangling + case class SomeMut[@mutable T](v: T) extends OptionMut[T] + @cCode.noMangling + case class NoneMut[@mutable T]() extends OptionMut[T] + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @cCode.function( + code = + """array_int8 __FUNCTION__(int32_t length) { + | int8_t* data = malloc(length); + | if (!data) { + | exit(-1); + | } + | memset(data, 0, length); + | return (array_int8) { .data = data, .length = length }; + |}""", + headerIncludes = "stdlib.h", + cIncludes = "" + ) + def allocArray(size: Int): Array[Byte] = { + Array.fill(size)(0: Byte) + }.ensuring(_.length == size) + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + def colorPos(px: Int): Int = { + ((Pixel.r(px) & 0xff) * 3 + (Pixel.g(px) & 0xff) * 5 + (Pixel.b(px) & 0xff) * 7 + (Pixel.a(px) & 0xff) * 11) % 64 + }.ensuring(x => x >= 0 && x < 64) + + def write16(data: Array[Byte], i: Int, value: Short): Unit = { + require(data.length >= 2 && i >= 0 && i < data.length - 1) + data(i) = ((0xff00 & value) >>> 8).toByte + data(i + 1) = (0xff & value).toByte + } ensuring(_ => read16(data, i) == value && old(data).length == data.length) + + def write32(data: Array[Byte], i: Int, value: Int): Unit = { + require(data.length >= 4 && i >= 0 && i < data.length - 3) + write16(data, i, (value >>> 16).toShort) + write16(data, i + 2, value.toShort) + } ensuring(_ => read32(data, i) == value && old(data).length == data.length) + + def read16(data: Array[Byte], i: Int): Short = { + require(data.length >= 2 && i >= 0 && i < data.length - 1) + (((((data(i) & 0xff) << 8) & 0xffff) | (data(i + 1) & 0xff)) & 0xffff).toShort + } + + def read32(data: Array[Byte], i: Int): Int = { + require(data.length >= 4 && i >= 0 && i < data.length - 3) + (read16(data, i) << 16) | (read16(data, i + 2) & 0xffff) + } + + @inline + def bool2int(b: Boolean): Int = if (b) 1 else 0 + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + @opaque + @inlineOnce + def modSumLemma(x: Long, chan: Long): Unit = { + require(0 <= x && x <= Long.MaxValue - chan) + require(3 <= chan && chan <= 4) + require(x % chan == 0) + () + }.ensuring(_ => (x + chan) % chan == 0) + + @ghost + @opaque + @inlineOnce + def modMultLemma(a: Long, b: Long, chan: Long): Unit = { + require(3 <= chan && chan <= 4) + require(0 <= a && a <= Short.MaxValue) + require(0 <= b && b <= Short.MaxValue) + () + }.ensuring(_ => (a * b * chan) % chan == 0) + + @ghost + @opaque + @inlineOnce + def modLeqLemma(x: Long, y: Long, chan: Long): Unit = { + require(y >= 0) + require(0 <= x && x < y) + require(3 <= chan && chan <= 4) + require(x % chan == 0) + require(y % chan == 0) + () + }.ensuring(_ => x + chan <= y) + + @ghost + @opaque + @inlineOnce + def modLtLemma(x: Long, y: Long, chan: Long): Unit = { + require(y >= 0) + require(3 <= chan && chan <= 4) + require(0 <= x && x <= Long.MaxValue - chan) + require(x + chan <= y) + require(x % chan == 0) + require(y % chan == 0) + () + }.ensuring(_ => x < y) + + @ghost + @pure + def samePixels(pixels: Array[Byte], px: Int, pxPos: Long, chan: Long): Boolean = { + require(3 <= chan && chan <= 4) + require(pixels.length % chan == 0) + require(0 <= pxPos && pxPos < pixels.length) + require(pxPos % chan == 0) + + modLeqLemma(pxPos, pixels.length, chan) + assert(pxPos + chan <= pixels.length) + + pixels(pxPos.toInt) == Pixel.r(px) && + pixels(pxPos.toInt + 1) == Pixel.g(px) && + pixels(pxPos.toInt + 2) == Pixel.b(px) && + ((chan == 4) ==> (pixels(pxPos.toInt + 3) == Pixel.a(px))) + } + + @ghost + @pure + @opaque + @inlineOnce + def samePixelsSingleElementRange(pixels: Array[Byte], px: Int, pxPos: Long, chan: Long): Unit = { + require(3 <= chan && chan <= 4) + require(0 <= pxPos && pxPos < pixels.length) + require(pxPos % chan == 0 && pixels.length % chan == 0) + require(samePixels(pixels, px, pxPos, chan)) + }.ensuring(_ => samePixelsForall(pixels, px, pxPos, pxPos + chan, chan)) + + @ghost + @pure + @opaque + @inlineOnce + def samePixelsExtendedRangeLemma(pixels: Array[Byte], px: Int, pxPosStart: Long, pxPosEnd: Long, chan: Long): Unit = { + decreases(pxPosEnd - pxPosStart) + require(3 <= chan && chan <= 4) + require(pixels.length % chan == 0) + require(0 <= pxPosStart && pxPosStart < pxPosEnd) + require(pxPosEnd < pixels.length) + require(pxPosStart % chan == 0) + require(pxPosEnd % chan == 0) + require(samePixelsForall(pixels, px, pxPosStart, pxPosEnd, chan)) + require(samePixels(pixels, px, pxPosEnd, chan)) + + modSumLemma(pxPosEnd, chan) + modLeqLemma(pxPosStart, pxPosEnd, chan) + modLeqLemma(pxPosEnd, pixels.length, chan) + assert((pxPosStart + chan) % chan == 0) + assert(pxPosStart + chan <= pxPosEnd) + assert(pxPosEnd + chan <= pixels.length) + + if (pxPosStart + chan == pxPosEnd) { + check(samePixelsForall(pixels, px, pxPosStart, pxPosEnd + chan, chan)) + } else { + samePixelsExtendedRangeLemma(pixels, px, pxPosStart + chan, pxPosEnd, chan) + assert(samePixelsForall(pixels, px, pxPosStart + chan, pxPosEnd + chan, chan)) + check(samePixelsForall(pixels, px, pxPosStart, pxPosEnd + chan, chan)) + } + }.ensuring { _ => + modSumLemma(pxPosEnd, chan) + modLeqLemma(pxPosEnd, pixels.length, chan) + samePixelsForall(pixels, px, pxPosStart, pxPosEnd + chan, chan) + } + + @ghost + @pure + def samePixelsForall(pixels: Array[Byte], px: Int, pxPosStart: Long, pxPosEnd: Long, chan: Long): Boolean = { + decreases(pxPosEnd - pxPosStart) + require(3 <= chan && chan <= 4) + require(pixels.length % chan == 0) + require(0 <= pxPosStart && pxPosStart < pxPosEnd) + require(pxPosEnd <= pixels.length) + require(pxPosEnd % chan == 0) + require(pxPosStart % chan == 0) + + modLeqLemma(pxPosStart, pxPosEnd, chan) + assert(pxPosStart + chan <= pxPosEnd) + + samePixels(pixels, px, pxPosStart, chan) && + (if (pxPosStart + chan == pxPosEnd) true + else { + assert(pxPosStart + chan <= pxPosEnd) + modSumLemma(pxPosStart, chan) + assert((pxPosStart + chan) % chan == 0) + assert(pixels.length % chan == 0) + samePixelsForall(pixels, px, pxPosStart + chan, pxPosEnd, chan) + }) + // forall[Int](i => (pxPosStart / chan <= i && i < pxPosEnd / chan) ==> samePixels(pixels, px, chan * i)) + } + + @ghost + @pure + @opaque + @inlineOnce + def samePixelsForallCombinedLemma(pixels: Array[Byte], px: Int, pxPosStart: Long, pxMiddle: Long, pxPosEnd: Long, chan: Long): Unit = { + decreases(pxPosEnd - pxPosStart) + require(3 <= chan && chan <= 4) + require(0 <= pxPosStart && pxPosStart < pxMiddle && pxMiddle < pxPosEnd) + require(pixels.length % chan == 0) + require(pxPosEnd <= pixels.length) + require(pxPosEnd % chan == 0) + require(pxMiddle % chan == 0) + require(pxPosStart % chan == 0) + require(samePixelsForall(pixels, px, pxPosStart, pxMiddle, chan)) + require(samePixelsForall(pixels, px, pxMiddle, pxPosEnd, chan)) + + if (pxMiddle == pxPosStart + chan) { + check(samePixelsForall(pixels, px, pxPosStart, pxPosEnd, chan)) + } else { + assert(samePixelsForall(pixels, px, pxPosStart + chan, pxMiddle, chan)) + samePixelsForallCombinedLemma(pixels, px, pxPosStart + chan, pxMiddle, pxPosEnd, chan) + assert(samePixelsForall(pixels, px, pxPosStart + chan, pxPosEnd, chan)) + check(samePixelsForall(pixels, px, pxPosStart, pxPosEnd, chan)) + } + }.ensuring(_ => samePixelsForall(pixels, px, pxPosStart, pxPosEnd, chan)) + + @ghost + @pure + @opaque + @inlineOnce + def samePixelsForallArraysEq(pixels1: Array[Byte], pixels2: Array[Byte], px: Int, pxPosStart: Long, pxPosEnd: Long, chan: Long): Unit = { + decreases(pxPosEnd - pxPosStart) + require(3 <= chan && chan <= 4) + require(0 <= pxPosStart && pxPosStart < pxPosEnd && pxPosEnd <= pixels1.length) + require(pixels1.length % chan == 0) + require(pxPosStart % chan == 0) + require(pxPosEnd % chan == 0) + require(pixels1.length == pixels2.length) + require(samePixelsForall(pixels1, px, pxPosStart, pxPosEnd, chan)) + require(samePixelsForall(pixels2, px, pxPosStart, pxPosEnd, chan)) + + assert(samePixels(pixels1, px, pxPosStart, chan)) + assert(samePixels(pixels2, px, pxPosStart, chan)) + if (pxPosStart + chan == pxPosEnd) { + check(arraysEq(pixels1, pixels2, pxPosStart, pxPosEnd)) + } else { + modSumLemma(pxPosStart, chan) + modLtLemma(pxPosStart, pxPosEnd, chan) + samePixelsForallArraysEq(pixels1, pixels2, px, pxPosStart + chan, pxPosEnd, chan) + assert(arraysEq(pixels1, pixels2, pxPosStart + chan, pxPosEnd)) + assert(arraysEq(pixels1, pixels2, pxPosStart, pxPosStart + chan)) + arraysEqCombinedLemma(pixels1, pixels2, pxPosStart, pxPosStart + chan, pxPosEnd) + check(arraysEq(pixels1, pixels2, pxPosStart, pxPosEnd)) + } + }.ensuring(_ => arraysEq(pixels1, pixels2, pxPosStart, pxPosEnd)) + + @ghost + @pure + @opaque + @inlineOnce + def samePixelsForallUnchangedLemma(pixels1: Array[Byte], pixels2: Array[Byte], px: Int, pxPosStart: Long, pxPosEnd: Long, chan: Long): Unit = { + decreases(pxPosEnd - pxPosStart) + require(3 <= chan && chan <= 4) + require(0 <= pxPosStart && pxPosStart < pxPosEnd && pxPosEnd <= pixels1.length) + require(pixels1.length % chan == 0) + require(pxPosStart % chan == 0) + require(pxPosEnd % chan == 0) + require(pixels1.length == pixels2.length) + require(samePixelsForall(pixels1, px, pxPosStart, pxPosEnd, chan)) + require(arraysEq(pixels1, pixels2, pxPosStart, pxPosEnd)) + + assert(samePixels(pixels2, px, pxPosStart, chan)) + if (pxPosStart + chan == pxPosEnd) { + check(samePixelsForall(pixels2, px, pxPosStart, pxPosEnd, chan)) + } else { + modSumLemma(pxPosStart, chan) + modLtLemma(pxPosStart, pxPosEnd, chan) + assert(samePixelsForall(pixels1, px, pxPosStart + chan, pxPosEnd, chan)) + arraysEqDropLeftLemma(pixels1, pixels2, pxPosStart, pxPosStart + chan, pxPosEnd) + assert(arraysEq(pixels1, pixels2, pxPosStart + chan, pxPosEnd)) + assert(pxPosStart + chan < pxPosEnd) + assert((pxPosStart + chan) % chan == 0) + samePixelsForallUnchangedLemma(pixels1, pixels2, px, pxPosStart + chan, pxPosEnd, chan) + check(samePixelsForall(pixels2, px, pxPosStart, pxPosEnd, chan)) + } + }.ensuring(_ => samePixelsForall(pixels2, px, pxPosStart, pxPosEnd, chan)) + + + @ghost + @pure + def arraysEq[T](arr1: Array[T], arr2: Array[T], from: Long, until: Long): Boolean = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from <= until && until <= arr1.length) + if (from == until) true + else arr1(from.toInt) == arr2(from.toInt) && arraysEq(arr1, arr2, from + 1, until) + } + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqCombinedLemma[T](arr1: Array[T], arr2: Array[T], from: Long, middle: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from < middle && middle < until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, middle)) + require(arraysEq(arr1, arr2, middle, until)) + + if (middle == from + 1) { + check(arraysEq(arr1, arr2, from, until)) + } else { + assert(arraysEq(arr1, arr2, from + 1, middle)) + arraysEqCombinedLemma(arr1, arr2, from + 1, middle, until) + assert(arraysEq(arr1, arr2, from + 1, until)) + check(arraysEq(arr1, arr2, from, until)) + } + }.ensuring(_ => arraysEq(arr1, arr2, from, until)) + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqCombinedLemma2[T](arr1: Array[T], arr2: Array[T], from: Long, middle: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from <= middle && middle <= until && until <= arr1.length) + require(from < until) + require(arraysEq(arr1, arr2, from, middle)) + require(arraysEq(arr1, arr2, middle, until)) + + if (from == middle || middle == until) { + check(arraysEq(arr1, arr2, from, until)) + } else { + arraysEqCombinedLemma(arr1, arr2, from, middle, until) + check(arraysEq(arr1, arr2, from, until)) + } + }.ensuring(_ => arraysEq(arr1, arr2, from, until)) + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqAtIndex[T](arr1: Array[T], arr2: Array[T], from: Long, until: Long, at: Long): Unit = { + decreases(until - at) + require(arr1.length == arr2.length) + require(0 <= from && from <= until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, until)) + require(from <= at && at < until) + if (at == from) { + check(arr1(at.toInt) == arr2(at.toInt)) + } else { + arraysEqAtIndex(arr1, arr2, from + 1, until, at) + check(arr1(at.toInt) == arr2(at.toInt)) + } + }.ensuring(_ => arr1(at.toInt) == arr2(at.toInt)) + + // This function reduces at compile-time to calls to arraysEqAtIndex, we ignore it as such. + @ignore + @ghost + inline def arraysEqAtIndices[T](arr1: Array[T], arr2: Array[T], from: Long, until: Long, inline fromIndice: Long, inline toIndice: Long): Unit = { + inline if (fromIndice >= toIndice) () + else { + arraysEqAtIndex(arr1, arr2, from, until, fromIndice) + arraysEqAtIndices(arr1, arr2, from, until, fromIndice + 1, toIndice) + } + } + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqSymLemma[T](arr1: Array[T], arr2: Array[T], from: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from <= until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, until)) + if (from == until) { + check(arraysEq(arr2, arr1, from, until)) + } else { + arraysEqSymLemma(arr1, arr2, from + 1, until) + check(arraysEq(arr2, arr1, from, until)) + } + }.ensuring(_ => arraysEq(arr2, arr1, from, until)) + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqDropRightLemma[T](arr1: Array[T], arr2: Array[T], from: Long, middle: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from <= middle && middle <= until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, until)) + + if (middle == from) { + check(arraysEq(arr1, arr2, from, until)) + } else { + arraysEqDropRightLemma(arr1, arr2, from + 1, middle, until) + check(arraysEq(arr1, arr2, from, middle)) + } + }.ensuring(_ => arraysEq(arr1, arr2, from, middle)) + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqDropLeftLemma[T](arr1: Array[T], arr2: Array[T], from: Long, middle: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length) + require(0 <= from && from <= middle && middle <= until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, until)) + + if (middle == from) { + check(arraysEq(arr1, arr2, middle, until)) + } else { + arraysEqDropLeftLemma(arr1, arr2, from + 1, middle, until) + check(arraysEq(arr1, arr2, middle, until)) + } + }.ensuring(_ => arraysEq(arr1, arr2, middle, until)) + + @ghost + @pure + @opaque + @inlineOnce + def arraysEqTransLemma[T](arr1: Array[T], arr2: Array[T], arr3: Array[T], from: Long, until: Long): Unit = { + decreases(until - from) + require(arr1.length == arr2.length && arr2.length == arr3.length) + require(0 <= from && from <= until && until <= arr1.length) + require(arraysEq(arr1, arr2, from, until)) + require(arraysEq(arr2, arr3, from, until)) + + if (from == until) { + check(arraysEq(arr1, arr3, from, until)) + } else { + arraysEqTransLemma(arr1, arr2, arr3, from + 1, until) + check(arraysEq(arr1, arr3, from, until)) + } + }.ensuring(_ => arraysEq(arr1, arr3, from, until)) + + @ghost + @pure + @opaque + @inlineOnce + def eqImpliesArraysEq[T](arr1: Array[T], arr2: Array[T], from: Long, until: Long): Unit = { + decreases(until - from) + require(arr1 == arr2) + require(0 <= from && from <= until && until <= arr1.length) + + if (from == until) { + check(arraysEq(arr1, arr2, from, until)) + } else { + eqImpliesArraysEq(arr1, arr2, from + 1, until) + check(arraysEq(arr1, arr2, from, until)) + } + }.ensuring(_ => arraysEq(arr1, arr2, from, until)) + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEq[T](arr: Array[T], i: Long, t: T): Unit = { + require(0 <= i && i < arr.length) + val upd = arr.updated(i.toInt, t) + + { + updatedAtArraysEqLeftProof(arr, i, t, 0) + updatedAtArraysEqRightProof(arr, i, t, arr.length) + }.ensuring { _ => + arraysEq(arr, upd, 0, i) &&& + arraysEq(arr, upd, i + 1, arr.length) + } + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEqLeftProof[T](arr: Array[T], i: Long, t: T, j: Long): Unit = { + decreases(i - j) + require(0 <= i && i < arr.length) + require(0 <= j && j <= i) + val upd = arr.updated(i.toInt, t) + + { + if (i == j) { + check(arraysEq(arr, upd, j, i)) + } else { + updatedAtArraysEqLeftProof(arr, i, t, j + 1) + check(arraysEq(arr, upd, j, i)) + } + }.ensuring(_ => arraysEq(arr, upd, j, i)) + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEqRightProof[T](arr: Array[T], i: Long, t: T, j: Long): Unit = { + decreases(j - i) + require(0 <= i && i < arr.length) + require(i < j && j <= arr.length) + val upd = arr.updated(i.toInt, t) + + { + if (i + 1 == j) { + check(arraysEq(arr, upd, i + 1, j)) + } else { + updatedAtArraysEqRightProof(arr, i, t, j - 1) + assert(arraysEq(arr, upd, i + 1, j - 1)) + assert(arraysEq(arr, upd, j - 1, j)) + arraysEqCombinedLemma2(arr, upd, i + 1, j - 1, j) + check(arraysEq(arr, upd, i + 1, j)) + } + }.ensuring(_ => arraysEq(arr, upd, i + 1, j)) + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEq[T](arr: Array[T], i: Long, t1: T, t2: T): Unit = { + require(0 <= i && i < arr.length) + require(i + 1 < arr.length) + val upd = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2) + + { + val updt1 = arr.updated(i.toInt, t1) + updatedAtArraysEq(arr, i, t1) + assert(arraysEq(arr, updt1, 0, i)) + + updatedAtArraysEq(updt1, i + 1, t2) + assert(arraysEq(updt1, upd, 0, i + 1)) + + arraysEqDropRightLemma(updt1, upd, 0, i, i + 1) + arraysEqTransLemma(arr, updt1, upd, 0, i) + check(arraysEq(arr, upd, 0, i)) + }.ensuring { _ => + arraysEq(arr, upd, 0, i) + } + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEq[T](arr: Array[T], i: Long, t1: T, t2: T, t3: T): Unit = { + require(0 <= i && i < arr.length) + require(i + 1 < arr.length) + require(i + 2 < arr.length) + val upd = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2).updated(i.toInt + 2, t3) + + { + val updt1t2 = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2) + updatedAtArraysEq(arr, i, t1, t2) + assert(arraysEq(arr, updt1t2, 0, i)) + + updatedAtArraysEq(updt1t2, i + 2, t3) + assert(arraysEq(updt1t2, upd, 0, i + 2)) + + arraysEqDropRightLemma(updt1t2, upd, 0, i, i + 2) + arraysEqTransLemma(arr, updt1t2, upd, 0, i) + check(arraysEq(arr, upd, 0, i)) + }.ensuring { _ => + arraysEq(arr, upd, 0, i) + } + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEq[T](arr: Array[T], i: Long, t1: T, t2: T, t3: T, t4: T): Unit = { + require(0 <= i && i < arr.length) + require(i + 1 < arr.length) + require(i + 2 < arr.length) + require(i + 3 < arr.length) + val upd = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2).updated(i.toInt + 2, t3).updated(i.toInt + 3, t4) + + { + val updt1t2t3 = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2).updated(i.toInt + 2, t3) + updatedAtArraysEq(arr, i, t1, t2, t3) + assert(arraysEq(arr, updt1t2t3, 0, i)) + + updatedAtArraysEq(updt1t2t3, i + 3, t4) + assert(arraysEq(updt1t2t3, upd, 0, i + 3)) + + arraysEqDropRightLemma(updt1t2t3, upd, 0, i, i + 3) + arraysEqTransLemma(arr, updt1t2t3, upd, 0, i) + check(arraysEq(arr, upd, 0, i)) + }.ensuring { _ => + arraysEq(arr, upd, 0, i) + } + } + + @ghost + @pure + @opaque + @inlineOnce + def updatedAtArraysEq[T](arr: Array[T], i: Long, t1: T, t2: T, t3: T, t4: T, t5: T): Unit = { + require(0 <= i && i < arr.length) + require(i + 1 < arr.length) + require(i + 2 < arr.length) + require(i + 3 < arr.length) + require(i + 4 < arr.length) + val upd = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2).updated(i.toInt + 2, t3).updated(i.toInt + 3, t4).updated(i.toInt + 4, t5) + + { + val updt1t2t3t4 = arr.updated(i.toInt, t1).updated(i.toInt + 1, t2).updated(i.toInt + 2, t3).updated(i.toInt + 3, t4) + updatedAtArraysEq(arr, i, t1, t2, t3, t4) + assert(arraysEq(arr, updt1t2t3t4, 0, i)) + + updatedAtArraysEq(updt1t2t3t4, i + 4, t5) + assert(arraysEq(updt1t2t3t4, upd, 0, i + 4)) + + arraysEqDropRightLemma(updt1t2t3t4, upd, 0, i, i + 4) + arraysEqTransLemma(arr, updt1t2t3t4, upd, 0, i) + check(arraysEq(arr, upd, 0, i)) + }.ensuring { _ => + arraysEq(arr, upd, 0, i) + } + } +} diff --git a/qoi/verified/decoder.scala b/qoi/verified/decoder.scala new file mode 100644 index 00000000..9443eff4 --- /dev/null +++ b/qoi/verified/decoder.scala @@ -0,0 +1,1012 @@ +import stainless.* +import stainless.lang.{inline as inlineCall, ghost as ghostExpr, *} +import stainless.collection.* +import stainless.annotation.{inlineOnce, ghost, mutable, opaque, pure, cCode} +import stainless.proof.* +import StaticChecks.* +import common.* + +object decoder { + + case class Ctx(bytes: Array[Byte], w: Long, h: Long, chan: Long) { + require(0 < w && w <= MaxWidth && + 0 < h && h <= MaxHeight && + 3 <= chan && chan <= 4 && + HeaderSize <= bytes.length) + + @inline + def pixelsLen: Long = { + w * h * chan + }.ensuring(_ % chan == 0) + + @inline + def chunksLen: Long = bytes.length - Padding + } + + @cCode.`export` + case class DecodedResult(pixels: Array[Byte], w: Long, h: Long, chan: Long) + + case class WriteRunPixelsResult(remainingRun: Long, pxPos: Long) + + sealed trait DecodedNext + case class Run(run: Long) extends DecodedNext + case class DiffOrIndexOrColor(px: Int) extends DecodedNext + + case class DecodingIteration(px: Int, inPos: Long, pxPos: Long, remainingRun: Long) + + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @pure + @cCode.`export` + def decode(bytes: Array[Byte], to: Long): OptionMut[DecodedResult] = { + if (!(bytes.length > HeaderSize + Padding && HeaderSize + Padding < to && to <= bytes.length)) + NoneMut() + else { + decodeHeader(bytes) match { + case Some((w, h, chan)) => + assert(0 < w && w <= MaxWidth) + assert(0 < h && h <= MaxHeight) + assert(3 <= chan && chan <= 4) + val index = Array.fill(64)(0) + val pixels = allocArray(w.toInt * h.toInt * chan.toInt) + val px = Pixel.fromRgba(0, 0, 0, 255.toByte) + val decIter = decodeLoop(index, pixels, px, HeaderSize, to - Padding, 0)(using Ctx(bytes, w, h, chan)) + @ghost val pixelsPreWrite = freshCopy(pixels) + ghostExpr { + check(0 <= decIter.pxPos && decIter.pxPos <= pixels.length) + check(decIter.pxPos % chan == 0) + } + if (decIter.pxPos != pixels.length) { + writeRemainingPixels(pixels, decIter.px, decIter.pxPos, decIter.pxPos)(using Ctx(bytes, w, h, chan)) + } + ghostExpr { + if (decIter.pxPos != pixels.length) { + check(arraysEq(pixelsPreWrite, pixels, 0, decIter.pxPos)) + } else { + eqImpliesArraysEq(pixelsPreWrite, pixels, 0, decIter.pxPos) + check(arraysEq(pixelsPreWrite, pixels, 0, decIter.pxPos)) + } + check((decIter.pxPos != pixels.length) ==> samePixelsForall(pixels, decIter.px, decIter.pxPos, pixels.length, chan)) + } + SomeMut(DecodedResult(pixels, w, h, chan)) + + case None() => NoneMut() + } + } + } + + @opaque + @inlineOnce + def writeRemainingPixels(pixels: Array[Byte], pxPrev: Int, pxPos: Long, @ghost pxPosOrig: Long)(using Ctx): Unit = { + decreases(pixels.length - pxPos) + require(pixels.length == pixelsLen) + require(0 <= pxPos && pxPos < pixels.length) + require(0 <= pxPosOrig && pxPosOrig <= pxPos) + require(pxPos % chan == 0) + require(pxPosOrig % chan == 0) + require((pxPosOrig != pxPos) ==> samePixelsForall(pixels, pxPrev, pxPosOrig, pxPos, chan)) + + @ghost val oldPixels = freshCopy(pixels) + ghostExpr { + modLeqLemma(pxPos, pixels.length, chan) + assert(pxPos + chan <= pixels.length) + } + val pxPosPlusChan = pxPos + chan + ghostExpr { + modSumLemma(pxPos, chan) + assert(pxPosPlusChan % chan == 0) + assert(pxPosPlusChan >= 0) + assert(pxPosPlusChan >= pxPosOrig) + } + writePixel(pixels, pxPrev, pxPos) + ghostExpr { + modMultLemma(w, h, chan) + assert(pixelsLen == w * h * chan) + check(pixelsLen % chan == 0) + check(pixels.length % chan == 0) + + assert(arraysEq(oldPixels, pixels, 0, pxPos)) + check(oldPixels.length == pixels.length) + arraysEqDropRightLemma(oldPixels, pixels, 0, pxPosOrig, pxPos) + check(arraysEq(oldPixels, pixels, 0, pxPosOrig)) + + assert(samePixels(pixels, pxPrev, pxPos, chan)) + samePixelsSingleElementRange(pixels, pxPrev, pxPos, chan) + assert(samePixelsForall(pixels, pxPrev, pxPos, pxPosPlusChan, chan)) + if (pxPosOrig != pxPos) { + arraysEqDropLeftLemma(oldPixels, pixels, 0, pxPosOrig, pxPos) + assert(samePixelsForall(oldPixels, pxPrev, pxPosOrig, pxPos, chan)) + samePixelsForallUnchangedLemma(oldPixels, pixels, pxPrev, pxPosOrig, pxPos, chan) + assert(samePixelsForall(pixels, pxPrev, pxPosOrig, pxPos, chan)) + + samePixelsForallCombinedLemma(pixels, pxPrev, pxPosOrig, pxPos, pxPosPlusChan, chan) + check(samePixelsForall(pixels, pxPrev, pxPosOrig, pxPosPlusChan, chan)) + } + } + assert((pxPosOrig != pxPos) ==> samePixelsForall(pixels, pxPrev, pxPosOrig, pxPosPlusChan, chan)) + + if (pxPosPlusChan < pixels.length) { + ghostExpr { + modLeqLemma(pxPosPlusChan, pixels.length, chan) + assert(pxPosPlusChan + chan <= pixels.length) + } + @ghost val pixelsPreRec = freshCopy(pixels) + writeRemainingPixels(pixels, pxPrev, pxPosPlusChan, pxPosOrig) + ghostExpr { + assert(pixelsPreRec.length == pixels.length) + check(pixelsPreRec.length == pixels.length) + assert(arraysEq(pixelsPreRec, pixels, 0, pxPosOrig)) + assert(arraysEq(oldPixels, pixelsPreRec, 0, pxPosOrig)) + arraysEqTransLemma(oldPixels, pixelsPreRec, pixels, 0, pxPosOrig) + check(arraysEq(oldPixels, pixels, 0, pxPosOrig)) + + check(samePixelsForall(pixels, pxPrev, pxPosOrig, pixels.length, chan)) + } + } else { + ghostExpr { + assert(pxPosPlusChan == pixels.length) + + if (pxPosOrig != pxPos) { + assert(samePixelsForall(pixels, pxPrev, pxPosOrig, pxPosPlusChan, chan)) + check(samePixelsForall(pixels, pxPrev, pxPosOrig, pixels.length, chan)) + } else { + check(samePixelsForall(pixels, pxPrev, pxPosOrig, pixels.length, chan)) + } + } + } + }.ensuring { _ => + old(pixels).length == pixels.length &&& + arraysEq(old(pixels), pixels, 0, pxPosOrig) &&& + samePixelsForall(pixels, pxPrev, pxPosOrig, pixels.length, chan) + } + + def decodeLoop(index: Array[Int], pixels: Array[Byte], pxPrev: Int, + inPos0: Long, untilInPos: Long, pxPos0: Long)(using Ctx): DecodingIteration = { + decreases(untilInPos - inPos0) + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < untilInPos && untilInPos <= chunksLen) + + @ghost val indexPre = freshCopy(index) + @ghost val pixelsPre = freshCopy(pixels) + val (res, decIter) = decodeNext(index, pixels, pxPrev, inPos0, pxPos0) + ghostExpr { check(decIter.pxPos % chan == 0) } + + if (decIter.inPos < untilInPos && decIter.pxPos + chan <= pixels.length) { + decodeLoop(index, pixels, decIter.px, decIter.inPos, untilInPos, decIter.pxPos) + } else { + decIter + } + }.ensuring { decIter => + index.length == 64 &&& + old(pixels).length == pixels.length &&& + pxPosInv(decIter.pxPos) &&& + HeaderSize <= decIter.inPos + } + + def decodeNext(index: Array[Int], pixels: Array[Byte], pxPrev: Int, inPos0: Long, pxPos0: Long)(using Ctx): (DecodedNext, DecodingIteration) = { + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < chunksLen) + + @ghost val pixelsPre = freshCopy(pixels) + @ghost val indexPre = freshCopy(index) + val (decRes, inPos) = doDecodeNext(index, pxPrev, inPos0) + + decRes match { + case Run(run) => + val WriteRunPixelsResult(resRun, resPxPos) = writeRunPixels(pixels, pxPrev, run, pxPos0) + ghostExpr { + check(pxPos0 < resPxPos && resPxPos <= pixels.length && resPxPos % chan == 0) + check(pixelsPre.length == pixels.length) + check(arraysEq(pixelsPre, pixels, 0, pxPos0)) + check(samePixelsForall(pixels, pxPrev, pxPos0, resPxPos, chan)) + check(writeRunPixelsInv(pxPos0, run, resPxPos, resRun)) + unfold(writeRunPixelsInv(pxPos0, run, resPxPos, resRun)) + check((pxPos0 + chan + chan * run <= pixelsLen) ==> (resRun == 0 && resPxPos == pxPos0 + chan * (run + 1))) // Slow (~60s) + } + (decRes, DecodingIteration(pxPrev, inPos, resPxPos, resRun)) + + case DiffOrIndexOrColor(px) => + writePixel(pixels, px, pxPos0) + index(colorPos(px)) = px + ghostExpr { + check(pixelsPre.length == pixels.length) + check(arraysEq(pixelsPre, pixels, 0, pxPos0)) + check(indexPre.updated(colorPos(px), px) == index) + check(samePixels(pixels, px, pxPos0, chan)) + } + (decRes, DecodingIteration(px, inPos, pxPos0 + chan, 0)) + } + } + + @pure + def doDecodeNext(index: Array[Int], pxPrev: Int, inPos0: Long)(using Ctx): (DecodedNext, Long) = { + require(index.length == 64) + require(HeaderSize <= inPos0 && inPos0 < chunksLen) + var px = pxPrev + var inPos = inPos0 + var run = 0L + + val b1 = bytes(inPos.toInt).toInt & 0xff + inPos += 1 + val res: DecodedNext = if (b1 == OpRgb) { + val px = Pixel.withRgba(pxPrev)(r = bytes(inPos.toInt), g = bytes(inPos.toInt + 1), b = bytes(inPos.toInt + 2)) + inPos += 3 + DiffOrIndexOrColor(px) + } else if (b1 == OpRgba) { + val px = Pixel.withRgba(pxPrev)(r = bytes(inPos.toInt), g = bytes(inPos.toInt + 1), b = bytes(inPos.toInt + 2), a = bytes(inPos.toInt + 3)) + inPos += 4 + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpIndex) { + val px = index(b1) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpDiff) { + val px = decodeDiff(pxPrev, b1) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpLuma) { + val b2 = bytes(inPos.toInt).toInt & 0xff + inPos += 1 + val px = decodeLuma(pxPrev, b1, b2) + DiffOrIndexOrColor(px) + } else if ((b1 & Mask2) == OpRun) { + val run = decodeRun(b1) + Run(run) + } else { + DiffOrIndexOrColor(pxPrev) + } + (res, inPos) + }.ensuring { case (_, inPos) => inPos <= inPos0 + 5 } + + @pure + def decodeHeader(bytes: Array[Byte]): Option[(Long, Long, Long)] = { + require(bytes.length > HeaderSize + Padding) + + val magic = read32(bytes, 0) + val w = read32(bytes, 4) + val h = read32(bytes, 8) + val chan = bytes(12) + + if (0 < w && w <= MaxWidth && 0 < h && h <= MaxHeight && magic == MagicNumber && 3 <= chan && chan <= 4) + Some((w.toLong, h.toLong, chan.toLong)) + else + None() + } + + // Note: not opaque, as writeRunPixelsPureBytesEqLemma needs to see its definition. + def writeRunPixels(pixels: Array[Byte], px: Int, run0: Long, pxPos0: Long)(using Ctx): WriteRunPixelsResult = { + decreases(pixelsLen - pxPos0) + require(pixels.length == pixelsLen) + require(runInv(run0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + + // Note: a run of 0 is possible: it means that we repeat the previous pixel once. + + val pxPos0PlusChan = pxPos0 + chan + ghostExpr { + modSumLemma(pxPos0, chan) + assert(pxPos0PlusChan % chan == 0) + } + @ghost val pixelsPre = freshCopy(pixels) + + writePixel(pixels, px, pxPos0) + + ghostExpr { + assert(pixels.length == pixelsPre.length) + assert(arraysEq(pixelsPre, pixels, 0, pxPos0)) + + assert(samePixels(pixels, px, pxPos0, chan)) + samePixelsSingleElementRange(pixels, px, pxPos0, chan) + check(samePixelsForall(pixels, px, pxPos0, pxPos0PlusChan, chan)) + } + + if (run0 > 0 && pxPos0 + chan < pixels.length) { + val run0Minus1 = run0 - 1 + @ghost val pixelsPreRec = freshCopy(pixels) + + ghostExpr { + modLeqLemma(pxPos0PlusChan, pixels.length, chan) + assert(pxPos0PlusChan + chan <= pixels.length) + eqImpliesArraysEq(pixelsPreRec, pixels, 0, pxPos0) + assert(arraysEq(pixelsPreRec, pixels, 0, pxPos0)) + assert(arraysEq(pixelsPre, pixelsPreRec, 0, pxPos0)) + assert(samePixelsForall(pixelsPreRec, px, pxPos0, pxPos0PlusChan, chan)) + } + + val WriteRunPixelsResult(run2, pxPos2) = writeRunPixels(pixels, px, run0Minus1, pxPos0PlusChan) + + ghostExpr { + assert(pixels.length == pixelsPre.length) + assert(pxPos0PlusChan < pxPos2) + assert(arraysEq(pixelsPreRec, pixels, 0, pxPos0PlusChan)) + assert(samePixelsForall(pixels, px, pxPos0PlusChan, pxPos2, chan)) + + // 2. + arraysEqDropRightLemma(pixelsPreRec, pixels, 0, pxPos0, pxPos0PlusChan) + assert(arraysEq(pixelsPreRec, pixels, 0, pxPos0)) + arraysEqTransLemma(pixelsPre, pixelsPreRec, pixels, 0, pxPos0) + check(arraysEq(pixelsPre, pixels, 0, pxPos0)) + + // 3. + arraysEqDropLeftLemma(pixelsPreRec, pixels, 0, pxPos0, pxPos0PlusChan) + assert(arraysEq(pixelsPreRec, pixels, pxPos0, pxPos0PlusChan)) + samePixelsForallUnchangedLemma(pixelsPreRec, pixels, px, pxPos0, pxPos0PlusChan, chan) + assert(samePixelsForall(pixels, px, pxPos0, pxPos0PlusChan, chan)) + samePixelsForallCombinedLemma(pixels, px, pxPos0, pxPos0PlusChan, pxPos2, chan) + check(samePixelsForall(pixels, px, pxPos0, pxPos2, chan)) + + // 4. + assert(pxPos0 + chan < pxPos2) + check(pxPos0 < pxPos2) + + // 7. + check(((pxPos0 + chan + run0 * chan <= pixels.length) ==> (run2 == 0 && pxPos2 == pxPos0 + chan * (run0 + 1))) because { + val lhs0 = pxPos0PlusChan + chan + run0Minus1 * chan + val rhs0 = pxPos0PlusChan + chan * (run0Minus1 + 1) + assert((lhs0 <= pixels.length) ==> (run2 == 0 && pxPos2 == rhs0)) + + val lhs1 = pxPos0 + chan + chan + (run0 - 1) * chan + val rhs1 = pxPos0 + chan + chan * run0 + assert(lhs1 == lhs0) + assert(rhs1 == rhs0) + + val lhs2 = pxPos0 + chan + chan + run0 * chan - chan + val rhs2 = pxPos0 + chan * (run0 + 1) + assert(lhs2 == lhs1) + assert(rhs2 == rhs1) + + val lhs3 = pxPos0 + chan + run0 * chan + assert(lhs3 == lhs2) + + assert(lhs3 == lhs0) + assert(rhs2 == rhs0) + assert((lhs3 <= pixels.length) ==> (run2 == 0 && pxPos2 == rhs2)) + + check((pxPos0 + chan + run0 * chan <= pixels.length) ==> (run2 == 0 && pxPos2 == pxPos0 + chan * (run0 + 1))) + trivial + }) + } + WriteRunPixelsResult(run2, pxPos2) + } else { + WriteRunPixelsResult(run0, pxPos0 + chan) + } + }.ensuring { case WriteRunPixelsResult(run2, pxPos2) => + old(pixels).length == pixels.length &&& + arraysEq(old(pixels), pixels, 0, pxPos0) &&& + samePixelsForall(pixels, px, pxPos0, pxPos2, chan) &&& + pxPos0 < pxPos2 &&& + pxPos2 <= pixels.length &&& + pxPos2 % chan == 0 &&& + writeRunPixelsInv(pxPos0, run0, pxPos2, run2) + } + + // Note: not opaque, as writePixelPureBytesEqLemma needs to see its definition. + def writePixel(pixels: Array[Byte], px: Int, pxPos: Long)(using Ctx): Unit = { + require(pixels.length == pixelsLen) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + + @ghost val pixelsPre = freshCopy(pixels) + ghostExpr { + eqImpliesArraysEq(pixelsPre, pixels, 0, pxPos) + assert(arraysEq(pixelsPre, pixels, 0, pxPos)) + } + + pixels(pxPos.toInt) = Pixel.r(px) + pixels(pxPos.toInt + 1) = Pixel.g(px) + pixels(pxPos.toInt + 2) = Pixel.b(px) + + if (chan == 4) { + pixels(pxPos.toInt + 3) = Pixel.a(px) + ghostExpr { + updatedAtArraysEq(pixelsPre, pxPos, Pixel.r(px), Pixel.g(px), Pixel.b(px), Pixel.a(px)) + check(arraysEq(pixelsPre, pixels, 0, pxPos)) + } + } else { + ghostExpr { + updatedAtArraysEq(pixelsPre, pxPos, Pixel.r(px), Pixel.g(px), Pixel.b(px)) + check(arraysEq(pixelsPre, pixels, 0, pxPos)) + } + } + }.ensuring { _ => + old(pixels).length == pixels.length &&& + arraysEq(old(pixels), pixels, 0, pxPos) &&& + samePixels(pixels, px, pxPos, chan) + } + + def decodeDiff(pxPrev: Int, b1: Int): Int = { + require((b1 & Mask2) == OpDiff) + Pixel.incremented(pxPrev)( + (((b1 >>> 4) & 0x03) - 2).toByte, + (((b1 >>> 2) & 0x03) - 2).toByte, + ((b1 & 0x03) - 2).toByte) + } + + def decodeLuma(pxPrev: Int, b1: Int, b2: Int): Int = { + require((b1 & Mask2) == OpLuma) + val vg = (b1 & 0x3f) - 32 + Pixel.incremented(pxPrev)( + (vg - 8 + ((b2 >>> 4) & 0x0f)).toByte, + vg.toByte, + (vg - 8 + (b2 & 0x0f)).toByte) + } + + def decodeRun(b1: Int): Int = { + require((b1 & Mask2) == OpRun) + b1 & 0x3f + } + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + def runInv(run: Long): Boolean = + 0 <= run && run <= 62 + + @ghost + def pxPosInv(pxPos: Long)(using Ctx): Boolean = + 0 <= pxPos && pxPos <= pixelsLen && pxPos % chan == 0 + + @ghost + def inPosInv(inPos: Long)(using Ctx): Boolean = + HeaderSize <= inPos && inPos <= bytes.length - Padding + + @ghost + def rangesInv(run: Long, inPos: Long, pxPos: Long)(using Ctx): Boolean = + pxPosInv(pxPos) && runInv(run) && inPosInv(inPos) + + @ghost + def rangesInv(indexLen: Long, run: Long, inPos: Long, pxPos: Long)(using Ctx): Boolean = + pxPosInv(pxPos) && runInv(run) && inPosInv(inPos) && indexLen == 64 + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + @pure + def writePixelPure(pixels: Array[Byte], px: Int, pxPos: Long)(using Ctx): Array[Byte] = { + require(pixels.length == pixelsLen) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + val pixelsCpy = freshCopy(pixels) + writePixel(pixelsCpy, px, pxPos) + pixelsCpy + }.ensuring { newPixels => + pixels.length == newPixels.length &&& + arraysEq(pixels, newPixels, 0, pxPos) &&& + samePixels(newPixels, px, pxPos, chan) + } + + @ghost + @pure + @opaque + @inlineOnce + def writePixelPureBytesEqLemma(pixels: Array[Byte], px: Int, pxPos: Long, from: Long, until: Long, bytes2: Array[Byte])(using ctx1: Ctx): Unit = { + require(pixels.length == pixelsLen) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(bytes.length == bytes2.length) + require(0 <= from && from <= until && until <= bytes.length) + require(arraysEq(bytes, bytes2, from, until)) + + val ctx2 = Ctx(freshCopy(bytes2), w, h, chan) + val pix1 = writePixelPure(pixels, px, pxPos)(using ctx1) + val pix2 = writePixelPure(pixels, px, pxPos)(using ctx2) + + { + () + }.ensuring(_ => pix1 == pix2) + } + + @ghost + def writeRunPixelsInv(pxPos0: Long, run0: Long, pxPos2: Long, run2: Long)(using Ctx): Boolean = { + require(runInv(run0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixelsLen) + (pxPos0 + chan + chan * run0 <= pixelsLen) ==> (run2 == 0 && pxPos2 == pxPos0 + chan * (run0 + 1)) + } + + @ghost + @pure + def writeRunPixelsPure(pixels: Array[Byte], px: Int, run0: Long, pxPos0: Long)(using Ctx): (Array[Byte], Long, Long) = { + require(pixels.length == pixelsLen) + require(runInv(run0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + val pixelsCpy = freshCopy(pixels) + val WriteRunPixelsResult(run2, pxPos2) = writeRunPixels(pixelsCpy, px, run0, pxPos0) + (pixelsCpy, run2, pxPos2) + }.ensuring { case (newPixels, run2, pxPos2) => + pixels.length == newPixels.length &&& + arraysEq(pixels, newPixels, 0, pxPos0) &&& + samePixelsForall(newPixels, px, pxPos0, pxPos2, chan) &&& + pxPos0 < pxPos2 &&& + pxPos2 <= newPixels.length &&& + pxPos2 % chan == 0 &&& + writeRunPixelsInv(pxPos0, run0, pxPos2, run2) + } + + @ghost + @pure + @opaque + @inlineOnce + def writeRunPixelPureBytesEqLemma(pixels: Array[Byte], px: Int, run0: Long, pxPos0: Long, from: Long, until: Long, bytes2: Array[Byte])(using ctx1: Ctx): Unit = { + decreases(pixelsLen - pxPos0) + require(pixels.length == pixelsLen) + require(runInv(run0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(bytes.length == bytes2.length) + require(0 <= from && from <= until && until <= bytes.length) + require(arraysEq(bytes, bytes2, from, until)) + + val ctx2 = Ctx(freshCopy(bytes2), w, h, chan) + val (pix1, run1, pxPos1) = writeRunPixelsPure(pixels, px, run0, pxPos0)(using ctx1) + val (pix2, run2, pxPos2) = writeRunPixelsPure(pixels, px, run0, pxPos0)(using ctx2) + + { + if (run0 == 0 || pxPos0 + chan >= pixels.length) { + check(pix1 == pix2) + check(run1 == run2) + check(pxPos1 == pxPos2) + } else { + val pxPos0PlusChan = pxPos0 + chan + modSumLemma(pxPos0, chan) + modLeqLemma(pxPos0, pixels.length, chan) + assert(pxPos0PlusChan % chan == 0) + assert(pxPos0PlusChan + chan <= pixels.length) + val pixNext1 = writePixelPure(pixels, px, pxPos0)(using ctx1) + val (pixRec1, runRec1, pxPosRec1) = writeRunPixelsPure(pixNext1, px, run0 - 1, pxPos0PlusChan)(using ctx1) + assert(pixRec1 == pix1) + assert(runRec1 == run1) + assert(pxPosRec1 == pxPos1) + val pixNext2 = writePixelPure(pixels, px, pxPos0)(using ctx2) + val (pixRec2, runRec2, pxPosRec2) = writeRunPixelsPure(pixNext2, px, run0 - 1, pxPos0PlusChan)(using ctx2) + assert(pixRec2 == pix2) + assert(runRec2 == run2) + assert(pxPosRec2 == pxPos2) + + writePixelPureBytesEqLemma(pixels, px, pxPos0, from, until, bytes2) + assert(pixNext1 == pixNext2) + writeRunPixelPureBytesEqLemma(pixNext1, px, run0 - 1, pxPos0PlusChan, from, until, bytes2) + assert(pixRec1 == pixRec2) + assert(runRec1 == runRec2) + assert(pxPosRec1 == pxPosRec2) + + check(pix1 == pix2) + check(run1 == run2) + check(pxPos1 == pxPos2) + } + }.ensuring { _ => + pix1 == pix2 &&& + run1 == run2 &&& + pxPos1 == pxPos2 + } + } + + @ghost + @pure + def decodeNextPure(index: Array[Int], pixels: Array[Byte], pxPrev: Int, inPos0: Long, pxPos0: Long)(using Ctx): (Array[Int], Array[Byte], DecodedNext, DecodingIteration) = { + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < chunksLen) + + val indexPre = freshCopy(index) + val pixelsPre = freshCopy(pixels) + val (decRes, decIter) = decodeNext(indexPre, pixelsPre, pxPrev, inPos0, pxPos0) + (indexPre, pixelsPre, decRes, decIter) + }.ensuring { + case (indexPre, pixelsPre, _, decIter) => + indexPre.length == 64 &&& + pixelsPre.length == pixels.length &&& + 0 <= decIter.inPos &&& + inPos0 < decIter.inPos &&& + decIter.pxPos % chan == 0 + } + + @ghost + @pure + @opaque + @inlineOnce + def decodeHeaderBytesEqLemma(bytes: Array[Byte], bytes2: Array[Byte]): Unit = { + require(bytes.length > HeaderSize + Padding) + require(bytes.length == bytes2.length) + require(arraysEq(bytes, bytes2, 0, HeaderSize)) + + val res1 = decodeHeader(bytes) + val res2 = decodeHeader(bytes2) + + { + arraysEqAtIndices(bytes, bytes2, 0, HeaderSize, 0, 14) + check(res1 == res2) + }.ensuring(_ => res1 == res2) + } + + @ghost + @pure + @opaque + @inlineOnce + def decodeLemma(until: Long)(using Ctx): Unit = { + require(HeaderSize + Padding < until && until <= bytes.length) + val readMagic = read32(bytes, 0) + val readW = read32(bytes, 4) + val readH = read32(bytes, 8) + val readChan = bytes(12) + require(readMagic == MagicNumber) + require(readW == w.toInt) + require(readH == h.toInt) + require(readChan.toLong == chan) + + val SomeMut(DecodedResult(decodedPixels, ww, hh, cchan)) = decode(bytes, until) + val initIndex = Array.fill(64)(0) + val initPixels = Array.fill(w.toInt * h.toInt * chan.toInt)(0: Byte) + val initPx = Pixel.fromRgba(0, 0, 0, 255.toByte) + val (_, decodedPixLoop, decIter) = decodeLoopPure(initIndex, initPixels, initPx, HeaderSize, until - Padding, 0) + + { + check(initPixels.length == decodedPixLoop.length) + check(decodedPixels.length == initPixels.length) + modMultLemma(w, h, chan) + check(decodedPixels.length % chan == 0) + check(0 <= decIter.pxPos && decIter.pxPos <= decodedPixels.length) + check(decIter.pxPos % chan == 0) + }.ensuring { _ => + arraysEq(decodedPixLoop, decodedPixels, 0, decIter.pxPos) &&& + ((decIter.pxPos != decodedPixels.length) ==> samePixelsForall(decodedPixels, decIter.px, decIter.pxPos, decodedPixels.length, chan)) + } + } + + @ghost + @pure + def decodeLoopPure(index: Array[Int], pixels: Array[Byte], pxPrev: Int, + inPos0: Long, untilInPos: Long, pxPos0: Long)(using Ctx): (Array[Int], Array[Byte], DecodingIteration) = { + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < untilInPos && untilInPos <= chunksLen) + + val indexCpy = freshCopy(index) + val pixelsCpy = freshCopy(pixels) + val res = decodeLoop(indexCpy, pixelsCpy, pxPrev, inPos0, untilInPos, pxPos0) + (indexCpy, pixelsCpy, res) + }.ensuring { case (indexCpy, pixelsCpy, res) => + 0 <= res.pxPos &&& res.pxPos <= pixels.length &&& + res.pxPos % chan == 0 + indexCpy.length == 64 &&& + pixelsCpy.length == pixels.length + } + + @ghost + @opaque + @inlineOnce + @pure + def decodeLoopPureNextLemma(index: Array[Int], + pixels: Array[Byte], + pxPrev: Int, + inPos0: Long, + inPos1: Long, + pxPos0: Long)(using Ctx): Unit = { + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < inPos1 && inPos1 <= chunksLen) + + val (index1, pixels1, decIter1) = decodeLoopPure(index, pixels, pxPrev, inPos0, inPos1, pxPos0) + val (indexNext, pixelNext, _, decIterNext) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0) + require(decIterNext.inPos < inPos1 && decIterNext.pxPos < pixels.length && decIterNext.pxPos + chan <= pixels.length) + val (index2, pixels2, decIter2) = decodeLoopPure(indexNext, pixelNext, decIterNext.px, decIterNext.inPos, inPos1, decIterNext.pxPos) + + { + () + }.ensuring { _ => + decIter1 == decIter2 &&& + index1 == index2 &&& + pixels1 == pixels2 + } + } + + @ghost + @opaque + @inlineOnce + @pure + def decodeLoopPureBytesEqLemma(index: Array[Int], + pixels: Array[Byte], + pxPrev: Int, + inPos0: Long, + untilInPos: Long, + pxPos0: Long, + bytes2: Array[Byte])(using ctx1: Ctx): Unit = { + decreases(untilInPos - inPos0) + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < untilInPos && untilInPos <= chunksLen) + require(bytes.length == bytes2.length) + require(arraysEq(bytes, bytes2, inPos0, untilInPos)) + + val ctx2 = Ctx(freshCopy(bytes2), w, h, chan) + val (ix1, pix1, decIter1) = decodeLoopPure(index, pixels, pxPrev, inPos0, untilInPos, pxPos0)(using ctx1) + require(decIter1.pxPos <= pixels.length) + require(decIter1.inPos == untilInPos) + val (ix2, pix2, decIter2) = decodeLoopPure(index, pixels, pxPrev, inPos0, untilInPos, pxPos0)(using ctx2) + + { + assert(ctx2.bytes == bytes2) + val (ix1Next, pix1Next, _, decIter1Next) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0)(using ctx1) + val (ix2Next, pix2Next, _, decIter2Next) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0)(using ctx2) + + assert(inPos0 < decIter1Next.inPos) + assert(decIter1Next.inPos <= decIter1.inPos) + arraysEqDropRightLemma(bytes, bytes2, inPos0, decIter1Next.inPos, untilInPos) + assert(arraysEq(bytes, bytes2, inPos0, decIter1Next.inPos)) + decodeNextPureBytesEqLemma(index, pixels, pxPrev, inPos0, pxPos0, bytes2) + assert(decIter1Next == decIter2Next) + assert(ix1Next == ix2Next) + assert(pix1Next == pix2Next) + + if (decIter1Next.inPos == decIter1.inPos) { + assert(decIter1Next.inPos == untilInPos) + assert(ix1Next == ix1) + assert(pix1Next == pix1) + assert(decIter1Next == decIter1) + check(ix1 == ix2) + check(pix1 == pix2) + check(decIter1 == decIter2) + } else { + assert(0 <= decIter1Next.inPos && decIter1Next.inPos < decIter1.inPos) + assert(ix1Next.length == 64 && pixels.length == pix1Next.length) + assert(decIter1Next.pxPos < pixels.length) + assert(decIter1Next.pxPos + chan <= pixels.length) + + arraysEqDropLeftLemma(bytes, bytes2, inPos0, decIter1Next.inPos, untilInPos) + assert(arraysEq(bytes, bytes2, decIter1Next.inPos, untilInPos)) + decodeLoopPureBytesEqLemma(ix1Next, pix1Next, decIter1Next.px, decIter1Next.inPos, untilInPos, decIter1Next.pxPos, bytes2) + val (ix1Rec, pix1Rec, decIter1Rec) = decodeLoopPure(ix1Next, pix1Next, decIter1Next.px, decIter1Next.inPos, untilInPos, decIter1Next.pxPos)(using ctx1) + val (ix2Rec, pix2Rec, decIter2Rec) = decodeLoopPure(ix1Next, pix1Next, decIter1Next.px, decIter1Next.inPos, untilInPos, decIter1Next.pxPos)(using ctx2) + assert(ix1Rec == ix2Rec) + assert(pix1Rec == pix2Rec) + assert(decIter1Rec == decIter2Rec) + + decodeLoopPureNextLemma(index, pixels, pxPrev, inPos0, untilInPos, pxPos0) + assert(ix1 == ix1Rec) + assert(pix1 == pix1Rec) + assert(decIter1 == decIter1Rec) + + check(ix1 == ix2) + check(pix1 == pix2) + check(decIter1 == decIter2) + } + }.ensuring { _ => + ix1 == ix2 &&& + pix1 == pix2 &&& + decIter1 == decIter2 + } + } + + @ghost + @opaque + @inlineOnce + @pure + def decodeLoopPureMergedLemma(index: Array[Int], + pixels: Array[Byte], + pxPrev: Int, + inPos0: Long, + inPos1: Long, + inPos2: Long, + pxPos0: Long)(using Ctx): Unit = { + decreases(inPos2 - inPos0) + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < inPos1 && inPos1 < inPos2 && inPos2 <= chunksLen) + val (index1, pixels1, decIter1) = decodeLoopPure(index, pixels, pxPrev, inPos0, inPos1, pxPos0) + require(decIter1.pxPos < pixels.length && decIter1.pxPos + chan <= pixels.length) + require(decIter1.inPos == inPos1) + val (index2, pixels2, decIter2) = decodeLoopPure(index1, pixels1, decIter1.px, inPos1, inPos2, decIter1.pxPos) + val (index3, pixels3, decIter3) = decodeLoopPure(index, pixels, pxPrev, inPos0, inPos2, pxPos0) + + { + val (indexNext, pixelNext, _, decIterNext) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0) + if (decIterNext.inPos == decIter1.inPos) { + check(decIter2.inPos == decIter3.inPos) + check(decIter2.pxPos == decIter3.pxPos) + } else { + assert(0 <= decIterNext.inPos && decIterNext.inPos < decIter1.inPos) + assert(indexNext.length == 64 && pixels.length == pixelNext.length) + assert(decIterNext.pxPos < pixels.length) + assert(decIterNext.pxPos + chan <= pixels.length) + + val (index1Rec, pixels1Rec, decIter1Rec) = decodeLoopPure(indexNext, pixelNext, decIterNext.px, decIterNext.inPos, inPos1, decIterNext.pxPos) + decodeLoopPureNextLemma(index, pixels, pxPrev, inPos0, inPos1, pxPos0) + assert(decIter1Rec == decIter1) + assert(index1Rec == index1) + assert(pixels1Rec == pixels1) + + val (index2Rec, pixels2Rec, decIter2Rec) = decodeLoopPure(index1Rec, pixels1Rec, decIter1Rec.px, inPos1, inPos2, decIter1Rec.pxPos) + + val (index3Rec, pixels3Rec, decIter3Rec) = decodeLoopPure(indexNext, pixelNext, decIterNext.px, decIterNext.inPos, inPos2, decIterNext.pxPos) + + decodeLoopPureMergedLemma(indexNext, pixelNext, decIterNext.px, decIterNext.inPos, inPos1, inPos2, decIterNext.pxPos) + assert(decIter2Rec == decIter3Rec) + + decodeLoopPureNextLemma(index, pixels, pxPrev, inPos0, inPos2, pxPos0) + assert(decIter3Rec == decIter3) + assert(index3Rec == index3) + assert(pixels3Rec == pixels3) + + assert(decodeLoopPure(index1Rec, pixels1Rec, decIter1Rec.px, inPos1, inPos2, decIter1Rec.pxPos) == + decodeLoopPure(index1, pixels1, decIter1.px, inPos1, inPos2, decIter1.pxPos)) + assert(decIter2Rec == decIter2) + assert(index2Rec == index2) + assert(pixels2Rec == pixels2) + + check(decIter2 == decIter3) + assert(index2 == index3) + assert(pixels2 == pixels3) + } + }.ensuring { _ => + decIter2 == decIter3 &&& + index2 == index3 &&& + pixels2 == pixels3 + } + } + + @ghost + @opaque + @inlineOnce + @pure + def decodeNextPureBytesEqLemma(index: Array[Int], pixels: Array[Byte], pxPrev: Int, inPos0: Long, pxPos0: Long, bytes2: Array[Byte])(using ctx1: Ctx): Unit = { + require(index.length == 64) + require(pixels.length == pixelsLen) + require(inPosInv(inPos0)) + require(pxPosInv(pxPos0)) + require(pxPos0 + chan <= pixels.length) + require(inPos0 < chunksLen) + require(bytes.length == bytes2.length) + val ctx2 = Ctx(freshCopy(bytes2), w, h, chan) + val (ix1, pix1, res1, decIter1) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0)(using ctx1) + require(arraysEq(bytes, bytes2, inPos0, decIter1.inPos)) + val (ix2, pix2, res2, decIter2) = decodeNextPure(index, pixels, pxPrev, inPos0, pxPos0)(using ctx2) + + { + doDecodeNextBytesEqLemma(index, pxPrev, inPos0, bytes2) + check(res1 == res2) + check(ix1 == ix2) + assert(bytes(inPos0.toInt) == bytes2(inPos0.toInt)) + assert(decIter1.inPos <= inPos0 + 5) + assert((inPos0 + 1 < decIter1.inPos) ==> { + arraysEqAtIndex(bytes, bytes2, inPos0, decIter1.inPos, inPos0 + 1) + bytes((inPos0 + 1).toInt) == bytes2((inPos0 + 1).toInt) + }) + assert((inPos0 + 2 < decIter1.inPos) ==> { + arraysEqAtIndex(bytes, bytes2, inPos0, decIter1.inPos, inPos0 + 2) + bytes((inPos0 + 2).toInt) == bytes2((inPos0 + 2).toInt) + }) + assert((inPos0 + 3 < decIter1.inPos) ==> { + arraysEqAtIndex(bytes, bytes2, inPos0, decIter1.inPos, inPos0 + 3) + bytes((inPos0 + 3).toInt) == bytes2((inPos0 + 3).toInt) + }) + assert((inPos0 + 4 < decIter1.inPos) ==> { + arraysEqAtIndex(bytes, bytes2, inPos0, decIter1.inPos, inPos0 + 4) + bytes((inPos0 + 4).toInt) == bytes2((inPos0 + 4).toInt) + }) + val b1 = bytes(inPos0.toInt).toInt & 0xff + if (b1 == OpRgb) { + assert(decIter1.inPos == inPos0 + 4) + assert(bytes((inPos0 + 1).toInt) == bytes2((inPos0 + 1).toInt)) + assert(bytes((inPos0 + 2).toInt) == bytes2((inPos0 + 2).toInt)) + assert(bytes((inPos0 + 3).toInt) == bytes2((inPos0 + 3).toInt)) + check(decIter1 == decIter2) + check(pix1 == pix2) + } else if (b1 == OpRgba) { + assert(decIter1.inPos == inPos0 + 5) + assert(bytes((inPos0 + 1).toInt) == bytes2((inPos0 + 1).toInt)) + assert(bytes((inPos0 + 2).toInt) == bytes2((inPos0 + 2).toInt)) + assert(bytes((inPos0 + 3).toInt) == bytes2((inPos0 + 3).toInt)) + assert(bytes((inPos0 + 4).toInt) == bytes2((inPos0 + 4).toInt)) + check(decIter1 == decIter2) + check(pix1 == pix2) + } else if ((b1 & Mask2) == OpLuma) { + assert(decIter1.inPos == inPos0 + 2) + assert(bytes((inPos0 + 1).toInt) == bytes2((inPos0 + 1).toInt)) + check(decIter1 == decIter2) + check(pix1 == pix2) + } else if ((b1 & Mask2) == OpIndex || (b1 & Mask2) == OpDiff) { + assert(decIter1.inPos == inPos0 + 1) + check(decIter1 == decIter2) + check(pix1 == pix2) + } else if ((b1 & Mask2) == OpRun) { + assert(decIter1.inPos == inPos0 + 1) + assert(res1 == Run(decodeRun(b1))) + assert(res2 == Run(decodeRun(b1))) + writeRunPixelPureBytesEqLemma(pixels, pxPrev, decodeRun(b1), pxPos0, inPos0, decIter1.inPos, bytes2) + check(decIter1 == decIter2) + check(pix1 == pix2) + } else { + assert(res1 == DiffOrIndexOrColor(pxPrev)) + assert(res2 == DiffOrIndexOrColor(pxPrev)) + assert(decIter1.inPos == inPos0 + 1) + check(decIter1 == decIter2) + check(pix1 == pix2) + } + check(decIter1 == decIter2) + check(pix1 == pix2) + }.ensuring { _ => + res1 == res2 &&& + ix1 == ix2 &&& + pix1 == pix2 &&& + decIter1 == decIter2 + } + } + + @ghost + @opaque + @inlineOnce + @pure + def doDecodeNextBytesEqLemma(index: Array[Int], pxPrev: Int, inPos0: Long, bytes2: Array[Byte])(using ctx1: Ctx): Unit = { + require(index.length == 64) + require(inPosInv(inPos0)) + require(inPos0 < chunksLen) + require(bytes.length == bytes2.length) + + val ctx2 = Ctx(freshCopy(bytes2), w, h, chan) + val (res1, inPos1) = doDecodeNext(index, pxPrev, inPos0)(using ctx1) + require(arraysEq(bytes, bytes2, inPos0, inPos1)) + val (res2, inPos2) = doDecodeNext(index, pxPrev, inPos0)(using ctx2) + + { + assert(bytes(inPos0.toInt) == bytes2(inPos0.toInt)) + check(inPos1 == inPos2) + check(res1 == res2) + () + }.ensuring { _ => + res1 == res2 &&& + inPos1 == inPos2 + } + } + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @inline + @cCode.inline + def w(using ctx: Ctx): Long = ctx.w + + @inline + @cCode.inline + def h(using ctx: Ctx): Long = ctx.h + + @inline + @cCode.inline + def chan(using ctx: Ctx): Long = ctx.chan + + @inline + @cCode.inline + def pixelsLen(using ctx: Ctx): Long = ctx.pixelsLen + + @inline + @cCode.inline + def chunksLen(using ctx: Ctx): Long = ctx.chunksLen + + @inline + @cCode.inline + def bytes(using ctx: Ctx): Array[Byte] = ctx.bytes +} diff --git a/qoi/verified/encoder.scala b/qoi/verified/encoder.scala new file mode 100644 index 00000000..9418f002 --- /dev/null +++ b/qoi/verified/encoder.scala @@ -0,0 +1,1501 @@ +import stainless.* +import stainless.lang.{ghost as ghostExpr, inline as inlineCall, *} +import stainless.collection.* +import stainless.annotation.{cCode, inlineOnce, mutable, opaque, pure, ghost} +import stainless.proof.* +import StaticChecks.* +import common.* + +object encoder { + + case class Ctx(pixels: Array[Byte], w: Long, h: Long, chan: Long) { + require(0 < w && w <= MaxWidth && + 0 < h && h <= MaxHeight && + 3 <= chan && chan <= 4 && + w * h * chan == pixels.length) + + @inline + def maxSize: Long = w * h * (chan + 1) + HeaderSize + Padding + + @inline + def pxEnd: Long = pixels.length - chan + } + + @cCode.`export` + case class EncodedResult(encoded: Array[Byte], length: Long) + + case class EncodeSingleStepResult(px: Int, outPos: Long, run: Long) + + case class EncodingIteration(px: Int, outPos: Long) + + case class RunUpdate(reset: Boolean, run: Long, outPos: Long) + + case class LoopIter(px: Int, pxPrev: Int, pxPos: Long) + + case class GhostDecoded(var index: Array[Int], + var pixels: Array[Byte], + var inPos: Long, + var pxPos: Long) + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + @pure + def decodeEncodeIsIdentityThm(using Ctx): Boolean = { + val SomeMut(EncodedResult(bytes, outPos)) = doEncode + + decoder.decode(bytes, outPos) match { + case SomeMut(decoder.DecodedResult(decodedPixels, ww, hh, cchan)) => + ww == w && + hh == h && + cchan == chan && + arraysEq(pixels, decodedPixels, 0, pixels.length) + case NoneMut() => false + } + }.holds + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @pure + @cCode.`export` + def encode(pixels: Array[Byte], w: Long, h: Long, chan: Long): OptionMut[EncodedResult] = { + if (!(0 < w && w <= MaxWidth && + 0 < h && h <= MaxHeight && + 3 <= chan && chan <= 4 && + w * h * chan == pixels.length)) + NoneMut() + else doEncode(using Ctx(pixels, w, h, chan)) + } + + @pure + def doEncode(using Ctx): OptionMut[EncodedResult] = { + val bytes = allocArray(maxSize.toInt) + writeHeader(bytes) + val index = Array.fill(64)(0) + val pxPrev = Pixel.fromRgba(0, 0, 0, 255.toByte) + @ghost val decoded = GhostDecoded(freshCopy(index), Array.fill(pixels.length)(0: Byte), HeaderSize, 0) + @ghost val initDecoded = freshCopy(decoded) + @ghost val oldBytes = freshCopy(bytes) + val EncodingIteration(pxRes, outPos) = encodeLoop(index, bytes, pxPrev, 0, HeaderSize, 0, decoded) + + ghostExpr { + assert(decodeLoopEncodeProp(bytes, pxPrev, HeaderSize, outPos, initDecoded, pxRes, decoded)) + assert(decoded.pxPos == pxEnd + chan) + assert(pxEnd + chan == pixels.length) + assert(decoded.pxPos == pixels.length) + assert(arraysEq(oldBytes, bytes, 0, HeaderSize)) + assert(HeaderSize <= outPos && outPos <= maxSize - Padding) + + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + + assert(pixels.length == w * h * chan) + val (decIndex, decPixels, decIter) = decoder.decodeLoopPure(initDecoded.index, initDecoded.pixels, pxPrev, HeaderSize, outPos, 0) + assert(decIter.pxPos % chan == 0) + assert(decIter.inPos == outPos) + assert(decIndex == decoded.index) + assert(decPixels == decoded.pixels) + assert(decIter.remainingRun == 0) + assert(decIter.pxPos == decoded.pxPos) + assert(decIter.pxPos == pixels.length) + assert(decIter.inPos == decoded.inPos) + assert(0 < w && w <= MaxWidth) + assert(0 < h && h <= MaxHeight) + assert(3 <= chan && chan <= 4) + assert(pixels.length > 0) + assert(arraysEq(pixels, decPixels, 0, pixels.length)) + + { + val readMagic = read32(oldBytes, 0) + val readW = read32(oldBytes, 4) + val readH = read32(oldBytes, 8) + val readChan = oldBytes(12) + assert(readMagic == MagicNumber) + assert(readW == w.toInt) + assert(readH == h.toInt) + assert(readChan.toLong == chan) + assert(0 < readW && readW <= MaxWidth) + assert(0 < readH && readH <= MaxHeight) + assert(readMagic == MagicNumber) + assert(3 <= readChan && readChan <= 4) + assert(readW.toLong == w) + assert(readH.toLong == h) + assert(readChan.toLong == chan) + unfold(decoder.decodeHeader(oldBytes)) + assert(decoder.decodeHeader(oldBytes) == Some((w, h, chan))) + decoder.decodeHeaderBytesEqLemma(oldBytes, bytes) + check(decoder.decodeHeader(bytes) == Some((w, h, chan))) + } + + decoder.decodeLemma(outPos + Padding) + val SomeMut(decoder.DecodedResult(actuallyDecoded, ww, hh, cchan)) = decoder.decode(bytes, outPos + Padding) + check(ww == w) + check(hh == h) + check(cchan == chan) + assert(actuallyDecoded == decoder.decodeLoopPure(initDecoded.index, initDecoded.pixels, pxPrev, HeaderSize, outPos, 0)._2) + assert(arraysEq(pixels, actuallyDecoded, 0, pixels.length)) + } + + SomeMut(EncodedResult(bytes, outPos + Padding)) + } + + @opaque + @inlineOnce + def encodeLoop(index: Array[Int], bytes: Array[Byte], pxPrev: Int, run0: Long, outPos0: Long, pxPos: Long, @ghost decoded: GhostDecoded)(using Ctx): EncodingIteration = { + decreases(pixels.length - pxPos) + require(rangesInv(index.length, bytes.length, run0, outPos0, pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(run0, outPos0, pxPos)) + require((chan == 3) ==> (Pixel.a(pxPrev) == 255.toByte)) + require(decoded.inPos == outPos0) + require(decoded.index == index) + require(decoded.pixels.length == pixels.length) + require(pxPosInv(decoded.pxPos)) + require(decoded.pxPos + chan <= decoded.pixels.length) + require(decoded.pxPos + chan * run0 == pxPos) + require(arraysEq(decoded.pixels, pixels, 0, decoded.pxPos)) + require((run0 > 0) ==> samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPos, chan)) + + @ghost val oldDecoded = freshCopy(decoded) + @ghost val oldBytes = freshCopy(bytes) + val EncodeSingleStepResult(px, outPos2, run1) = encodeSingleStep(index, bytes, pxPrev, run0, outPos0, pxPos, decoded) + ghostExpr { + assert(decoded.pixels.length == pixels.length) + check(bytes.length == maxSize) + check(oldBytes.length == bytes.length) + check(index.length == 64) + check(index == decoded.index) + check(HeaderSize <= outPos2 && outPos2 <= maxSize - Padding) + assert(rangesInv(index.length, bytes.length, run1, outPos2, pxPos)) + assert(positionsIneqInv(run1, outPos2, pxPos + chan)) + assert((chan == 3) ==> (Pixel.a(px) == 255.toByte)) + assert(decoded.inPos == outPos2) + assert(samePixels(pixels, px, pxPos, chan)) // Precond 2 very slow (~100s), precond 4 slow (~40s) + assert(decoded.index == index) + assert(decoded.pxPos + chan * run1 == pxPos + chan) + assert(decoded.pxPos % chan == 0) // Slow (~50s) + assert(0 <= decoded.pxPos) + assert(arraysEq(oldBytes, bytes, 0, outPos0)) // Precond 2 slow (~45s) + assert(decoded.pixels.length == pixels.length) + assert((outPos0 < outPos2) ==> decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decoded)) + + modMultLemma(w, h, chan) + assert((w * h * chan) % chan == 0) + assert(pixels.length == w * h * chan) + assert(pixels.length % chan == 0) + assert(pxPosInv(pxPos)) + unfold(pxPosInv(pxPos)) + assert(pxPos % chan == 0) + + assert(0 <= outPos0 && outPos0 <= bytes.length - Padding) + assert(outPos0 <= bytes.length) + } + + if (pxPos + chan < pixels.length) { + val pxPosPlusChan = pxPos + chan + ghostExpr { + assert(pxPosPlusChan >= 0) + modSumLemma(pxPos, chan) + assert(pxPosPlusChan % chan == 0) + modLeqLemma(pxPosPlusChan, pixels.length, chan) + assert(pxPosPlusChan + chan <= pixels.length) + assert(rangesInv(index.length, bytes.length, run1, outPos2, pxPosPlusChan)) // Slow (~60s) + + if (outPos0 == outPos2) { + assert(run1 == run0 + 1) + assert(px == pxPrev) + check(decoded == oldDecoded) + + check(decoded.pxPos + chan <= decoded.pixels.length) + check(decoded.pxPos % chan == 0) + + check(decoded.pxPos < decoded.pixels.length) + check(decoded.pxPos + chan <= decoded.pixels.length) + check(arraysEq(decoded.pixels, pixels, 0, decoded.pxPos)) + + samePixelsSingleElementRange(pixels, px, decoded.pxPos, chan) + if (run0 != 0) { + assert(0 <= decoded.pxPos) + assert(decoded.pxPos < pxPos) + assert(samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPos, chan)) + samePixelsForallCombinedLemma(pixels, px, decoded.pxPos, pxPos, pxPosPlusChan, chan) + check(samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPosPlusChan, chan)) + } else { + assert(decoded.pxPos == pxPos) + check(samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPosPlusChan, chan)) + } + check((run1 > 0) ==> samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPosPlusChan, chan)) + check(decoded.pxPos + chan * run1 == pxPosPlusChan) + } else { + assert(outPos0 < outPos2) + assert(run1 == 0) + assert(decoded.pxPos == pxPos + chan) + assert(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decoded)) + + check(decoded.pxPos < decoded.pixels.length) + modLeqLemma(decoded.pxPos, decoded.pixels.length, chan) + check(decoded.pxPos + chan <= decoded.pixels.length) + + assert(arraysEq(pixels, decoded.pixels, 0, decoded.pxPos)) + arraysEqSymLemma(pixels, decoded.pixels, 0, decoded.pxPos) + check(arraysEq(decoded.pixels, pixels, 0, decoded.pxPos)) + check((run1 > 0) ==> samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPosPlusChan, chan)) + check(decoded.pxPos + chan * run1 == pxPosPlusChan) + } + } + @ghost val bytesPreRec = freshCopy(bytes) + @ghost val decodedPreRec = freshCopy(decoded) + // We do not de-structure `res` here, as this causes GenC to emit extra-instruction, which hinder TCE by gcc and clang. + val res = encodeLoop(index, bytes, px, run1, outPos2, pxPosPlusChan, decoded) + + ghostExpr { + val EncodingIteration(pxRes, outPosRes) = res + check(bytes.length == maxSize && index == decoded.index) + check(decoded.pixels.length == pixels.length) + check(oldBytes.length == bytes.length) + assert(index.length == 64 && decoded.index.length == 64) + check(HeaderSize <= outPosRes && outPosRes <= maxSize - Padding) + check(outPos0 < outPosRes) + check(outPos2 < outPosRes) + check(outPosRes <= bytes.length - Padding) + assert(decodeLoopEncodeProp(bytes, px, outPos2, outPosRes, decodedPreRec, pxRes, decoded)) + assert(arraysEq(bytesPreRec, bytes, 0, outPos2)) + + // From encodeSingleStep, which holds for bytes and decoded prior to the recursive call: + // (outPos0 < outPos2) ==> decodeEncodeProp(bytesPreRec, pxPrev, pxPos, outPos0, outPos2, oldDecoded, decodedPreRec) + + if (outPos0 == outPos2) { + assert(px == pxPrev) + assert(oldDecoded == decodedPreRec) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPosRes, oldDecoded, pxRes, decoded)) + } else { + assert(outPos0 < outPos2) + assert(decodeLoopEncodeProp(bytesPreRec, pxPrev, outPos0, outPos2, oldDecoded, px, decodedPreRec)) + + arraysEqDropLeftLemma(bytesPreRec, bytes, 0, outPos0, outPos2) + assert(arraysEq(bytesPreRec, bytes, outPos0, outPos2)) + decodeLoopEncodePropBytesEqLemma(bytesPreRec, bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decodedPreRec) + assert(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decodedPreRec)) + + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + + val (ix1, pix1, decIter1) = decoder.decodeLoopPure(oldDecoded.index, oldDecoded.pixels, pxPrev, outPos0, outPos2, oldDecoded.pxPos) + assert(decIter1.pxPos == decodedPreRec.pxPos) + assert(decIter1.inPos == decodedPreRec.inPos) + assert(decIter1.inPos == outPos2) + assert(decIter1.px == px) + + assert(decodedPreRec.pxPos + chan * run1 == pxPos + chan) + assert(decodedPreRec.pxPos < pixels.length) + assert(decIter1.pxPos + chan <= pixels.length) + assert(decIter1.pxPos < pixels.length) + assert(ix1 == decodedPreRec.index) + assert(pix1 == decodedPreRec.pixels) + + assert(decodedPreRec.index.length == 64 && decodedPreRec.pixels.length == pixels.length) + assert(HeaderSize <= outPos2 && outPos2 <= bytes.length - Padding) + assert(outPos2 <= bytes.length) + assert(0 <= decodedPreRec.pxPos && decodedPreRec.pxPos <= decodedPreRec.pixels.length) + assert(decodedPreRec.pxPos % chan == 0) + assert(decodedPreRec.index.length == 64) + assert(decodedPreRec.pixels.length == pixels.length) + assert(decodedPreRec.pixels.length == w * h * chan) + assert((w * h * chan) % chan == 0) + assert(decodedPreRec.pixels.length % chan == 0) + assert(decodedPreRec.pxPos <= w * h * chan) + assert(decoder.pxPosInv(decodedPreRec.pxPos)) + val (ix2, pix2, decIter2) = decoder.decodeLoopPure(decodedPreRec.index, decodedPreRec.pixels, px, outPos2, outPosRes, decodedPreRec.pxPos) + assert(decIter2.pxPos == decoded.pxPos) + assert(decIter2.inPos == decoded.inPos) + assert(ix2 == decoded.index) + assert(pix2 == decoded.pixels) + + assert(oldDecoded.pxPos % chan == 0) + assert(HeaderSize <= outPos0 && outPos0 <= bytes.length) + assert(oldDecoded.index.length == 64) + assert(oldDecoded.pixels.length == pixels.length) + assert(oldDecoded.pixels.length == w * h * chan) + assert((w * h * chan) % chan == 0) + assert(oldDecoded.pixels.length % chan == 0) + assert(0 <= oldDecoded.pxPos && oldDecoded.pxPos <= oldDecoded.pixels.length) + val (ix3, pix3, decIter3) = decoder.decodeLoopPure(oldDecoded.index, oldDecoded.pixels, pxPrev, outPos0, outPosRes, oldDecoded.pxPos) + + assert(outPosRes <= bytes.length - Padding) + assert(outPos0 < outPos2) + assert(outPos2 < outPosRes) + assert(ix1 == decodedPreRec.index) + assert(pix1 == decodedPreRec.pixels) + assert(decIter1.px == px) + assert(decIter1.inPos == outPos2) + assert(decIter1.pxPos + chan <= pixels.length) + assert(decIter1.pxPos < pixels.length) + assert(oldDecoded.pixels.length == pixels.length) + assert(decIter1.pxPos + chan <= oldDecoded.pixels.length) + assert(decIter1.pxPos < oldDecoded.pixels.length) + decoder.decodeLoopPureMergedLemma(oldDecoded.index, oldDecoded.pixels, pxPrev, outPos0, outPos2, outPosRes, oldDecoded.pxPos) + assert(ix2 == ix3) + assert(pix2 == pix3) + assert(decIter2 == decIter3) + + assert(decIter3.pxPos == decoded.pxPos) + assert(decIter3.inPos == decoded.inPos) + assert(ix3 == decoded.index) + assert(pix3 == decoded.pixels) + + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPosRes, oldDecoded, pxRes, decoded)) + } + check(oldBytes.length == bytes.length) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPosRes, oldDecoded, pxRes, decoded)) + + arraysEqDropRightLemma(bytesPreRec, bytes, 0, outPos0, outPos2) + arraysEqTransLemma(oldBytes, bytesPreRec, bytes, 0, outPos0) + check(arraysEq(oldBytes, bytes, 0, outPos0)) + check(decoded.pxPos == pxEnd + chan) + } + + res + } else { + @ghost val bytesPrePadded = freshCopy(bytes) + bytes(outPos2.toInt + Padding - 1) = 1 + ghostExpr { + check(outPos0 < outPos2) + + // Showing arraysEq(oldBytes, bytes, 0, outPos0) + check(oldBytes.length == bytesPrePadded.length) + check(bytesPrePadded.length == bytes.length) + assert(arraysEq(oldBytes, bytesPrePadded, 0, outPos0)) + + updatedAtArraysEq(bytesPrePadded, outPos2 + Padding - 1, 1) + assert(arraysEq(bytesPrePadded, bytes, 0, outPos2 + Padding - 1)) + arraysEqDropRightLemma(bytesPrePadded, bytes, 0, outPos0, outPos2 + Padding - 1) + assert(arraysEq(bytesPrePadded, bytes, 0, outPos0)) + arraysEqTransLemma(oldBytes, bytesPrePadded, bytes, 0, outPos0) + check(arraysEq(oldBytes, bytes, 0, outPos0)) + + assert(pxPos + chan == pixels.length) + assert(pxPos == pixels.length - chan) + assert((pxPos == pxEnd) ==> (run1 == 0)) + assert(pxPos == pxEnd) + assert(run1 == 0) + assert(decoded.pxPos + chan * run1 == pxPos + chan) + assert(decoded.pxPos == pxPos + chan) + check(decoded.pxPos == pxEnd + chan) + + // Showing decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPosRes, oldDecoded, pxRes, decoded) + assert(decodeLoopEncodeProp(bytesPrePadded, pxPrev, outPos0, outPos2, oldDecoded, px, decoded)) + // The two arraysEqX are for the precond arraysEq(oldBytes, bytes, outPos0, outPos2) + arraysEqDropRightLemma(bytesPrePadded, bytes, 0, outPos2, outPos2 + Padding - 1) + arraysEqDropLeftLemma(bytesPrePadded, bytes, 0, outPos0, outPos2) + decodeLoopEncodePropBytesEqLemma(bytesPrePadded, bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decoded) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, oldDecoded, px, decoded)) + } + EncodingIteration(px, outPos2) + } + }.ensuring { case EncodingIteration(pxRes, outPosRes) => + bytes.length == maxSize &&& + index.length == 64 &&& + index == decoded.index &&& + decoded.pixels.length == pixels.length &&& + HeaderSize <= outPosRes &&& outPosRes <= maxSize - Padding &&& + outPos0 < outPosRes &&& + old(bytes).length == bytes.length &&& + arraysEq(old(bytes), bytes, 0, outPos0) &&& + decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPosRes, old(decoded), pxRes, decoded) &&& + decoded.pxPos == pxEnd + chan + } + + @opaque + @inlineOnce + def writeHeader(bytes: Array[Byte])(using Ctx): Unit = { + require(bytes.length >= HeaderSize) + write32(bytes, 0, MagicNumber) + assert(read32(bytes, 0) == MagicNumber) + + write32(bytes, 4, w.toInt) + assert(read32(bytes, 4) == w.toInt) + + write32(bytes, 8, h.toInt) + assert(read32(bytes, 8) == h.toInt) + + bytes(12) = chan.toByte + assert(bytes(12) == chan.toByte) + + bytes(13) = 0 // Color-space (unused) + }.ensuring { _ => + old(bytes).length == bytes.length &&& + read32(bytes, 0) == MagicNumber &&& + read32(bytes, 4) == w.toInt &&& + read32(bytes, 8) == h.toInt &&& + bytes(12) == chan.toByte &&& + bytes(13) == 0 + } + + @opaque + @inlineOnce + def encodeSingleStep(index: Array[Int], bytes: Array[Byte], pxPrev: Int, run0: Long, outPos0: Long, pxPos: Long, @ghost decoded: GhostDecoded)(using Ctx): EncodeSingleStepResult = { + require(rangesInv(index.length, bytes.length, run0, outPos0, pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(run0, outPos0, pxPos)) + require((chan == 3) ==> (Pixel.a(pxPrev) == 255.toByte)) + require(decoded.inPos == outPos0) + require(decoded.index == index) + require(decoded.pixels.length == pixels.length) + require(pxPosInv(decoded.pxPos)) + require(decoded.pxPos + chan <= decoded.pixels.length) + require(decoded.pxPos + chan * run0 == pxPos) + require(arraysEq(decoded.pixels, pixels, 0, decoded.pxPos)) + require((run0 > 0) ==> samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPos, chan)) + + assert(decoded.index.length == 64) + + val px = + if (chan == 4) read32(pixels, pxPos.toInt) + else { + assert(chan == 3 && Pixel.a(pxPrev) == 255.toByte) + Pixel.fromRgba(pixels(pxPos.toInt), pixels(pxPos.toInt + 1), pixels(pxPos.toInt + 2), Pixel.a(pxPrev)) + } + assert((chan == 3) ==> (Pixel.a(px) == Pixel.a(pxPrev))) + assert(samePixels(pixels, px, pxPos, chan)) // Precond 2 slow (~55s) + given li: LoopIter = LoopIter(px, pxPrev, pxPos) + + @ghost val bytesPre = freshCopy(bytes) + val runUpd = updateRun(bytes, run0, outPos0) + ghostExpr { + withinBoundsLemma(run0, outPos0, pxPos) + assert((run0 + bool2int(px == pxPrev) > 0 && runUpd.reset) ==> updateRunProp(bytes, run0, outPos0, runUpd)) + assert(bytesPre.length == bytes.length) + assert(arraysEq(bytesPre, bytes, 0, outPos0)) + } + val run1 = runUpd.run + val outPos1 = runUpd.outPos + + @ghost val indexPre = freshCopy(index) + + val outPos2 = if (px != pxPrev) { + assert(run1 == 0) + assert(bool2int(px == pxPrev) == 0) + + ghostExpr { + if (run0 == 0) { + positionsIneqIncrementedLemma(run1, outPos1, pxPos) + check(positionsIneqInv(run1, outPos1 + chan + 1, pxPos + chan)) + } else { + assert(runUpd.reset && outPos1 <= outPos0 + 2 && run0 > 0) + val lhs0 = chan * (outPos0 - HeaderSize + chan) + val rhs0 = (chan + 1) * pxPos + assert(lhs0 <= rhs0) + val lhs1 = lhs0 + (chan + 1) * chan + val rhs1 = rhs0 + (chan + 1) * chan + assert(lhs1 <= rhs1) // Slow (~45s) + val lhs2 = chan * (outPos0 - HeaderSize + chan + chan + 1) + val rhs2 = (chan + 1) * (pxPos + chan) + assert(lhs2 == lhs1) + assert(rhs2 == rhs1) + assert(lhs2 <= rhs2) + val lhs3 = chan * (outPos1 - HeaderSize + chan + 1) + assert((lhs3 <= lhs2) because (chan >= 3 && outPos1 <= outPos0 + 2)) + assert(lhs3 <= rhs2) + check(positionsIneqInv(0, outPos1 + chan + 1, pxPos + chan)) + } + assert(positionsIneqInv(0, outPos1 + chan + 1, pxPos + chan)) + } + + @ghost val bytesPreEncNoRun = freshCopy(bytes) + assert(arraysEq(bytesPre, bytesPreEncNoRun, 0, outPos0)) + val outPos2 = encodeNoRun(index, bytes, outPos1) + + ghostExpr { + assert(bytesPreEncNoRun.length == bytes.length) + assert(arraysEq(bytesPreEncNoRun, bytes, 0, outPos1)) + check(positionsIneqInv(0, outPos2, pxPos + chan)) + check(HeaderSize <= outPos2 && outPos1 < outPos2 && outPos2 <= maxSize - Padding) + check(rangesInv(index.length, bytes.length, 0, outPos1, pxPos)) + + // Note: bool2int(px == pxPrev) == 0 + check(((run0 > 0 && runUpd.reset) ==> updateRunProp(bytes, run0, outPos0, runUpd)) because { + if (run0 > 0 && runUpd.reset) { + updateRunPropBytesEqLemma(bytesPreEncNoRun, run0, outPos0, runUpd, bytes) + trivial + } else trivial + }) + + assert((outPos2 <= bytes.length - Padding) because (bytes.length == maxSize)) + check(positionsIneqInv(0, outPos1, pxPos)) + // Precond 6 slow (~40s) + check(encodeNoRunProp(indexPre, index, bytes, outPos1, outPos2)) + + arraysEqDropRightLemma(bytesPreEncNoRun, bytes, 0, outPos0, outPos1) + arraysEqTransLemma(bytesPre, bytesPreEncNoRun, bytes, 0, outPos0) + check(arraysEq(bytesPre, bytes, 0, outPos0)) + } + + outPos2 + } else { + ghostExpr { + if (runUpd.reset) { + assert(outPos1 <= outPos0 + 2) + assert(run1 == 0) + val lhs0 = chan * (outPos0 - HeaderSize + run0 * chan) + val rhs0 = (chan + 1) * pxPos + assert(lhs0 <= rhs0) + val lhs1 = lhs0 + (chan + 1) * chan + val rhs1 = rhs0 + (chan + 1) * chan + assert(lhs1 <= rhs1) // Slow (~75s) + val lhs2 = chan * (outPos0 - HeaderSize + run0 * chan + chan + 1) + val rhs2 = (chan + 1) * (pxPos + chan) + assert(lhs1 == lhs2) + assert(rhs1 == rhs2) + assert(lhs2 <= rhs2) + val lhs3 = chan * (outPos0 - HeaderSize + run0 * chan + chan + 1) + assert(lhs3 <= rhs2) + val lhs4 = chan * (outPos1 - HeaderSize) + assert((lhs4 <= lhs3) because (chan >= 3 && outPos1 <= outPos0 + 2 && run0 >= 0)) + assert(lhs4 <= rhs2) + check(positionsIneqInv(run1, outPos1, pxPos + chan)) + } else { + assert(outPos1 == outPos0 && run1 == run0 + 1 && run0 < 0x2020) + assert(positionsIneqInv(run0, outPos0, pxPos)) + val lhs0 = chan * ((outPos0 - HeaderSize) + chan * run0) + val rhs0 = (chan + 1) * pxPos + assert(lhs0 <= rhs0) + val lhs1 = lhs0 + (chan + 1) * chan + val rhs1 = rhs0 + (chan + 1) * chan + assert(lhs1 <= rhs1) // Slow (~60s) + val lhs2 = chan * ((outPos0 - HeaderSize) + chan * run0 + chan + 1) + val rhs2 = (chan + 1) * (pxPos + chan) + assert(lhs2 == lhs1) + assert(rhs2 == rhs1) + assert(lhs2 <= rhs2) + val lhs3 = chan * ((outPos0 - HeaderSize) + chan * (run0 + 1)) + assert(lhs3 <= lhs2) // We actually have lhs3 == lhs2 but lhs3 <= lhs2 is sufficient and takes way less time to verify + assert(lhs3 <= rhs2) + val lhs4 = chan * ((outPos1 - HeaderSize) + chan * (run0 + 1)) + assert(lhs4 == lhs3) + assert(lhs4 <= rhs2) + check(positionsIneqInv(run1, outPos1, pxPos + chan) because (run1 == run0 + 1)) + } + + check(rangesInv(index.length, bytes.length, run1, outPos1, pxPos)) + check(positionsIneqInv(run1, outPos1, pxPos + chan)) + } + outPos1 + } + assert(HeaderSize <= outPos2 && outPos2 <= maxSize - Padding) + assert((px != pxPrev) ==> (outPos1 < outPos2)) + assert((outPos2 <= bytes.length - Padding) because (bytes.length == maxSize)) + assert((run0 + bool2int(px == pxPrev) > 0 && runUpd.reset) ==> updateRunProp(bytes, run0, outPos0, runUpd)) + assert((px != pxPrev) ==> encodeNoRunProp(indexPre, index, bytes, outPos1, outPos2)) + + @ghost val newDecoded = if (run0 > 0 && runUpd.reset && px != pxPrev) { // Note: bool2int(px == pxPrev) == 0 + check(outPos1 < outPos2) + assert(updateRunProp(bytes, run0, outPos0, runUpd)) + assert(encodeNoRunProp(indexPre, index, bytes, outPos1, outPos2)) + assert(indexPre.updated(colorPos(px), px) == index) + + assert(decoded.pxPos + chan * run0 <= pixels.length) + val decoded1 = { + val decoded1 = decodeUpdateRunPass(bytes, run0, outPos0, runUpd, decoded) + check(pixels.length == decoded1.pixels.length) + check(arraysEq(pixels, decoded1.pixels, 0, decoded1.pxPos)) + arraysEqDropLeftLemma(pixels, decoded1.pixels, 0, pxPos, decoded1.pxPos) + arraysEqSymLemma(pixels, decoded1.pixels, 0, pxPos) + check(arraysEq(decoded1.pixels, pixels, 0, pxPos)) + check(decoded1.index == indexPre) + + assert(decoded1.pxPos == decoded.pxPos + chan * run0) + assert(decoded.pxPos + chan * run0 == pxPos) + check(decoded1.pxPos == pxPos) + check(decoded1.inPos == outPos1) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos1, decoded, pxPrev, decoded1)) + + freshCopy(decoded1) + } + val decoded2 = { + val decoded2 = decodeEncodeNoRunPass(indexPre, index, bytes, outPos1, outPos2, decoded1) + check(pixels.length == decoded2.pixels.length) + check(decoded2.pxPos == decoded1.pxPos + chan) + check(decoded2.pxPos % chan == 0) + check(decoded2.inPos == outPos2) + check(arraysEq(pixels, decoded2.pixels, 0, decoded2.pxPos)) + check(decoded2.index == decoded1.index.updated(colorPos(px), px)) + assert(decoded1.pxPos == pxPos) + assert(pxPos + chan <= pixels.length) + assert(decoded1.pxPos + chan <= pixels.length) + assert(decoded1.pixels.length == pixels.length) + assert(decoded1.pxPos + chan <= decoded1.pixels.length) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos1, outPos2, decoded1, px, decoded2)) + + freshCopy(decoded2) + } + + + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + + val (ix3, pix3, decIter3) = { + assert(decoded.pixels.length == pixels.length) + assert(pxPosInv(decoded.pxPos)) + assert(0 <= decoded.pxPos && decoded.pxPos <= decoded.pixels.length) + assert(outPosInv(outPos0)) + unfold(outPosInv(outPos0)) + assert(HeaderSize <= outPos0 && outPos0 <= bytes.length - Padding) + assert(decoder.inPosInv(outPos0)) + assert(decoded.index.length == 64) + assert(outPos0 < outPos1 && outPos1 <= bytes.length - Padding) + // For precond 2, 4, 5 + assert(pxPosInv(decoded.pxPos)) + unfold(pxPosInv(decoded.pxPos)) + assert(0 <= decoded.pxPos && decoded.pxPos <= pixels.length && decoded.pxPos % chan == 0) + assert(decoded.pxPos + chan * run0 == pxPos) + assert(pxPos + chan <= pixels.length) + assert(decoded.pxPos + chan <= pixels.length) + assert(decoded.pixels.length == pixels.length) + assert(decoded.pxPos + chan <= decoded.pixels.length) // 5 + assert(pixels.length == w * h * chan) // 2 (and part of 4) + assert(decoder.pxPosInv(decoded.pxPos)) // 4 Slow (~60s) + val (ix1, pix1, decIter1) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos1, decoded.pxPos) + assert(decIter1.pxPos == decoded1.pxPos) + assert(decIter1.pxPos == pxPos) + assert(decIter1.inPos == outPos1) + + decoder.doDecodeNext(decoded.index, pxPrev, outPos0) match { + case (decoder.Run(r), inPos) => + assert(decIter1.px == pxPrev) + () + case _ => + assert(false) + () + } + + assert(decIter1.px == pxPrev) + assert(decIter1.remainingRun == 0) + assert(ix1 == decoded1.index) + assert(pix1 == decoded1.pixels) + + assert(decoded1.pxPos == pxPos) + assert(decoded1.pixels.length == pixels.length) + assert(pxPos + chan <= pixels.length) + assert(decoded1.pxPos + chan <= decoded1.pixels.length) + assert(outPos1 < outPos2 && outPos2 <= bytes.length - Padding) + assert(pixels.length == w * h * chan) + assert(decoder.pxPosInv(decoded1.pxPos)) + val (ix2, pix2, decIter2) = decoder.decodeLoopPure(decoded1.index, decoded1.pixels, decIter1.px, outPos1, outPos2, decoded1.pxPos) + assert(decIter2.pxPos == decoded2.pxPos) + assert(decIter2.inPos == outPos2) + assert(decIter2.remainingRun == 0) + assert(decIter2.px == px) + assert(ix2 == decoded2.index) + assert(pix2 == decoded2.pixels) + + val (ix3, pix3, decIter3) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos) + assert(pxPos < pixels.length && pxPos + chan <= pixels.length) + assert(decIter1.pxPos < decoded.pixels.length) + assert(decIter1.pxPos + chan <= decoded.pixels.length) // Very slow (~90s) + // Precond 7 slow (~45s) + decoder.decodeLoopPureMergedLemma(decoded.index, decoded.pixels, pxPrev, outPos0, outPos1, outPos2, decoded.pxPos) + check(ix3 == decoded2.index) + check(pix3 == decoded2.pixels) + check(decIter3.pxPos == decoded2.pxPos) + check(decIter3.inPos == outPos2) + check(decIter3.remainingRun == 0) + check(decIter3.px == px) + + assert(arraysEq(pixels, decoded2.pixels, 0, decoded2.pxPos)) + assert(decIter3.pxPos == pxPos + chan) + assert(pix3 == decoded2.pixels && decoded2.pxPos == pxPos + chan) + check(arraysEq(pixels, pix3, 0, decIter3.pxPos)) + + (freshCopy(ix3), freshCopy(pix3), decIter3) + } + assert((ix3, pix3, decIter3) == decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos)) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, decoded, px, decoded2)) // Very slow (from ~75s up to ~140s) + + check(index == decoded2.index) + assert(run1 == 0) + check(decoded2.pxPos + chan * run1 == pxPos + chan) + + assert(outPos0 != outPos2) + check((outPos0 == outPos2) ==> (decoded == decoded2 && run1 == run0 + 1 && px == pxPrev)) + + freshCopy(decoded2) + } else if (!(run0 > 0 && runUpd.reset) && px != pxPrev) { // Note: bool2int(px == pxPrev) == 0 + check(outPos0 == outPos1) + assert(encodeNoRunProp(indexPre, index, bytes, outPos1, outPos2)) + assert(indexPre.updated(colorPos(px), px) == index) + val decoded2 = decodeEncodeNoRunPass(indexPre, index, bytes, outPos1, outPos2, decoded) + assert(decoded2.pixels.length == pixels.length) + assert(decoded2.pxPos % chan == 0) + assert(decoded2.pxPos == decoded.pxPos + chan) + assert(arraysEq(pixels, decoded2.pixels, 0, decoded2.pxPos)) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, decoded, px, decoded2)) + check(index == decoded2.index) + + assert(run0 == 0 && run1 == 0) + assert(decoded.pxPos == pxPos) + check(decoded2.pxPos + chan * run1 == pxPos + chan) + + assert(outPos0 != outPos2) + check((outPos0 == outPos2) ==> (decoded == decoded2 && run1 == run0 + 1 && px == pxPrev)) + + freshCopy(decoded2) + } else if (run0 + 1 > 0 && runUpd.reset && px == pxPrev) { // Note: bool2int(px == pxPrev) == 1 + assert(decoded.pxPos + chan * run0 <= pixels.length) + check(outPos1 == outPos2) + assert(updateRunProp(bytes, run0, outPos0, runUpd)) + val decoded2 = decodeUpdateRunPass(bytes, run0, outPos0, runUpd, decoded) + assert(decoded2.pixels.length == pixels.length) + assert(decoded2.pxPos == decoded.pxPos + chan * (run0 + 1)) + assert(decoded.pxPos + chan * run0 == pxPos) + assert(decoded.pxPos + chan * run0 + chan == pxPos + chan) + assert(decoded.pxPos + chan * (run0 + 1) == pxPos + chan) + assert(decoded2.pxPos == pxPos + chan) + assert(decoded2.pxPos % chan == 0) + assert(arraysEq(pixels, decoded2.pixels, 0, decoded2.pxPos)) + check(decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, decoded, px, decoded2)) + check(index == decoded2.index) + + assert(run1 == 0) + check(decoded2.pxPos + chan * run1 == pxPos + chan) + + assert(outPos0 != outPos2) + check((outPos0 == outPos2) ==> (decoded == decoded2 && run1 == run0 + 1 && px == pxPrev)) + + freshCopy(decoded2) + } else { + check(outPos0 == outPos1 && outPos1 == outPos2) + check(run1 == run0 + 1 && px == pxPrev) + + assert(decoded.pxPos + chan * run0 == pxPos) + assert(decoded.pxPos + chan * run0 + chan == pxPos + chan) + assert(decoded.pxPos + chan * (run0 + 1) == pxPos + chan) + check(decoded.pxPos + chan * run1 == pxPos + chan) + + assert(outPos0 == outPos2) + check((outPos0 == outPos2) ==> (decoded == decoded && run1 == run0 + 1 && px == pxPrev)) + + freshCopy(decoded) + } + + ghostExpr { + check(index == newDecoded.index) + check((outPos0 < outPos2) ==> decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, decoded, px, newDecoded)) + assert(pxPosInv(pxPos)) + unfold(pxPosInv(pxPos)) + check(pxPos % chan == 0) + check(pixels.length % chan == 0) // Slow (~75) + check(samePixels(pixels, px, pxPos, chan)) + check(arraysEq(bytesPre, bytes, 0, outPos0)) + check(rangesInv(index.length, bytes.length, run1, outPos2, pxPos)) + check(positionsIneqInv(run1, outPos2, pxPos + chan)) // Slow (~45s) + check(pxPosInv(decoded.pxPos)) + check(newDecoded.pxPos + chan * run1 == pxPos + chan) + check((outPos0 == outPos2) ==> (decoded == newDecoded && run1 == run0 + 1 && px == pxPrev)) + + decoded.index = freshCopy(newDecoded.index) + decoded.pixels = freshCopy(newDecoded.pixels) + decoded.inPos = newDecoded.inPos + decoded.pxPos = newDecoded.pxPos + } + + EncodeSingleStepResult(px, outPos2, run1) + }.ensuring { case EncodeSingleStepResult(px, outPos2, run1) => + // Bytes and index length are unchanged + bytes.length == maxSize &&& + index.length == 64 &&& + // Range for outPos2 + outPos0 <= outPos2 &&& + outPosInv(outPos2) &&& + // Run-related props + runInv(run1) &&& + ((run1 > 0) ==> (bytes == old(bytes) && outPos2 == outPos0)) &&& + ((pxPos == pxEnd) ==> (run1 == 0)) &&& + // Ranges and inequality still holding + positionsIneqInv(run1, outPos2, pxPos + chan) &&& + rangesInv(index.length, bytes.length, run1, outPos2, pxPos) &&& + // Alpha-channel unchanged when we only have 3 channels + ((chan == 3) ==> (Pixel.a(px) == 255.toByte)) &&& + index == decoded.index &&& + decoded.pixels.length == pixels.length &&& + pxPosInv(decoded.pxPos) &&& + decoded.pxPos + chan * run1 == pxPos + chan &&& + ((outPos0 < outPos2) ==> decodeLoopEncodeProp(bytes, pxPrev, outPos0, outPos2, old(decoded), px, decoded)) &&& + ((outPos0 == outPos2) ==> (old(decoded) == decoded && run1 == run0 + 1 && px == pxPrev)) &&& + samePixels(pixels, px, pxPos, chan) &&& // Precond 2 slow (~70s) + arraysEq(old(bytes), bytes, 0, outPos0) + } + + @inline + def updateRun(bytes: Array[Byte], run0: Long, outPos0: Long)(using Ctx, LoopIter): RunUpdate = { + require(bytes.length == maxSize) + require(runInv(run0)) + require(outPosInv(outPos0)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(run0, outPos0, pxPos)) + + ghostExpr { + withinBoundsLemma(run0, outPos0, pxPos) + assert(outPos0 + run0 * chan + Padding <= maxSize) + assert(outPos0 + 2 <= maxSize) + } + @ghost val bytesPre = freshCopy(bytes) + + var run = run0 + var outPos = outPos0 + + if (px == pxPrev) + run += 1 + + var runReset = false + if (run > 0 && (run == 62 || px != pxPrev || pxPos == pxEnd)) { + val b1 = (OpRun | (run - 1)).toByte + bytes(outPos.toInt) = b1 + outPos += 1 + ghostExpr(check(decoder.decodeRun(b1) == run - 1)) + run = 0 + runReset = true + ghostExpr { + updatedAtArraysEq(bytesPre, outPos0, b1) + check(arraysEq(bytesPre, bytes, 0, outPos0)) + } + } else { + ghostExpr { + assert(bytesPre == bytes) + eqImpliesArraysEq(bytesPre, bytes, 0, outPos0) + check(arraysEq(bytesPre, bytes, 0, outPos0)) + } + } + ghostExpr(check(arraysEq(bytesPre, bytes, 0, outPos0))) + + RunUpdate(runReset, run, outPos) + } ensuring { ru => + old(bytes).length == bytes.length &&& + arraysEq(old(bytes), bytes, 0, outPos0) &&& + runInv(ru.run) &&& + outPosInv(ru.outPos) &&& + (ru.reset ==> (ru.run == 0 && ru.outPos == outPos0 + 1)) &&& + (!ru.reset ==> (ru.outPos == outPos0)) &&& + ((px != pxPrev && run0 == 0) ==> (ru.run == 0)) &&& + ((px == pxPrev && !ru.reset) ==> (ru.run == run0 + 1)) &&& + ((run0 + bool2int(px == pxPrev) > 0 && ru.reset) ==> updateRunProp(bytes, run0, outPos0, ru)) + } + + @opaque + @inlineOnce + def encodeNoRun(index: Array[Int], bytes: Array[Byte], outPos1: Long)(using Ctx, LoopIter): Long = { + require(index.length == 64) + require(bytes.length == maxSize) + require(outPosInv(outPos1)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(0, outPos1, pxPos)) + require((chan == 3) ==> (Pixel.a(px) == Pixel.a(pxPrev))) + + ghostExpr { + withinBoundsLemma(0, outPos1, pxPos) + assert(outPos1 + chan + 1 <= maxSize) + + positionsIneqIncrementedLemma(0, outPos1, pxPos) + assert(positionsIneqInv(0, outPos1 + chan + 1, pxPos + chan)) // Needed for loopInvUpperOutPosLemma + } + + @ghost val oldIndex = freshCopy(index) + @ghost val oldBytes = freshCopy(bytes) + + val indexPos = colorPos(px) + var newOutPos = outPos1 + + if (index(indexPos) == px) { + val b1 = (OpIndex | indexPos) & 0xff + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + ghostExpr { + check(newOutPos <= maxSize - Padding) + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + updatedAtArraysEq(oldBytes, outPos1, b1.toByte) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + } else { + index(indexPos) = px + + if (Pixel.a(px) == Pixel.a(pxPrev)) { + // Note: these 5 variables are declared as signed char in the reference implementation + val vr = ((Pixel.r(px).toInt & 0xff) - (Pixel.r(pxPrev).toInt & 0xff)).toByte + val vg = ((Pixel.g(px).toInt & 0xff) - (Pixel.g(pxPrev).toInt & 0xff)).toByte + val vb = ((Pixel.b(px).toInt & 0xff) - (Pixel.b(pxPrev).toInt & 0xff)).toByte + val vgR = (vr - vg).toByte + val vgB = (vb - vg).toByte + + if (vr > -3 && vr < 2 && vg > -3 && vg < 2 && vb > -3 && vb < 2) { + val b1 = OpDiff | (((vr + 2) << 4) & 0xff) | (((vg + 2) << 2) & 0xff) | ((vb + 2) & 0xff) + check(decoder.decodeDiff(pxPrev, b1) == px) + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + + ghostExpr { + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + updatedAtArraysEq(oldBytes, outPos1, b1.toByte) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + } else if (vgR > -9 && vgR < 8 && vg > -33 && vg < 32 && vgB > -9 && vgB < 8) { + val b1 = OpLuma | ((vg + 32) & 0xff) + val b2 = (((vgR + 8) << 4) & 0xff) | ((vgB + 8) & 0xff) + check(decoder.decodeLuma(pxPrev, b1, b2) == px) + bytes(newOutPos.toInt) = b1.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = b2.toByte + newOutPos += 1 + + ghostExpr { + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + updatedAtArraysEq(oldBytes, outPos1, b1.toByte, b2.toByte) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + } else { + bytes(newOutPos.toInt) = OpRgb.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.r(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.g(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.b(px) + newOutPos += 1 + + ghostExpr { + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + updatedAtArraysEq(oldBytes, outPos1, OpRgb.toByte, Pixel.r(px), Pixel.g(px), Pixel.b(px)) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + } + + ghostExpr { + loopInvUpperOutPosLemma(0, outPos1, pxPos + chan, newOutPos) // For precond. 2 of withinBoundsLemma2 + withinBoundsLemma2(0, newOutPos, pxPos + chan) + check(newOutPos <= maxSize - Padding) + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + } else { + bytes(newOutPos.toInt) = OpRgba.toByte + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.r(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.g(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.b(px) + newOutPos += 1 + bytes(newOutPos.toInt) = Pixel.a(px) + newOutPos += 1 + + ghostExpr { + updatedAtArraysEq(oldBytes, outPos1, OpRgba.toByte, Pixel.r(px), Pixel.g(px), Pixel.b(px), Pixel.a(px)) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + + loopInvUpperOutPosLemma(0, outPos1, pxPos + chan, newOutPos) // For precond. 2 of withinBoundsLemma2 + withinBoundsLemma2(0, newOutPos, pxPos + chan) + check(newOutPos <= maxSize - Padding) + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + } + } + } + + ghostExpr { + assert(newOutPos <= outPos1 + chan + 1) + check(newOutPos <= maxSize - Padding) + loopInvUpperOutPosLemma(0, outPos1, pxPos + chan, newOutPos) + check(rangesInv(index.length, bytes.length, 0, newOutPos, pxPos)) + check(positionsIneqInv(0, newOutPos, pxPos + chan)) + check(encodeNoRunProp(oldIndex, index, bytes, outPos1, newOutPos)) + check(arraysEq(oldBytes, bytes, 0, outPos1)) + } + newOutPos + } ensuring { newOutPos => + old(bytes).length == bytes.length &&& + outPosInv(newOutPos) &&& + outPos1 < newOutPos &&& + positionsIneqInv(0, newOutPos, pxPos + chan) &&& + encodeNoRunProp(old(index), index, bytes, outPos1, newOutPos) &&& + arraysEq(old(bytes), bytes, 0, outPos1) + } + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + def positionsIneqInv(run: Long, outPos: Long, pxPos: Long)(using Ctx): Boolean = + chan * (outPos - HeaderSize + chan * run) <= (chan + 1) * pxPos + + @ghost + def runInv(run: Long): Boolean = + 0 <= run && run < 62 + + @ghost + def pxPosInv(pxPos: Long)(using Ctx): Boolean = + 0 <= pxPos && pxPos <= pixels.length && pxPos % chan == 0 + + @ghost + def outPosInv(outPos: Long)(using Ctx): Boolean = + HeaderSize <= outPos && outPos <= maxSize - Padding + + @ghost + def rangesInv(run: Long, outPos: Long, pxPos: Long)(using Ctx): Boolean = + pxPosInv(pxPos) && runInv(run) && outPosInv(outPos) + + @ghost + def rangesInv(indexLen: Long, bytesLen: Long, run: Long, outPos: Long, pxPos: Long)(using Ctx): Boolean = + pxPosInv(pxPos) && runInv(run) && outPosInv(outPos) && bytesLen == maxSize && indexLen == 64 + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + @opaque + @inlineOnce + def withinBoundsLemma(run: Long, outPos: Long, pxPos: Long)(using Ctx): Unit = { + require(rangesInv(run, outPos, pxPos)) + require(positionsIneqInv(run, outPos, pxPos)) + require(pxPos + chan <= pixels.length) + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) <= (chan + 1) * (pxPos + chan)) + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) <= (chan + 1) * pixels.length) + assert(pixels.length * (chan + 1) == w * h * chan * (chan + 1)) + assert(chan * (maxSize - HeaderSize - Padding) == w * h * chan * (chan + 1)) + assert((chan + 1) * pixels.length == chan * (maxSize - HeaderSize - Padding)) + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) <= chan * (maxSize - HeaderSize - Padding)) + assert(outPos - HeaderSize + chan * run + chan + 1 <= maxSize - HeaderSize - Padding) // ~23s + }.ensuring(_ => outPos + run * chan + chan + 1 + Padding <= maxSize) + + @ghost + @opaque + @inlineOnce + def withinBoundsLemma2(run: Long, outPos: Long, pxPos: Long)(using Ctx): Unit = { + require(rangesInv(run, outPos, pxPos)) + require(positionsIneqInv(run, outPos, pxPos)) + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) <= (chan + 1) * (pxPos + chan)) + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) <= (chan + 1) * (pixels.length + chan)) + assert(pixels.length * (chan + 1) == w * h * chan * (chan + 1)) + assert(chan * (maxSize - HeaderSize - Padding) == w * h * chan * (chan + 1)) + assert((chan + 1) * pixels.length == chan * (maxSize - HeaderSize - Padding)) + + assert(chan * (outPos - HeaderSize + chan * run + chan + 1) - (chan + 1) * chan <= (chan + 1) * (pixels.length + chan) - (chan + 1) * chan) + assert(chan * (outPos - HeaderSize + chan * run) <= (chan + 1) * pixels.length) + assert(chan * (outPos - HeaderSize + chan * run) <= chan * (maxSize - HeaderSize - Padding)) + assert(outPos - HeaderSize + chan * run <= maxSize - HeaderSize - Padding) + assert(outPos + chan * run <= maxSize - Padding) + assert(outPos + chan * run + Padding <= maxSize) + }.ensuring(_ => outPos + chan * run + Padding <= maxSize) + + @ghost + @opaque + @inlineOnce + def positionsIneqIncrementedLemma(run: Long, outPos: Long, pxPos: Long)(using Ctx): Unit = { + require(rangesInv(run, outPos, pxPos)) + require(positionsIneqInv(run, outPos, pxPos)) + }.ensuring(_ => positionsIneqInv(run, outPos + chan + 1, pxPos + chan)) + + @ghost + @opaque + @inlineOnce + def loopInvUpperOutPosLemma(run: Long, oldOutPos: Long, pxPos: Long, newOutPos: Long)(using Ctx): Unit = { + require(rangesInv(run, oldOutPos, pxPos)) + require(HeaderSize <= newOutPos && newOutPos <= oldOutPos + chan + 1) + require(positionsIneqInv(run, oldOutPos + chan + 1, pxPos)) + assert(chan * ((oldOutPos - HeaderSize) + chan * run) <= (chan + 1) * pxPos) + assert(chan * (newOutPos - HeaderSize) <= chan * (oldOutPos + chan + 1 - HeaderSize)) + assert(chan * ((newOutPos - HeaderSize) + chan * run) <= chan * ((oldOutPos + chan + 1 - HeaderSize) + chan * run)) + }.ensuring(_ => positionsIneqInv(run, newOutPos, pxPos)) + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @ghost + @pure + def decodeLoopEncodeProp(bytes: Array[Byte], + pxPrev: Int, + outPos0: Long, + outPos2: Long, + decoded: GhostDecoded, + px: Int, + newDecoded: GhostDecoded)(using Ctx): Boolean = { + require(bytes.length == maxSize) + require(outPosInv(outPos0)) + require(outPosInv(outPos2)) + require(decoded.pixels.length == pixels.length) + require(decoded.index.length == 64) + require(pxPosInv(decoded.pxPos)) + require(decoded.pxPos + chan <= decoded.pixels.length) + require(outPos0 < outPos2) + + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + assert(decoded.pixels.length == w * h * chan) + assert(decoded.pixels.length % chan == 0) + val (decIndex, decPixels, decIter) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos) + + decoded.pxPos < decIter.pxPos && + decIter.pxPos <= decoded.pixels.length && + decIter.pxPos % chan == 0 && + decIter.inPos == outPos2 && + arraysEq(pixels, decPixels, 0, decIter.pxPos) && + decIter.remainingRun == 0 && + decIter.pxPos == newDecoded.pxPos && + decIter.inPos == newDecoded.inPos && + decIndex == newDecoded.index && + decPixels == newDecoded.pixels && + decIter.px == px + } + + @ghost + @pure + @opaque + @inlineOnce + def decodeLoopEncodePropBytesEqLemma(bytes1: Array[Byte], + bytes2: Array[Byte], + pxPrev: Int, + outPos0: Long, + outPos2: Long, + decoded: GhostDecoded, + px: Int, + newDecoded: GhostDecoded)(using Ctx): Unit = { + require(bytes1.length == maxSize) + require(outPosInv(outPos0)) + require(outPosInv(outPos2)) + require(decoded.pixels.length == pixels.length) + require(decoded.index.length == 64) + require(decoded.pxPos + chan <= decoded.pixels.length) + require(pxPosInv(decoded.pxPos)) + require(outPos0 < outPos2) + require(decodeLoopEncodeProp(bytes1, pxPrev, outPos0, outPos2, decoded, px, newDecoded)) + require(bytes1.length == bytes2.length) + require(arraysEq(bytes1, bytes2, outPos0, outPos2)) + require(pxPosInv(newDecoded.pxPos)) + + val ctx1 = decoder.Ctx(freshCopy(bytes1), w, h, chan) + assert(decoded.pixels.length % chan == 0) + assert(decoded.pixels.length == w * h * chan) + assert(outPos0 <= bytes1.length) + val (ix1, pix1, decIter1) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos)(using ctx1) + assert(decIter1.pxPos == newDecoded.pxPos) + assert(decIter1.pxPos <= pixels.length) + assert(decIter1.inPos == outPos2) + + val ctx2 = decoder.Ctx(freshCopy(bytes2), w, h, chan) + decoder.decodeLoopPureBytesEqLemma(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos, bytes2)(using ctx1) + val (ix2, pix2, decIter2) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, outPos2, decoded.pxPos)(using ctx2) + assert(ix1 == ix2) + assert(pix1 == pix2) + assert(decIter1 == decIter2) + check(decodeLoopEncodeProp(bytes2, pxPrev, outPos0, outPos2, decoded, px, newDecoded)) + }.ensuring(_ => decodeLoopEncodeProp(bytes2, pxPrev, outPos0, outPos2, decoded, px, newDecoded)) + + @ghost + @pure + @opaque + @inlineOnce + def decodeUpdateRunPass(bytes: Array[Byte], run0: Long, outPos0: Long, ru: RunUpdate, decoded: GhostDecoded)(using Ctx, LoopIter): GhostDecoded = { + require(bytes.length == maxSize) + require(runInv(run0)) + require(outPosInv(outPos0)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(outPos0 < bytes.length - Padding) + val run = run0 + bool2int(px == pxPrev) + require(run > 0) + require(updateRunProp(bytes, run0, outPos0, ru)) + require(outPos0 < ru.outPos && ru.outPos <= bytes.length - Padding) + require(decoded.index.length == 64) + require(decoded.pixels.length == pixels.length) + require(pxPosInv(decoded.pxPos)) + require(decoded.pxPos + chan <= decoded.pixels.length) + require(decoded.pxPos + chan * run0 == pxPos) + require(arraysEq(decoded.pixels, pixels, 0, decoded.pxPos)) + require(samePixels(pixels, px, pxPos, chan)) + require((run0 > 0) ==> samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPos, chan)) + + assert(3 <= chan && chan <= 4) + assert(decoded.pxPos + chan * run0 + chan <= pixels.length) + assert(pxPos % chan == 0) + assert(pixels.length % chan == 0) + + given dctx: decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + assert(decoded.pixels.length == w * h * chan) + assert(decoded.pixels.length % chan == 0) + assert(HeaderSize <= outPos0 && outPos0 <= bytes.length) + assert(decoded.pxPos % chan == 0) + assert(0 <= decoded.pxPos && decoded.pxPos <= pixels.length) + assert(decoder.pxPosInv(pxPos)) + val (decIndex, decPixels, decIter) = { + val (decIndex, decPixels, decIter) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos0, ru.outPos, decoded.pxPos) + check(decIter.pxPos % chan == 0) + updateRunPropAnyIndexLemma(bytes, decoded.index, run0, outPos0, ru) + + check(decIndex == decoded.index) + check(decIter.px == pxPrev) + check(decIter.inPos == ru.outPos) + check(decIter.remainingRun == 0) + check(decPixels.length == pixels.length) + + assert(decoder.writeRunPixelsInv(decoded.pxPos, run - 1, decIter.pxPos, decIter.remainingRun)) // Slow (~40-70s) and may spuriously fail to verify!! + unfold(decoder.writeRunPixelsInv(decoded.pxPos, run - 1, decIter.pxPos, decIter.remainingRun)) + assert((decoded.pxPos + chan + chan * (run - 1) <= decoded.pixels.length) ==> (decIter.remainingRun == 0 && decIter.pxPos == decoded.pxPos + chan * run)) // Slow (~60s) + assert(run - 1 <= run0) + assert(decoded.pxPos + chan * run <= decoded.pxPos + chan * run0 + chan) + assert(decoded.pxPos + chan * run == decoded.pxPos + chan + chan * (run - 1)) + check(decoded.pxPos + chan + chan * (run - 1) <= pixels.length) + check(decIter.pxPos == decoded.pxPos + chan * run) + + check(decoded.pxPos < decIter.pxPos) + check(decIter.pxPos <= decoded.pixels.length) + assert(arraysEq(decoded.pixels, decPixels, 0, decoded.pxPos)) + assert(samePixelsForall(decPixels, pxPrev, decoded.pxPos, decIter.pxPos, chan)) + arraysEqSymLemma(decoded.pixels, pixels, 0, decoded.pxPos) + arraysEqTransLemma(pixels, decoded.pixels, decPixels, 0, decoded.pxPos) + check(arraysEq(pixels, decPixels, 0, decoded.pxPos)) + + (freshCopy(decIndex), freshCopy(decPixels), decIter) + } + + val pxPosPlusChan = pxPos + chan + assert(pxPosPlusChan <= pixels.length) + modSumLemma(pxPos, chan) + assert(pxPosPlusChan % chan == 0) + samePixelsSingleElementRange(pixels, px, pxPos, chan) + assert(samePixelsForall(pixels, px, pxPos, pxPosPlusChan, chan)) + if (run0 == 0) { + assert(px == pxPrev) + assert(decoded.pxPos == pxPos) + assert(decIter.pxPos == decoded.pxPos + chan) + assert(decIter.pxPos == pxPos + chan) + assert(decIter.pxPos == pxPosPlusChan) + assert(samePixelsForall(decPixels, px, pxPos, pxPosPlusChan, chan)) + samePixelsForallArraysEq(pixels, decPixels, px, pxPos, pxPosPlusChan, chan) + arraysEqCombinedLemma2(pixels, decPixels, 0, pxPos, pxPosPlusChan) + check(arraysEq(pixels, decPixels, 0, decIter.pxPos)) + } else { + assert(samePixelsForall(pixels, pxPrev, decoded.pxPos, pxPos, chan)) + assert(decoded.pxPos + chan * run0 == pxPos && decIter.pxPos == decoded.pxPos + chan * run) // From previous assertions + assert(decoded.pxPos == pxPos - chan * run0) + assert(decIter.pxPos == pxPos - chan * run0 + chan * run) + assert(decIter.pxPos == pxPos + chan * bool2int(px == pxPrev)) + if (px == pxPrev) { + assert(decIter.pxPos == pxPos + chan) + samePixelsForallCombinedLemma(pixels, px, decoded.pxPos, pxPos, pxPosPlusChan, chan) + assert(samePixelsForall(pixels, px, decoded.pxPos, decIter.pxPos, chan)) + samePixelsForallArraysEq(pixels, decPixels, px, decoded.pxPos, decIter.pxPos, chan) + assert(arraysEq(pixels, decPixels, decoded.pxPos, decIter.pxPos)) + arraysEqCombinedLemma2(pixels, decPixels, 0, decoded.pxPos, decIter.pxPos) + check(arraysEq(pixels, decPixels, 0, decIter.pxPos)) + } else { + assert(decIter.pxPos == pxPos) + assert(samePixelsForall(pixels, pxPrev, decoded.pxPos, decIter.pxPos, chan)) + samePixelsForallArraysEq(pixels, decPixels, pxPrev, decoded.pxPos, decIter.pxPos, chan) + assert(arraysEq(pixels, decPixels, decoded.pxPos, decIter.pxPos)) + arraysEqCombinedLemma2(pixels, decPixels, 0, decoded.pxPos, decIter.pxPos) + check(arraysEq(pixels, decPixels, 0, decIter.pxPos)) + } + check(arraysEq(pixels, decPixels, 0, decIter.pxPos)) + } + check(arraysEq(pixels, decPixels, 0, decIter.pxPos)) + + GhostDecoded(freshCopy(decIndex), freshCopy(decPixels), decIter.inPos, decIter.pxPos) + }.ensuring { newDecoded => + newDecoded.pxPos == decoded.pxPos + chan * (run0 + bool2int(px == pxPrev)) &&& + newDecoded.pxPos % chan == 0 &&& + newDecoded.inPos == ru.outPos &&& + newDecoded.index == decoded.index &&& + decodeLoopEncodeProp(bytes, pxPrev, outPos0, ru.outPos, decoded, pxPrev, newDecoded) + } + + @ghost + @pure + def updateRunProp(bytes: Array[Byte], run0: Long, outPos0: Long, ru: RunUpdate)(using Ctx, LoopIter): Boolean = { + require(bytes.length == maxSize) + require(runInv(run0)) + require(outPosInv(outPos0)) + require(outPos0 < bytes.length - Padding) + val run = run0 + bool2int(px == pxPrev) + require(run > 0) + + val dummyIndex = Array.fill(64)(0) + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + decoder.doDecodeNext(dummyIndex, pxPrev, outPos0) match { + case (decoder.Run(r), inPos) => r + 1 == run && inPos == ru.outPos + case _ => false + } + } + + @ghost + @pure + @opaque + @inlineOnce + def updateRunPropAnyIndexLemma(bytes: Array[Byte], index: Array[Int], run0: Long, outPos0: Long, ru: RunUpdate)(using Ctx, LoopIter): Unit = { + require(bytes.length == maxSize) + require(runInv(run0)) + require(outPosInv(outPos0)) + require(outPos0 < bytes.length - Padding) + val run = run0 + bool2int(px == pxPrev) + require(run > 0) + require(updateRunProp(bytes, run0, outPos0, ru)) + require(index.length == 64) + val ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + val decodedNextRes = decoder.doDecodeNext(index, pxPrev, outPos0)(using ctx) + + { + () + }.ensuring(_ => decodedNextRes match { + case (decoder.Run(r), inPos) => r + 1 == run && inPos == ru.outPos + case _ => false + }) + } + + @ghost + @pure + def updateRunPropBytesEqLemma(bytes: Array[Byte], run0: Long, outPos0: Long, + ru: RunUpdate, bytes2: Array[Byte])(using Ctx, LoopIter): Unit = { + require(bytes.length == maxSize) + require(runInv(run0)) + require(outPosInv(outPos0)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(outPos0 < bytes.length - Padding) + val run = run0 + bool2int(px == pxPrev) + require(run > 0) + require(updateRunProp(bytes, run0, outPos0, ru)) + require(bytes.length == bytes2.length) + require(arraysEq(bytes, bytes2, 0, ru.outPos)) + require(outPos0 + 1 <= ru.outPos) + + val dctx1: decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + val dctx2: decoder.Ctx = decoder.Ctx(freshCopy(bytes2), w, h, chan) + val dummyIndex = Array.fill(64)(0) + val res1 = decoder.doDecodeNext(dummyIndex, pxPrev, outPos0)(using dctx1) + val res2 = decoder.doDecodeNext(dummyIndex, pxPrev, outPos0)(using dctx2) + + val b1Bytes1 = bytes(outPos0.toInt).toInt & 0xff + val b1Bytes2 = bytes2(outPos0.toInt).toInt & 0xff + arraysEqAtIndex(bytes, bytes2, 0, ru.outPos, outPos0) + assert(b1Bytes1 == b1Bytes2) + assert((b1Bytes1 & Mask2) == OpRun) + check(res1 == res2) + check((outPos0 < bytes2.length - Padding) because (bytes.length == bytes2.length)) + }.ensuring(_ => updateRunProp(bytes2, run0, outPos0, ru)) + + @ghost + @opaque + @inlineOnce + def decodeEncodeNoRunPass(oldIndex: Array[Int], index: Array[Int], bytes: Array[Byte], outPos1: Long, outPos2: Long, decoded: GhostDecoded)(using Ctx, LoopIter): GhostDecoded = { + require(oldIndex.length == 64) + require(index.length == 64) + require(bytes.length == maxSize) + require(outPosInv(outPos1)) + require(outPosInv(outPos2)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(0, outPos1, pxPos)) + require(outPos1 < outPos2) + require(decoded.index.length == 64) + require(oldIndex == decoded.index) + require(encodeNoRunProp(oldIndex, index, bytes, outPos1, outPos2)) + require(decoded.pixels.length == pixels.length) + require(decoded.pxPos == pxPos) + require(samePixels(pixels, px, pxPos, chan)) + require(arraysEq(decoded.pixels, pixels, 0, pxPos)) + require(px != pxPrev) + + assert(pxPos % chan == 0 && pixels.length % chan == 0) + given decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + val (decIndex, decPixels, decIter) = decoder.decodeLoopPure(decoded.index, decoded.pixels, pxPrev, outPos1, outPos2, decoded.pxPos) + check(decPixels.length == decoded.pixels.length) + check(decoded.pxPos < decIter.pxPos) + check(decIter.pxPos <= decoded.pixels.length) + check(decIter.pxPos % chan == 0) + check(decIter.inPos == outPos2) + assert(decPixels.length % chan == 0) + check(decIter.remainingRun == 0) + check(decIter.pxPos == decoded.pxPos + chan) + + assert(encodeNoRunProp(oldIndex, index, bytes, outPos1, outPos2)) + check(decoded.index.updated(colorPos(px), px) == decIndex) + assert(samePixels(decPixels, px, decoded.pxPos, chan)) + + assert(arraysEq(decoded.pixels, decPixels, 0, decoded.pxPos)) + arraysEqSymLemma(decoded.pixels, pixels, 0, pxPos) + arraysEqTransLemma(pixels, decoded.pixels, decPixels, 0, pxPos) + assert(arraysEq(pixels, decPixels, 0, pxPos)) + samePixelsSingleElementRange(pixels, px, pxPos, chan) + samePixelsSingleElementRange(decPixels, px, pxPos, chan) + samePixelsForallArraysEq(pixels, decPixels, px, pxPos, pxPos + chan, chan) + assert(arraysEq(pixels, decPixels, pxPos, pxPos + chan)) + arraysEqCombinedLemma2(pixels, decPixels, 0, pxPos, pxPos + chan) + assert(arraysEq(pixels, decPixels, 0, pxPos + chan)) + check(arraysEq(pixels, decPixels, 0, decIter.pxPos) because (decIter.pxPos == pxPos + chan)) + + GhostDecoded(freshCopy(decIndex), freshCopy(decPixels), decIter.inPos, decIter.pxPos) + }.ensuring { newDecoded => + newDecoded.pxPos % chan == 0 &&& + newDecoded.inPos == outPos2 &&& + newDecoded.pxPos == pxPos + chan &&& + newDecoded.index == decoded.index.updated(colorPos(px), px) &&& + newDecoded.pixels.length == pixels.length &&& + decodeLoopEncodeProp(bytes, pxPrev, outPos1, outPos2, decoded, px, newDecoded) + } + + @ghost + @pure + def encodeNoRunProp(oldIndex: Array[Int], index: Array[Int], bytes: Array[Byte], outPos1: Long, outPos2: Long)(using Ctx, LoopIter): Boolean = { + require(oldIndex.length == 64) + require(index.length == 64) + require(bytes.length == maxSize) + require(outPosInv(outPos1)) + require(outPosInv(outPos2)) + require(pxPosInv(pxPos)) + require(pxPos + chan <= pixels.length) + require(positionsIneqInv(0, outPos1, pxPos)) + require(outPos1 < outPos2) + + given dctx: decoder.Ctx = decoder.Ctx(freshCopy(bytes), w, h, chan) + withinBoundsLemma(0, outPos1, pxPos) + assert(0 <= pxPos && pxPos <= pixels.length && w * h * chan == pixels.length && pixels.length % chan == 0) + + decoder.doDecodeNext(oldIndex, pxPrev, outPos1) match { + case (decoder.DiffOrIndexOrColor(decodedPx), inPosRes) => + decodedPx == px && inPosRes == outPos2 && oldIndex.updated(colorPos(px), px) == index + case _ => false + } + } + + ///////////////////////////////////////////////////////////////////////////////////////////////////// + ///////////////////////////////////////////////////////////////////////////////////////////////////// + + @inline + @cCode.inline + def maxSize(using ctx: Ctx): Long = ctx.maxSize + + @inline + @cCode.inline + def pxEnd(using ctx: Ctx): Long = ctx.pxEnd + + @inline + @cCode.inline + def w(using ctx: Ctx): Long = ctx.w + + @inline + @cCode.inline + def h(using ctx: Ctx): Long = ctx.h + + @inline + @cCode.inline + def chan(using ctx: Ctx): Long = ctx.chan + + @inline + @cCode.inline + def pixels(using ctx: Ctx): Array[Byte] = ctx.pixels + + @inline + @cCode.inline + def px(using li: LoopIter): Int = li.px + + @inline + @cCode.inline + def pxPrev(using li: LoopIter): Int = li.pxPrev + + @inline + @cCode.inline + def pxPos(using li: LoopIter): Long = li.pxPos +} diff --git a/qoi/verify.sh b/qoi/verify.sh new file mode 100755 index 00000000..887a60f1 --- /dev/null +++ b/qoi/verify.sh @@ -0,0 +1,6 @@ +#!/bin/bash + +# kill subprocesses on exit or kill +trap '[ -n "$(jobs -pr)" ] && kill -9 $(jobs -pr)' SIGINT SIGTERM EXIT + +stainless-dotty verified/common.scala verified/encoder.scala verified/decoder.scala -J-Xms10G -J-Xss20M "$@" \ No newline at end of file