From c9352407e190b0931208653af74679e8cf9ac2ef Mon Sep 17 00:00:00 2001 From: Light Date: Tue, 29 Jun 2021 11:01:11 +0430 Subject: [PATCH] Textures - Added Texture and glTexture class ( no dx implementation yet ) - Added TextureRendererProgram to Renderer - Added LoadTexture to ResourceManager - Added TextureShader - Added awesomeface.png - Removed unused shader files from Sandbox/res - Fixed gladLoadGLLoader call in glGraphicsContext - Enabled alpha blending in glGraphicsContext - Added stb_image library --- Dependencies/stb_image/stb_image.h | 7871 +++++++++++++++++ Engine/premake5.lua | 1 + Engine/res/Shaders/TextureShader.h | 35 + .../src/Engine/Graphics/GraphicsContext.cpp | 3 +- Engine/src/Engine/Graphics/Renderer.cpp | 72 +- Engine/src/Engine/Graphics/Renderer.h | 43 + Engine/src/Engine/Graphics/Texture.cpp | 28 + Engine/src/Engine/Graphics/Texture.h | 25 + Engine/src/Engine/Utility/ResourceManager.cpp | 22 + Engine/src/Engine/Utility/ResourceManager.h | 7 + .../GraphicsAPI/OpenGL/glGraphicsContext.cpp | 7 +- .../Platform/GraphicsAPI/OpenGL/glTexture.cpp | 49 + .../Platform/GraphicsAPI/OpenGL/glTexture.h | 21 + Sandbox/res/Textures/awesomeface.png | Bin 0 -> 76691 bytes Sandbox/res/fragment.fragment | 18 - Sandbox/res/vertex.vertex | 29 - Sandbox/src/SandboxLayer.h | 12 +- 17 files changed, 8187 insertions(+), 56 deletions(-) create mode 100644 Dependencies/stb_image/stb_image.h create mode 100644 Engine/res/Shaders/TextureShader.h create mode 100644 Engine/src/Engine/Graphics/Texture.cpp create mode 100644 Engine/src/Engine/Graphics/Texture.h create mode 100644 Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.cpp create mode 100644 Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.h create mode 100644 Sandbox/res/Textures/awesomeface.png delete mode 100644 Sandbox/res/fragment.fragment delete mode 100644 Sandbox/res/vertex.vertex diff --git a/Dependencies/stb_image/stb_image.h b/Dependencies/stb_image/stb_image.h new file mode 100644 index 0000000..e80e7b9 --- /dev/null +++ b/Dependencies/stb_image/stb_image.h @@ -0,0 +1,7871 @@ +/* stb_image - v2.26 - 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: + + 2.26 (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 + 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 + 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 [reserved] + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + 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. +// +// =========================================================================== +// +// 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: +// +// By default we convert iphone-formatted PNGs back to RGB, even though +// they are internally encoded differently. You can disable this conversion +// by calling stbi_convert_iphone_png_to_rgb(0), in which case +// you will always just get the native iphone "format" through (which +// is BGR stored in RGB). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// 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_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) >> (32 - (y)))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__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 +// assume GCC or Clang on ARM targets +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#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); +#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) +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) +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; + +#ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s, x, y, comp, req_comp, ri); +#endif +#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 +#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(_MSC_VER) && 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(_MSC_VER) && 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(_MSC_VER) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode))) + return 0; + +#if _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); + return z + (stbi__get16le(s) << 16); +} +#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 = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB + k = stbi_lrot(j->code_buffer, n); + if (n < 0 || n >= (int)(sizeof(stbi__bmask) / sizeof(*stbi__bmask))) return 0; + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & ~sgn); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg* j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg* j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static 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) 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 == -1) 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 << j->succ_low); + } + else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short)(1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg* j, short data[64], stbi__huffman* hac, stbi__int16* fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c, r, s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS) - 1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short)((r >> 8) << shift); + } + else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code", "Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } + else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short)(stbi__extend_receive(j, s) << shift); + } + } + } while (k <= j->spec_end); + } + else { + // refinement scan for these AC coefficients + + short bit = (short)(1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short* p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit) == 0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } + else { + k = j->spec_start; + do { + int r, s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code", "Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } + else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } + else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short* p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit) == 0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + else { + if (r == 0) { + *p = (short)s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int)x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc)x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) * 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; + } + + // 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; + + // 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)); + 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)); + 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))); + 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) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[288]; + stbi__uint16 value[288]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16 - bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman* z, 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 >= sizeof(z->size)) 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[288] = +{ + 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, 288)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } + else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf* a, char* obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char* stbi_zlib_decode_malloc_guesssize(const char* buffer, int len, int initial_size, int* outlen) +{ + stbi__zbuf a; + char* p = (char*)stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc*)buffer; + a.zbuffer_end = (stbi_uc*)buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int)(a.zout - a.zout_start); + return a.zout_start; + } + else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char* stbi_zlib_decode_malloc(char const* buffer, int len, int* outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char* stbi_zlib_decode_malloc_guesssize_headerflag(const char* buffer, int len, int initial_size, int* outlen, int parse_header) +{ + stbi__zbuf a; + char* p = (char*)stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc*)buffer; + a.zbuffer_end = (stbi_uc*)buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int)(a.zout - a.zout_start); + return a.zout_start; + } + else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char* obuffer, int olen, char const* ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc*)ibuffer; + a.zbuffer_end = (stbi_uc*)ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int)(a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char* stbi_zlib_decode_noheader_malloc(char const* buffer, int len, int* outlen) +{ + stbi__zbuf a; + char* p = (char*)stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc*)buffer; + a.zbuffer_end = (stbi_uc*)buffer + len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int)(a.zout - a.zout_start); + return a.zout_start; + } + else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char* obuffer, int olen, const char* ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc*)ibuffer; + a.zbuffer_end = (stbi_uc*)ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int)(a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context* s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context* s) +{ + static 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); + 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 = 0; +static int stbi__de_iphone_flag = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag = flag_true_if_should_convert; +} + +static void stbi__de_iphone(stbi__png* z) +{ + stbi__context* s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc* p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i = 0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } + else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i = 0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + 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 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"); + 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) { + if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } + else { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } + } + else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + 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 { + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + stbi__get32le(s); // discard color space + for (i = 0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void*)1; +} + + +static 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) { + STBI_ASSERT(info.offset == s->callback_already_read + (int)(s->img_buffer - s->img_buffer_original)); + if (info.offset != s->callback_already_read + (s->img_buffer - s->buffer_start)) { + 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; y < height; ++y) { + int packet_idx; + + for (packet_idx = 0; packet_idx < num_packets; ++packet_idx) { + stbi__pic_packet* packet = &packets[packet_idx]; + stbi_uc* dest = result + y * width * 4; + + switch (packet->type) { + default: + return stbi__errpuc("bad format", "packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for (x = 0; x < width; ++x, dest += 4) + if (!stbi__readval(s, packet->channel, 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; i < count; ++i, dest += 4) + stbi__copyval(packet->channel, 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; i < count; ++i, dest += 4) + stbi__copyval(packet->channel, dest, value); + } + else { // Raw + ++count; + if (count > left) return stbi__errpuc("bad file", "scanline overrun"); + + for (i = 0; i < count; ++i, dest += 4) + if (!stbi__readval(s, packet->channel, 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); + 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 (!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(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; + 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 (NULL == tmp) { + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + return stbi__errpuc("outofmem", "Out of memory"); + } + else { + out = (stbi_uc*)tmp; + out_size = layers * stride; + } + + if (delays) { + *delays = (int*)STBI_REALLOC_SIZED(*delays, delays_size, sizeof(int) * layers); + delays_size = layers * sizeof(int); + } + } + else { + out = (stbi_uc*)stbi__malloc(layers * stride); + out_size = layers * stride; + if (delays) { + *delays = (int*)stbi__malloc(layers * sizeof(int)); + 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); + stbi__rewind(s); + if (p == NULL) + 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; + } + (void)stbi__get32be(s); + (void)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) +// Does not support 16-bit-per-channel + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context* s) +{ + char p, t; + p = (char)stbi__get8(s); + t = (char)stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind(s); + return 0; + } + return 1; +} + +static 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); + + if (!stbi__pnm_info(s, (int*)&s->img_x, (int*)&s->img_y, (int*)&s->img_n)) + 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__mad3sizes_valid(s->img_n, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc*)stbi__malloc_mad3(s->img_n, s->img_x, s->img_y, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + stbi__getn(s, out, s->img_n * s->img_x * s->img_y); + + if (req_comp && req_comp != s->img_n) { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context* s, char* c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char)stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r') + *c = (char)stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context* s, char* c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value * 10 + (*c - '0'); + *c = (char)stbi__get8(s); + } + + return value; +} + +static int stbi__pnm_info(stbi__context* s, int* x, int* y, int* comp) +{ + int maxv, 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 > 255) + return stbi__err("max value > 255", "PPM image not 8-bit"); + else + return 1; +} +#endif + +static int stbi__info_main(stbi__context* s, int* x, int* y, int* comp) +{ +#ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; +#endif + +#ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; +#endif + + // test tga last because it's a crappy test! +#ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; +#endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +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 + + 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/Engine/premake5.lua b/Engine/premake5.lua index cb01812..4bf5ffc 100644 --- a/Engine/premake5.lua +++ b/Engine/premake5.lua @@ -48,6 +48,7 @@ project "Engine" (dependenciesdir .. "GLAD/include"), (dependenciesdir .. "imgui/backends"), (dependenciesdir .. "imgui/"), + (dependenciesdir .. "stb_image/"), (dependenciesdir .. "glm/"), } diff --git a/Engine/res/Shaders/TextureShader.h b/Engine/res/Shaders/TextureShader.h new file mode 100644 index 0000000..ce2d1fc --- /dev/null +++ b/Engine/res/Shaders/TextureShader.h @@ -0,0 +1,35 @@ +#define LT_ENGINE_RESOURCES_TEXTURE_SHADER_VS \ +R"( ++GLSL +#version 440 core + +layout(location = 0) in vec3 a_Position; +layout(location = 1) in vec2 a_TexCoords; + +out vec2 texCoords; + +void main() +{ + gl_Position = vec4(a_Position, 1.0); + texCoords = a_TexCoords; +} +-GLSL +)" + +#define LT_ENGINE_RESOURCES_TEXTURE_SHADER_PS \ +R"( ++GLSL +#version 440 core + +in vec2 texCoords; + +uniform sampler2D u_Texture; + +out vec4 FragmentColor; + +void main() +{ + FragmentColor = texture(u_Texture, texCoords); +} +-GLSL +)" \ No newline at end of file diff --git a/Engine/src/Engine/Graphics/GraphicsContext.cpp b/Engine/src/Engine/Graphics/GraphicsContext.cpp index 1a9e6cb..b86f4f7 100644 --- a/Engine/src/Engine/Graphics/GraphicsContext.cpp +++ b/Engine/src/Engine/Graphics/GraphicsContext.cpp @@ -68,7 +68,8 @@ namespace Light { s_Context->m_UserInterface = std::unique_ptr(UserInterface::Create(windowHandle, s_Context->m_SharedContext)); s_Context->m_Renderer = std::unique_ptr(Renderer::Create(s_Context->m_RenderCommand, s_Context->m_SharedContext)); - // sanity check + // check + LT_ENGINE_ASSERT(s_Context->m_ResourceManager, "GraphicsContext::Create: failed to create ResourceManager"); LT_ENGINE_ASSERT(s_Context->m_RenderCommand, "GraphicsContext::Create: failed to create RenderCommand"); LT_ENGINE_ASSERT(s_Context->m_UserInterface, "GraphicsContext::Create: failed to create UserInterface"); LT_ENGINE_ASSERT(s_Context->m_Renderer, "GraphicsContext::Create: failed to create Renderer"); diff --git a/Engine/src/Engine/Graphics/Renderer.cpp b/Engine/src/Engine/Graphics/Renderer.cpp index 61e37be..fcf7faf 100644 --- a/Engine/src/Engine/Graphics/Renderer.cpp +++ b/Engine/src/Engine/Graphics/Renderer.cpp @@ -6,9 +6,12 @@ #include "Utility/ResourceManager.h" #include "../res/Shaders/QuadShader.h" +#include "../res/Shaders/TextureShader.h" #include "RenderCommand.h" +#include "Texture.h" + namespace Light { Renderer* Renderer::s_Context; @@ -19,13 +22,21 @@ namespace Light { s_Context = this; ResourceManager::CreateShader("QuadShader", LT_ENGINE_RESOURCES_QUAD_SHADER_VS, LT_ENGINE_RESOURCES_QUAD_SHADER_PS); + ResourceManager::CreateShader("TextureShader", LT_ENGINE_RESOURCES_TEXTURE_SHADER_VS, LT_ENGINE_RESOURCES_TEXTURE_SHADER_PS); - // QUADRENDERER // + //** quad renderer **// m_QuadRenderer.shader = ResourceManager::GetShader("QuadShader"); m_QuadRenderer.vertexBuffer = std::unique_ptr(VertexBuffer::Create(nullptr, sizeof(QuadRendererProgram::QuadVertexData), LT_MAX_QUAD * 4, m_SharedContext)); m_QuadRenderer.indexBuffer = std::unique_ptr(IndexBuffer::Create(nullptr, LT_MAX_QUAD * 6, m_SharedContext)); m_QuadRenderer.vertexLayout = std::unique_ptr(VertexLayout::Create(m_QuadRenderer.vertexBuffer.get(), m_QuadRenderer.shader.get(), { { "POSITION", VertexElementType::Float3 },{ "COLOR", VertexElementType::Float4 } }, m_SharedContext)); - // QUADRENDERER // + //** quad renderer **// + + //** texture rendererc **// + m_TextureRenderer.shader = ResourceManager::GetShader("TextureShader"); + m_TextureRenderer.vertexBuffer = std::unique_ptr(VertexBuffer::Create(nullptr, sizeof(TextureRendererProgram::TextureVertexData), LT_MAX_QUAD * 4, m_SharedContext)); + m_TextureRenderer.indexBuffer = std::unique_ptr(IndexBuffer::Create(nullptr, LT_MAX_QUAD * 6, m_SharedContext)); + m_TextureRenderer.vertexLayout = std::unique_ptr(VertexLayout::Create(m_TextureRenderer.vertexBuffer.get(), m_TextureRenderer.shader.get(), { { "POSITION", VertexElementType::Float3 },{ "TEXCOORDS", VertexElementType::Float2 } }, m_SharedContext)); + //** texture rendererc **// } Renderer* Renderer::Create(std::shared_ptr renderCommand, std::shared_ptr sharedContext) @@ -38,6 +49,11 @@ namespace Light { s_Context->DrawQuadImpl(position, size, tint); } + void Renderer::DrawQuad(const glm::vec3& position, const glm::vec2& size, std::shared_ptr texture) + { + s_Context->DrawQuadImpl(position, size, texture); + } + void Renderer::DrawQuadImpl(const glm::vec3& position, const glm::vec2& size, const glm::vec4& tint) { // check @@ -58,7 +74,7 @@ namespace Light { // TOP_LEFT m_QuadRenderer.mapCurrent[0].position = { xMin, yMin, position.z }; m_QuadRenderer.mapCurrent[0].tint = tint; - + // TOP_RIGHT m_QuadRenderer.mapCurrent[1].position = { xMax, yMin, position.z }; m_QuadRenderer.mapCurrent[1].tint = tint; @@ -76,9 +92,51 @@ namespace Light { m_QuadRenderer.quadCount++; } + void Renderer::DrawQuadImpl(const glm::vec3& position, const glm::vec2& size, std::shared_ptr texture) + { + // check + if (m_TextureRenderer.mapCurrent + 4 >= m_TextureRenderer.mapEnd) + { + LT_ENGINE_WARN("Render::DrawQuad(Texture): max calls per frame exceeded the limit: {}", LT_MAX_QUAD); + + EndScene(); + m_TextureRenderer.Map(); + } + + // #todo: implement a proper binding + texture->Bind(); + + // locals + const float xMin = position.x; + const float yMin = position.y; + const float xMax = position.x + size.x; + const float yMax = position.y + size.y; + + // TOP_LEFT + m_TextureRenderer.mapCurrent[0].position = { xMin, yMin, position.z }; + m_TextureRenderer.mapCurrent[0].uv = { 0, 0 }; + + // TOP_RIGHT + m_TextureRenderer.mapCurrent[1].position = { xMax, yMin, position.z }; + m_TextureRenderer.mapCurrent[1].uv = { 1, 0 }; + + // BOTTOM_RIGHT + m_TextureRenderer.mapCurrent[2].position = { xMax, yMax, position.z }; + m_TextureRenderer.mapCurrent[2].uv = { 1, 1 }; + + // BOTTOM_LEFT + m_TextureRenderer.mapCurrent[3].position = { xMin, yMax, position.z }; + m_TextureRenderer.mapCurrent[3].uv = { 0, 1 }; + + // advance + m_TextureRenderer.mapCurrent += 4; + m_TextureRenderer.quadCount++; + } + void Renderer::BeginScene() { m_QuadRenderer.Map(); + m_TextureRenderer.Map(); } void Renderer::EndScene() @@ -90,6 +148,14 @@ namespace Light { m_RenderCommand->DrawIndexed(m_QuadRenderer.quadCount * 6); m_QuadRenderer.quadCount = 0; } + + if (m_TextureRenderer.quadCount) + { + m_TextureRenderer.Bind(); + + m_RenderCommand->DrawIndexed(m_TextureRenderer.quadCount * 6); + m_TextureRenderer.quadCount = 0; + } } } \ No newline at end of file diff --git a/Engine/src/Engine/Graphics/Renderer.h b/Engine/src/Engine/Graphics/Renderer.h index 75a2c39..861197b 100644 --- a/Engine/src/Engine/Graphics/Renderer.h +++ b/Engine/src/Engine/Graphics/Renderer.h @@ -17,6 +17,8 @@ namespace Light { class RenderCommand; class SharedContext; + class Texture; + struct RendererProgram { virtual void Map() = 0; @@ -62,17 +64,57 @@ namespace Light { } }; + struct TextureRendererProgram : public RendererProgram + { + // graphics context + std::shared_ptr shader; + std::shared_ptr vertexBuffer; + std::shared_ptr indexBuffer; + std::shared_ptr vertexLayout; + + // buffer data + struct TextureVertexData + { + glm::vec3 position; + glm::vec2 uv; + }; + + TextureVertexData* mapCurrent = nullptr; + TextureVertexData* mapEnd = nullptr; + unsigned int quadCount = 0u; + + // functions + void Map() + { + mapCurrent = (TextureVertexData*)vertexBuffer->Map(); + mapEnd = mapCurrent + LT_MAX_QUAD * 4; + } + + void Bind() + { + vertexBuffer->UnMap(); + shader->Bind(); + vertexLayout->Bind(); + vertexBuffer->Bind(); + indexBuffer->Bind(); + } + }; + static Renderer* s_Context; QuadRendererProgram m_QuadRenderer; + TextureRendererProgram m_TextureRenderer; std::shared_ptr m_RenderCommand; std::shared_ptr m_SharedContext; + + std::shared_ptr m_TestTexture; public: static Renderer* Create(std::shared_ptr renderCommand, std::shared_ptr sharedContext); static void DrawQuad(const glm::vec3& position, const glm::vec2& size, const glm::vec4& tint); + static void DrawQuad(const glm::vec3& position, const glm::vec2& size, std::shared_ptr texture); void BeginScene(); void EndScene(); @@ -81,6 +123,7 @@ namespace Light { Renderer(std::shared_ptr renderCommand, std::shared_ptr sharedContext); void DrawQuadImpl(const glm::vec3& position, const glm::vec2& size, const glm::vec4& tint); + void DrawQuadImpl(const glm::vec3& position, const glm::vec2& size, std::shared_ptr texture); }; } \ No newline at end of file diff --git a/Engine/src/Engine/Graphics/Texture.cpp b/Engine/src/Engine/Graphics/Texture.cpp new file mode 100644 index 0000000..abefe3c --- /dev/null +++ b/Engine/src/Engine/Graphics/Texture.cpp @@ -0,0 +1,28 @@ +#include "ltpch.h" +#include "Texture.h" + +#include "GraphicsContext.h" + +#include "OpenGL/glTexture.h" +#ifdef LIGHT_PLATFORM_WINDOWS + // #todo: +#endif + +namespace Light { + + Texture* Texture::Create(unsigned int width, unsigned int height, unsigned int components, unsigned char* pixels, std::shared_ptr sharedContext) + { + switch (GraphicsContext::GetGraphicsAPI()) + { + case GraphicsAPI::OpenGL: + return new glTexture(width, height, components, pixels); + + case GraphicsAPI::DirectX: LT_WIN() + + default: + LT_ENGINE_ASSERT(false, "Texture::Create: invalid/unsupported GraphicsAPI {}", GraphicsContext::GetGraphicsAPI()); + return nullptr; + } + } + +} \ No newline at end of file diff --git a/Engine/src/Engine/Graphics/Texture.h b/Engine/src/Engine/Graphics/Texture.h new file mode 100644 index 0000000..5d8d3d6 --- /dev/null +++ b/Engine/src/Engine/Graphics/Texture.h @@ -0,0 +1,25 @@ +#pragma once + +#include "Base.h" + +namespace Light { + + class SharedContext; + + class Texture + { + public: + static Texture* Create(unsigned int width, unsigned int height, unsigned int components, unsigned char* pixels, std::shared_ptr sharedContext); + + Texture(const Texture&) = delete; + Texture& operator=(const Texture&) = delete; + + virtual ~Texture() = default; + + virtual void Bind(unsigned int slot = 0) = 0; + + protected: + Texture() = default; + }; + +} \ No newline at end of file diff --git a/Engine/src/Engine/Utility/ResourceManager.cpp b/Engine/src/Engine/Utility/ResourceManager.cpp index ef02559..bdde555 100644 --- a/Engine/src/Engine/Utility/ResourceManager.cpp +++ b/Engine/src/Engine/Utility/ResourceManager.cpp @@ -2,7 +2,9 @@ #include "ResourceManager.h" #include "Graphics/GraphicsContext.h" + #include "Graphics/Shader.h" +#include "Graphics/Texture.h" #define STB_IMAGE_IMPLEMENTATION #include g @@ -21,6 +23,8 @@ namespace Light { { LT_ENGINE_ASSERT(!s_Context, "ResourceManager::ResourceManager: an instance of 'resource manager' already exists, do not construct this class"); s_Context = this; + + stbi_set_flip_vertically_on_load(true); } void ResourceManager::CreateShaderImpl(const std::string& name, const std::string& vertexSource, const std::string& pixelSource) @@ -86,6 +90,24 @@ namespace Light { m_Shaders[name] = std::shared_ptr(Shader::Create(vertexSource, pixelSource, m_SharedContext)); } + void ResourceManager::LoadTextureImpl(const std::string& name, const std::string& path, int desiredComponents) + { + // load image + int width, height, components; + unsigned char* pixels = stbi_load(path.c_str(), &width, &height, &components, desiredComponents); + + // check + LT_ENGINE_ASSERT(pixels, "ResourceManager::LoadTexture: failed to load texture <{}>, path: {}", name, path); + if (components != desiredComponents) + { + LT_ENGINE_WARN("ResourceManager::LoadTexture: image file compoenents != desired components ({} - {})", components, desiredComponents); + LT_ENGINE_WARN("ResourceManager::LoadTexture: <{}> path: {}", name, path); + } + + // create texture + m_Textures[name] = std::shared_ptr(Texture::Create(width, height, components, pixels, m_SharedContext)); + } + void ResourceManager::ExtractShaderSource(std::string& src, const std::string& delim) { size_t begDelimPos, endDelimPos; diff --git a/Engine/src/Engine/Utility/ResourceManager.h b/Engine/src/Engine/Utility/ResourceManager.h index 9591eb0..d306dbe 100644 --- a/Engine/src/Engine/Utility/ResourceManager.h +++ b/Engine/src/Engine/Utility/ResourceManager.h @@ -7,12 +7,15 @@ namespace Light { class Shader; + class Texture; + class SharedContext; class ResourceManager { private: std::unordered_map> m_Shaders; + std::unordered_map> m_Textures; std::shared_ptr m_SharedContext; @@ -25,7 +28,10 @@ namespace Light { static inline void CreateShader(const std::string& name, const std::string& vertexSource, const std::string& pixelSource) { s_Context->CreateShaderImpl(name, vertexSource, pixelSource); } static inline void LoadShader(const std::string& name, const std::string& vertexPath, const std::string& pixelPath) { s_Context->LoadShaderImpl(name, vertexPath, pixelPath); } + static inline void LoadTexture(const std::string& name, const std::string& path, int desiredComponents = 4) { s_Context->LoadTextureImpl(name, path, desiredComponents); } + static inline std::shared_ptr GetShader(const std::string& name) { return s_Context->m_Shaders[name]; } + static inline std::shared_ptr GetTexture(const std::string& name) { return s_Context->m_Textures[name]; } private: ResourceManager(std::shared_ptr sharedContext); @@ -33,6 +39,7 @@ namespace Light { void CreateShaderImpl(const std::string& name, const std::string& vertexSource, const std::string& pixelSource); void LoadShaderImpl(const std::string& name, const std::string& vertexPath, const std::string& pixelPath); + void LoadTextureImpl(const std::string& name, const std::string& path, int desiredComponents); private: void ExtractShaderSource(std::string& src, const std::string& delim); }; diff --git a/Engine/src/Platform/GraphicsAPI/OpenGL/glGraphicsContext.cpp b/Engine/src/Platform/GraphicsAPI/OpenGL/glGraphicsContext.cpp index 39de6ef..a99ad46 100644 --- a/Engine/src/Platform/GraphicsAPI/OpenGL/glGraphicsContext.cpp +++ b/Engine/src/Platform/GraphicsAPI/OpenGL/glGraphicsContext.cpp @@ -25,10 +25,11 @@ namespace Light { glfwMakeContextCurrent(windowHandle); - if(!gladLoadGLLoader((GLADloadproc)glfwGetProcAddress)) - { exit(1); } + LT_ENGINE_ASSERT(gladLoadGLLoader((GLADloadproc)glfwGetProcAddress), "glGraphicsContext::glGraphicsContext: failed to initialize opengl (glad)"); - + // #todo: add blender + glEnable(GL_BLEND); + glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); } void glGraphicsContext::OnWindowResize(const WindowResizedEvent& event) diff --git a/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.cpp b/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.cpp new file mode 100644 index 0000000..ebf11f5 --- /dev/null +++ b/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.cpp @@ -0,0 +1,49 @@ +#include "ltpch.h" +#include "glTexture.h" + +#include + +namespace Light { + + glTexture::glTexture(unsigned int width, unsigned int height, unsigned int components, unsigned char* pixels) + { + // create texture + glCreateTextures(GL_TEXTURE_2D, 1, &m_TextureID); + + // texture parameters + glTextureParameteri(m_TextureID, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR); + glTextureParameteri(m_TextureID, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glTextureParameteri(m_TextureID, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTextureParameteri(m_TextureID, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + + // vocab = 0 + unsigned int format = components == 4u ? GL_RGBA : + components == 3u ? GL_RGB : + components == 2u ? GL_RG : + components == 1u ? GL_RED : NULL; + + unsigned int internalFormat = format == GL_RGBA ? GL_RGBA8 : + format == GL_RGB ? GL_RGB8 : + format == GL_RG ? GL_RG8 : + format == GL_RED ? GL_R8 : NULL; + + // check + LT_ENGINE_ASSERT(format, "glTexture::glTexture: invalid number of components: {}", components); + + Bind(); + glTexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format, GL_UNSIGNED_BYTE, pixels); + glGenerateMipmap(GL_TEXTURE_2D); + } + + glTexture::~glTexture() + { + glDeleteTextures(1, &m_TextureID); + } + + void glTexture::Bind(unsigned int slot/*= 0*/) + { + glActiveTexture(GL_TEXTURE0 + slot); + glBindTexture(GL_TEXTURE_2D, m_TextureID); + } + +} \ No newline at end of file diff --git a/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.h b/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.h new file mode 100644 index 0000000..90c803a --- /dev/null +++ b/Engine/src/Platform/GraphicsAPI/OpenGL/glTexture.h @@ -0,0 +1,21 @@ +#pragma once + +#include "Base.h" +#include "Graphics/Texture.h" + +namespace Light { + + class glTexture : public Texture + { + private: + unsigned int m_TextureID; + unsigned int m_ActiveTexture; + + public: + glTexture(unsigned int width, unsigned int height, unsigned int components, unsigned char* pixels); + ~glTexture(); + + void Bind(unsigned int slot = 0) override; + }; + +} \ No newline at end of file diff --git a/Sandbox/res/Textures/awesomeface.png b/Sandbox/res/Textures/awesomeface.png new file mode 100644 index 0000000000000000000000000000000000000000..58daba04f1f35d41821194d972b510cbbb86278d GIT binary patch literal 76691 zcmX_obzGFq_x4SwpeX37h=3rmfRwaIN=hs!2uMqJckZgFEGpegr-UG#3#dp+OGrz1 zE*<-ueZHUf{X-GibI+NXGiR=Io$ChuOil6X<-3<52)e5DR89keNWnizA+n3$3zDAm zE%-v@rlBYU6?M}s6Mkg%1n~rdO2V$3np^i$}l>X+FBW==*$n3FPZkK9$*p#2}g=Y!3Y@@MBV)Z6WS1V)lB?XTLcP+ekbi1-}fW_W0K9dhh9B;8fDE|LC7 z`gl^r_*F(P_^{GjVhCc)=T{XJ?3=Ud8wujs5I`{ctUpYzyReWLU#4y?In}HNgCM!f z;Hjui_tgMRm>>223D1tlScC8=Nr(#VO&;NrD#mY4F0U%dB|^Zb+}MHF25C;a9_dpZ za;pK77|g4Pu(V~qH+k@!_1&Yl=&x|FBHUndsD~!$lyU3D4{T%$5^hw?7Dlu}CHh{iL@cRK(*=~_Qb?nm`=Tc6fZ%L zybKWpIWvxaDt?P@fFCFH%1h}Vz24F_lle%*tm(;OLs~NmafUj20n`Eeb>-1 z&tRRAK@K_g*S+9bTgo^CR$sb&@L{^s{vJ1e z8*ZBjp#p3uU-KfCHc(**EUR`z?CYQ2+8*cAw(lLsD;M>)vU;Z|eM@55KX zP>hBA>8+&i6ApWWCLe)^5GL;|D+E)-^opDvC)g1~H*7HE4)DVivP=n zuiUf{Y*gx1962QY21Ev`dDN>9S&LqQLyXEp3B4EJ=uz~Z6Nw=p=mrWzIw@~wa|GPI z5BUZy4uireaLP}xdd^N$P$C(zG>Wir56f+)O5pDkAGHKc4pTzys?>Ofc<-M<8&o5< z#LzS|@O7GDViH-6-c?=l2$)xepg@~CbdTz3PqP^O5r`YfG!R+a*%K!fW{r-UQUh{Y zU~e<@!HSVt_dpZ(pS>c6s4jshXOz^#^vdIF;CGivp|n4-@x4YwdK8CxY!Jjs2nFS$ z6!M8t$@eKG{yqy&q~p0WeSA@J|uzQCOEnV;$>{_SY=Fv1?J zDLXOn^O;~W!oYHd3nsWrf^zO%c>PmV+;YYZrbh>Mjo2&G z;Pw~v5%}hi1K4kACcxDmrt2KgJ1}*TLQVAV;0+JAlgK5Q!kvhrSKt7l?l{ck>Ywhd zG+YKo&M8J5U+Rc)!A&g55XxME2$JR>JE=6LJ}2=2Q`VhTwZ1>?16#Wcg-3&>eyZS` zV%)Q5Y@ZRL2I9WuBCI&Ex+lnpo9w@=gffpXHASKA5Oj8LQjt*jIIac4ib{K3sCO_0 z+Z`-;4@D$rh=7guZYsm}+=(Gns;=V<5`Is(5h>8`gzns#0Gz?+jlu@S!QMGp~#}xj4fmxs0r@v6@;P<)aJM)%rPu=i#w_~ov^Ie9C8|8?ppSq>M>j>zCc&f-z)&% z5Ghdw-f02vstd8;kRprFAVO|rEB{i>(`MDo*Y1c+94yV!o9D{Y;28K+*7Zp9fmYim zZd_M}p;YGr%Rzhe%h#{s>__N0$qL85$!dR{xZkPc5GAzgUO(}vs&mC-%wjN+zSQ#U z+6j`J<+mEFIR3B-tpxVFC>^9VA@^Oyw2p*#zSvPwjVjB~KY1oU{D+DSwexk3@b3b( z%9@6&_O`+{?!3hAWkr0j+a(wz=HFpVjjPK7XO@g>@0_+F>d&sqzte4s;FTd_W8Wm~Z(^VnD@rPopf z2{##+UM%lN1XsoE(Z{Neq<0vLUn}eT#iTQ)q2v?Sw+hM6=ZUF&!9oQY`oL1#n_aI`0_dR-R3o} zb}tw{fY@Ja!Eu+LOx*fr1Y+$Ih&8f;6?mY9Ox3YYFZuKLZ89VBq-u zP^db(XBqMV*a~J>1%aTP~48K}71x(2tBv^r~u-#mqEyEVIp0&1vY@2g)CY;J>XfmOhhz zPx<}*?vw7XUEdsOiq@r_|B$z8@uPfia`6qT{<<+y@TAw#Yjk76p#qp{gm(~exV2*Z z;al`iEQONN=GWnO*oOAaKRFrhj`~Wq9$&b$I4wpSiJe<>xa-V!et+h7+<8nRWs$6! zT+u#9Yre>1c!dca;5btZfA2>Fu@LeQx__*B9gb6$V!1w%Ggbe*sH#b-Fu-E-DzfhO zEfzM+XNF*&_4^NPS}q=$jQSoq&gdO^{?IG#L2WQe_xEv<;%!Xhyds9z>~xfTJ@Qtte>eL|N?#PUVFIL2E?GaHUftC0Tyn9S!!QiBt$2v76*G~4 z65ghS-6V!c--A?=w)6vk4R(g5h=@*!Wun*D!A)*>38km{wo0Nh`e@1Ay_#OarcS?U z!_}r3>pb1u8E0DJs|QHmK~O^`I2BGLh^2&9N2B*ol~lNy*Y&ldUtGp4@k9QyGj?2^BfO|rj`XN3+$ zZ_n4Gy7l58E>nqzSTt`v1vG=oeFSa0#-pVYA9Z(5v#j?P8y$9 z3YusTBgM5f1v}{s$TkI}CRG1wJ}|lnWlMuwz^3A1y5X5R-}j!ImxS_l)GqFaa36*( z=^BDe^sSM&$wPm~t8>dqyWKC9H%7hD=>9IXQ$-_j>(X=70b(f>)ee#k>-V3JQIYv} zGuklfgr@`K)%UCH6Y2iey{8a#F;BhQUwzN;__cX@6@~SEX)iK=OUVjpD0j&-R77>$ z-n2mtl6C?2)iVPv)zhKo3c`{4H~h^}*QpOVe!T*#Ub#ooJ4@_4pSE$$dvPF}-)sYZ z!s*k^fcBjx zGq`sqsXOc6Kv5^eq{3~NNPpznt0ht_JvR)cf2N`) zURwF0_{YVo5Nny^Pqia|oU#x8V>bWVu0?5JW`nar@K?TV9T;Wew8145Nd~|luW&rb zfU@51PO8U3`IhZUih~hRVMUy0bvm;ll8{TT7*Tn|G=_mYqDo4aiF4S#s__j~qKQn{ zh57{xXiGqxgLouW8-8wK`YrDMi`pVTSwsdp>N){q~D75;OFKurm(!nubr)*VvM*@uNgFGA6Fs;A%Oiqsgl6NTGi z1VW41WG@azT>JW9R2L%4WymRt6Oq#6;-%DY?nXKPwVi2%pIg3pa5a1m1av!s_SEpz z%r)5YJ_EUSjcE>rL~Ur2pR4R7j>I{3gTi~`>(Nf52zA?z*4P@?4SO^>H%`jRG)xV8 zDk-i<++I+IzNCuV;%8Acay$+jI}shAWK`bYrvNjS-#Hz-VzGWqzpoQ6J*&D@)W~Oa zkuh&8nw%xI170q>H3)aV1l?c+H+<-d<^b36YDaJ8sj-n40z zfhYFV;R1LHR7(P2&nUZd3>f~}O$ap?VL8JDLyDx!L!_!zy9eufh;D|_XKIqQq0zi? zjv$gEJ6pQy=01nr?kp?ri7Re2=Jp#2PhqE3J^kDK7Vf!9in0c2hUG&b+`SH|&nuf| zJKPrQre@1XojUn4@ei++6%k~+z2on*4g48VaWg+RUDx?Q8gDW)442&Fhav!Q4;FS< zJD`Lu#!vV7q1->>$w}<|x=tSf(EB0hjJV-WnQ{u!PEwKd#-apFC*kFs1;q4$O@lGf z-RUR>US20Pe0<@H$AGw~RdB<)6N}aZOR#U!E=$*6>rCpi^fXp9i7}z1=P`$IXfZ@@ z#r5Djmmz5!oym*|41uL+D0p#Swp*$u-N5|znH~Vl%{_Aov~zhni#|G+i>9gvx|tvW zgn@}^bsSMJDX?^Z!Jj^Otb*9G5`&V*UPP+&*smpT=6;ld4pp44dmof#?;NVQ&6Pdp zT<0R*N+g%K-K0&yNc@ro+8B^(T}-SVle?9SLYDGt&X=35c2rOc$Ay07B)A;jqOQ9s zboI}|YyBM?BL>tSJkV;!1vdT{dh`h(e&LCQ;;}Q2F)9B);AS##R%WijeQKGEMMBau z4x^=R83458U1d>j94l`*%)1H5(P*@!~#?OH%NPSS#UAx6YeIBue)K* zOcK0*@ms#kf{)X0VZyu`&)Y8e1||(qJ%r^BsGP@DX?Dd^5rFy#kX2; zN5O3=UJ@!j@fG5K+7m%MryqQ2X5b)AC;7+(9CgWeWy{a*5aw;+M(q&6W|YDi*6>J2 z*nZT;>bl})LNENa3YJ283FPS?DpSX+y7&&G_tjtY1r!6_T-<67%M~;l3~G}vKrB%J zEw|6|_Qaf|q;RO{^s+H_%7p`$JB9{5G!iW9DbLBmZ_ac_J@#ziHqh)wMmuQrv8W z(R+>6iR}DwcX<_2M`>;`5~wNN=4Vsgy#$`kfRY#c-^@S9g&~$4QSu4MByD89k~rhOVPy8Ggn`aX$ke;;nY*Q>~%GwjQ4P#emNnQu24ec zfP#2Rx4uy58=L zo9-e_^6*$6n*ZpO8?wcF17+u0^%Sg;0PR?%JX3c(@u0OaC{6v@&YfZocVwvsYXkvoAjRjdfTLum}Ops3>y`#Yh-5OD4q zxnM|PmrV4{Wa;g1TbH1q!AY4WGv2NkhcOP)=x|S6|E3@#nrc}#=n!yN)5VqW6oJ0U z&SzWz+Jgh?+8OgTOMA~=>i$ppniXU}|Ctiphe*s%exU*>i&4dC=%76i6 z8F~znwMrVIR=i8=>rQ_(_IW*9DWLn11R5`KUe>PJO}E*ZE7;9Kt&*si=A!ix;|_0> zFH>28WPUiXdX^%@*9Jx=@zaW{b6jSm*V8H&00o&9xCLEL0;1;kD5*Q|tw5~X@6c`t zXG2CRMS|hu{vhGwf`1n5r76kzZBtIRya9FE6$HagfJ-|V&Ly5V(3N#`r`^HTHeEr~ z1BaDAgF>32whK=QcNBP%*=TA#R{v&reDlwDk97(N>f8PMCC0T=imjuXV|9%p#MFU&q&$XJL>_-+ZPmz_R+;*>y# zhH=jW7F}$p{$|LN0=P~>FKsLo4FGNhC zuoUmndUY6vsm0;Ng)}|avi=3|Srbhw!}LLfMS$1|DgH=4u`@N&>HM^n;ko`s&DTu9 zqE-cRfR7#fO8t*!C8FkPs%|@*N%wW}ySzs;5bVHN4pFQ+l`>@->vVqhp`w5Ve=e)I z>~<&lj0k!ZT>=~KaKZ~eX*f5+Ol6YcPMf>n8@v3_2RT5V+%??74>!vjy?>jpFVJeY z<<+fp)f$ALG~d!S)UcERSJk-C^963x4}RAroMtY=^sy98MBCQa==9uR#Te^Zi$`gt zqBc7p3iLZY1VId?y{Oz|53fs&=kSsg)Y_@KHKI$`OaaS{k6*EX_BjUwiANgm!zTrv z_jEc-`*tHoq5|fti7HsQV@8=EC_5rHmI*{g%=U+i$E;=UUfhd(Weq8rS8pA2}UZQ)almh1ZZti_J%Wg_{lc!0dt z0FsjeSn^cE5YR458x|wH*pAHwPQ4BfYLU24?v*Db&~(+r1;4rUMa$0AjbtuDWl{DB zwTVcwHGLgNv;`J8aD@!UoA_nH;_!mQiq}T?#hsohkMI?((lstn5Qp3&KU_uTc98UR z$NrGnxGEkj1joH>QZ%}Vx(A`Me|?j}c=waW4Khd%@5ju4{@~ZCwI~$jD(b6DAnLf; z#1NqyI+_4uX%dR8PHl&>#y6?$d@ix%>lud?YEV1Rq!{GR;4BYYhPxX_-qXl%;P|0N zecF@w=x^uN6`z`A*Aowi`FuBw@`I}&^OIf03U3`%QMY+`eTws-eY6uu^;0#;%rq*F zNa?|(`p#&8W|i7%!S@Nc_89adAC0a;l<)JOzVSRjZiRMh3z29lT2t*UQqu?rEzb!dU~iHcNu{0kmYF5rOW2yCQBT#yb#`!MOdG z_X2$TGQAFs-JS^e?0ui?ueC`5`5EG|9%ag?!yOvDo9O$}WWNRNBHnxPy%4l-g?sZE?0QQ zRtk`V*&p%e69w($ibMN}q&&JQPRI+fl5-S-edEm^;MJPfm6#_lpgiI3jo)rVIPO(F zSd=Pcl6G*i>d8BC92T&Ye10%K`uyIJK6Rd9GU7U5&R!c0`~Bm%nw@w=w|92;Z;7VJ zqA7!eDxKAg8EmwkF1Q}}sFyHOw`V+czB+-?L_lcare-U6`7*;kUp1232fy6FpJwI3 zkr{VN0-@S^H}`*zYQkuy%ZW zF-)4+&~%0srkC_ai;q|Ke79B2_24Ugw!bL01_u1oswuT`J$Bj4(i#w_k+h7Nb6U*V z2#6L+*CAcmu;34*xjHb$8t0De_eSWXFk8gdY59Ps(&RFLQJxTR?cxx{6N z^lK8DS~Y3fV(Udg5-a8Sq^=r$;W`zdqxt-{JRm@54{i8)gB{42*nDgqyaQa96$0PNBaXJ=w|!`+uHS^H zgeFw$9b{;0o4!03QaL|~<8j?dd+d1;RF%1tc-Mr}70Rj3$Xf<9)szz5{u$)Q#Ekrp zs6Vrd{tDtb=k%qSD}=3d`l;_lseL`-2D#em){fZ1s-;yv%9><9v-+k((_vWkjzY!n}8rh^&4$k?7iD9(MO-M9@4r3V0?gJl&d z{%Z^uk+OaQZQlH2-XgWu%? zJsbDL3clU#g9_f!cMU)g#u7b|uBd^q0yZmTKoOfahu+{xjH$Yp#`O7m&Sx^%9*1hqb^D4j=kVL%{ zfjJ;31ioX@WUPr0Qz0%1kAa8cN?Ee6oaf>v5El=I^(nz}|5ybjAa<2Fu6WeaDl&Fb z+am*wFuE8-b^!{Ca7^``lJnwK(NM${^5E-o>;4qRIuMrYhr{6Ly6bnR#+Wb#t||eg zRX8H-5kcs9%&}81|-_FSJ$#wiN?O10fQcTsglhva~7Q}Eoxa(aP$oE(! zB(xgt);&lFBVF3>nnl!`y-5a_jwi&ZdN-lBh2t}pi+nj2Prb3q0f1r^z(#a+IACGmSNHZbSH)}JH7?u1+uujAIMwwETgrxVSmzl`QSWH2Z0o&ir z!g_|08f_-!CAksI43Uwn#~&x4v+AcDo#OC|8|(LdYG^>I1vLfuvP>7kXjx9UiAKkd zDPYE!mw-*r_(lQ}>`1d6+_j#o9gq$@>XU<7@9!B76Jkzo;9F0e?^WVqr*RvrNVNuZ zI$UyxpOJ_i@W16Zj-_Bqf0v8QRzr4H-eR*{B)Bhwq$4-=BU^tSSD@HAg0@>%?eAM@ zt$`cu@iV@nBa+SlZ72?yjZv(5w^+^V`lF`|dD^0#{|#mQZimLTJN#lPE3(R8>^{dw zsP_Ev-{u=IsU=b>mdl_Y--!^nXZqC{`Bd97U)i}>vg_Uq@9Db}*Y!_Id zQXgQx>RcF#yH!xtd6>5fjt0#unVu(~w)mSsMLyf|q0M0=$!app4U z^bnSJ`q|^EuPh%93V9f=881xS3^(`1wDCQ;4mG4hkknaM^VbrSo!*l(DZhWt%f0y- z1h^>ue6S#f*^18CSRtTMA+~7T(ke=U7ght8+~#NO27^)7PMrn9Msd7(Ndcw-J|`uT z8rh~f|CzePfKXbbuwewrGKQn9{CS|pvW?6XOTe0iG~IKs-sOL%JW5Pl*){07Dh_658h6zBj+aepR`3I*OWrtl?pq8ol;ojgsS0p#Romw_4O=$iV}Ios#V zlfmeN)D^bD0`tEB@j-_O3`t6W21x?HTjuY?#=qV0cRvwp{CxeLs2dp&s=wb7>DBta z{a>j7ZKvD8es~hs`5`bPIL`OrKgE`0Uu?r>tYi14R| z@<04L00GSPG}>@-aJU2T4n4gYaGQ3d;QfodXpYoR1`(b3o!x5aNuHD^cVDFNxo9xR<2igQKy_YxY^o0*8F?+nD$?nDq0KFq&T; zrDz=li~RN%0v7%!aF(t6(|`b1P!AL9tUppTybAhl6fW3BR59vOiPov7AoQ@=uT3}# z{+7#$lOH!(f6)Go?L@hv!~Ga&FPy)32#w8Ge{6_!(H#1l&^CY}QWM-xk0bg3Iu7!K zKbv#m9lpBF^Z+%~pF2CXb7L$CiGKPGXl~4ZvQy|hG3F@~h=_AC4M!JaDISsef)4R$ zLhmo65AVlnfGgPv7gA}>0wp+pqP~df<<`MnPm)iV(#dWGy;?#_6fTirLy?%gdn;>W zI&%f4h2@%YY{<~hN(YpKXh{Ncg>375{)?S|~DlLo1vC{q!-#c4t=RrKe za2a|OKMH@3`7 J;(Uk!Q_zAxwD5IVNsyC)`qGm{3A$}2F>#Z9!Z=Ty>N4XyhS^xyj0t#kwxa>Wum1#caiIw2p zX%KJ;LTA;$uFK03K~lMkA8>ag&W&!Hfa+>e2#Kv@k@q8PVb;Om$e`;?wsCt)m#?4E z-0~Hy8Db5jwKSkT$&T+`m&pZVEl->?NDin!t}L=9>_s>QU|xA!4IUb37psJLY!|@0 z1)ncjJoUvvO${4DF3;`n81C85w^8$64Hxdh1p_`4id5>d{p)&qy<&c%C$=MpJ7S7F zjlQTe;jppMUH?0LdV zamR0WRmC?gp6cC%dV88%;Y`>EL%;B?_*LGR2j(M$(<5#EGbh>T`ZPdo>#F9<+8@5d z%S9O@Z(cyn_XiOn z1|J0jU0u(bldO<8^P7Q*kfP#EPllGtijyGIF*%Uw!4;$IPIf z@TjkoX(j$#XueDF_^Ib@yf27|H1GZajk)sKm3F{Zm@m|Kc?w4Y6h!~s$&V(Im@{iG zomAWP*m1P|8CC)s{trUP(C>zE}{3pVuOt=GQO-&T~knd;n3tri*=b+<~ zjHtI`Dz)-=AH#FvEw;gdMja|4IGANair62LkApcd} zTmL!3032%QGt4QL_@e^Qg@qMUUJf~yTXmQ!m|YvumyNeKeh9cqrY}7vT42?D2eyDy zP{Wu$>7Cxp+GKQADBOteJBg>P5n%iTXK-)oBdW6~c^} zuf*C{ltm2uk*g$MCr1URLsSTA%%LAJka(;WL7BSFTVtUM&?Ch7a(S7CN+wY8W}Np6A9h8Act@9-9O^Sj8}lg0IsGzBOo95%_#Il8kO z^B6XGyP$>-3Hv}+haM1V`)3t+pH_IiTp&W@)r>cc!IvgUj;P7OS8?$r>{dOOOlg_M z{7uW32wkNr=eV5;*mV#R zAOAayNdo{+`YfNPObR%VYVq^^mNT!rr(if&`I(J;qGr1u4btp=4i1aC8b9B*@FyPs zO?Nuh@~T_5%bDxWKAVa~&pk03`s3K5cW50fg<3i^og;;*<`@1HdkNqFUmKtc_yvhn zI2b*7pz}ML@&j*`Ktw-5Y-kKN;t5fc#9M^8DU<6N?fEID()DMyo+^-pLgT{~0TvwW zBg~*Z?j z+hjCFzEFn8}Ur7-aX4- zkSVg2sDZDqM~-Dv^3z>UsCt$N#@712uP@tym1h5iaeK#{u{;<`PSaMIM)Y6XZ|abxV9{W(@ZFH z{s?@;u1*WWiv4J0J`zhG3JGi{qn}Px=mH$;?*KBGJZRWM;au+9Qzy|q=7Xbc>!iK; zAb>5sX=FHl1u`YMb9U*&nJNx@09*yR@qb5NDOU;X%myqV3*^~k^e0!ikx;~}1sUsQ z=->V~w04P!U=~mfP@fOT`nu+gnmK_hyi3O)F`b*mwhn}A zoUz7Z^_lKa;n@8tgr1cZ(=0n|6Z7()1=-a^=zUm9GL4lC1~1$ zzOsuGTwB(rGYunDCQpbM4R{{9sM+7SN)RVpQa^~RA9w!`)4VdSu^6b=FU{9!FBOPz zNk^BN$p#S>oBx#MG#_PW=J%Slu*m74=DilaLC^Ug7KJYH`<~vV@!pRsCA6N6E@Ef)Xy>Th2J~TJp6HCv~ zo6Q8mBJy^VJ?$lG@aoB#+BJmpLEKlY?~zdUs!TPdrmtI0MjbYViC?#PN)~I4)|zOR zJE0*b>(g61Y{_$=o1-FTiCicRVhAQoroi3lGCcVKY0;mgfIe1 z=7~hnofc*+E3{Tj<`pg#j0&M%z{Sscb|m0IgIB)b@C!D!&s zcOOa@j=m9U0uY_aB{Fn9M28=Zf9rMBw75-%y8Ti^ z>&DuH)@xrZ24zbqw$x2eLn!you1%#KzSGGVKOIKCo)1+@uWYrva@t zh|#9dpv>?w#d+JwCkQ*4R)LNIt$c!0;?_Toel>bc;SCM{gHupwzrv*Y#9)_SWf-ah zz7clyxgH%+aF_sM{4Ve%r>pW>K($P!d4YLJr}afL)4o3hbp`>K>qnd?{Yl#-%qu zB7WwprMfhPw>4V$#^b=*K3q&0{mEVUee(c75v6WA>0PF-o!M>`Ob{dP@|A1xycpsk?kbA^#D8S=h}gY^wh56?2ty9;bSf0Ui!aGMt7+DjBX*O4|Ak;9CEC#e;Uxn^va+DGlE{e zj-I>%?+&>-p~b_7XkZY23WTf`0wcru6${_3ehKTE-t)c9+Y@JDS#P@M`M2C6&Zq6~ zvc%HT%Ywq=8la#)YZQMJ(5s}b65(nwz#=9v=*O6^!J(O-_(g+HGheHH$CrOCdEmo+ ziIgW!3V`6bd2=n|0yu1iJ{XJRam2zv^J$mSk_@uyXiu(p^YK}?-Ltwv9jY(JVv2YF zaI-liE-C^2su3SiQ+k(fwftj-g9!V{+Oh6O>l`rZ*7xk>vDdTRqS0aC1T6>Ptx{i3 zS+?7#oI55+T1BZQR6Nh=nOq9!7N(R^(Q%Ag7YnIOqm%*E_W|RdG&Qc3F?H_Aa_bjH z9Sd218*c#auAg`haCR1J%@sSzXbyocuZcHZlln5lOJ3=>iARNdqK3LSn9b`?W4Hd=>cQk34egD61t~Ic#b$xtMSq#wt>wv*5ZuxWT|%S1@iI%*ZV_G~1qr0> zNy43Sez1t@8Aj`ykn%<`{G5umw!kOT;OKnfNd(fa9%O5PgHFtV%I7&8-ZMV!;G=rm zx7qC3RH7>oOwNrERn*U300^$n%sMp?D`Li?= z_ye~`&_C`ppb6}b@e^fbn&pxF3_3qKZqZ4(Md9u#BXUT|A;l;DpkcG&p5FXfYryL+6lHoAV=gQkZ+j*kb53;xT{uiQqfRLrCC zPC8FQo4C@hJi<~Cv`Ys3#?mRJ8ITn}R+T@=4onhRrT&p7XljPdsUWJr(Y}*kUbowT zx^LgJCf&R|B5i}>k;y9uUYsQ{LQ>R*{lR*%~m%DjpD+J?F}IG1$)>U^}>imwQ0g#0l3 z*Z*~A=bQY_8ttw{%Q(XG_I5@WNuAdhu3U;DHUi6A7fKb(v`*3waB96}+I8m?^rETk z%?lS$Gks$C+%a{l}^L>UE$&OW6}3 zgVWs4b`0K@vGr%|{p?AD9*O&8kMh;8iXP0pckO>(kvt_qzUDDCP&QX9)-v|nMm@3D zaFm@mW#MB{BRSToe|f5mP;7^#tZ#b!@_5&9<~{dIbn-3L;xR1O1?%d%chohvQ+8KH z8xFKDjFaWPU!9KeCu>{@{C1hi^6@~THxT-Daj$E+H-z5R8Dn(IcTJvoF7tZl`lqt; zOHIhw%oX#KYl~HE@(Ozo zbX|}BBIZP>g=ZZtpkNYg;{LzIMQ_Ei%DK(TGO{{H`Ujo!RzeX#<*H~x*R5A;iYaF} zO&?YrbMaNJ!=eq1-UDUtGeO^}#>tb-^04jhhBAIya`?V8cL2eE-j1`U0H*z61e7Xx{qwU%&@U@Kpo+&A36QLDOuBT{} zCYo;QSN$Z=*WF2XYv?Q9cVjzb3dsg^j3olWHztUjhsgKlH}u5-JFN@aZjqhTX4)MH zFSY51TUgu5Ts;s;6#dWp`$NG6P z!)^UtSG5dPdS(|?!I3sQlwj@n#<4)^r+U(vUz1$5{sO}r;pC%%3J3exm!qn4Nvkbw z>j#L4F8xQVsh@D~Kja!?c>n6=^;Z~L6l?NBa=$9q!A4~Y%;5?Y+S@j@Rqop%>0}y? zycTw!bIyZs^c`NQbTE3osSu=qS@7ZaX}Opn=@pgiqEQLn&7C~pKMhV5Yy=f>6+CJE zPsVU5;^^?sX1#94@p&1eYZgSMt*wRUL*3AJiTN*KpPfM;eusU9<)QSDx%r$iW1=_1>00uDs==;`C6(VFwTDGv zDd>nsKBU0X)g59b#lInmpI*f$yox8StFeaNr@QvZ5yV|7mwh)e=o zt37p$Jl?x3MslcFeC2fAf_r|UQ{Pv@9VPd&lxeE&)~kV7L(di^6;*55CVyTP*HghG zWL4s$`@-%6*eZ|cpMEc@bhmJ}jRHmYz>6zlXC%Z#An4=7@%zMp<>r}q{9<8^8P!X8 zAHhqCPgAYVHu`sBk1mqbYTO#9 z8arI=+h6E4JDa&+7TDiUMqfJMXh~rGPqeF!@JCWn^P!h=MrqQ_)gI@Eo^GaWXoS{9 z5CsJZ|EN7&aGL-5PfhBSi5PUnbqG?X;=ck^EZlGQL1lN;vTmlXhjp-ITCQ>i{G?Au z^ShFCzkZF*wh^z;o$h5)#HMJ&O64XN6p0zyU-kOGnaJiDH#Yn!(AjMLVuY=oo1a%Y z+L>o%X8w6%S5{WWDj=XKY(K)NUuhq4dg{Na=|=A~QL8mnsLR|L%N@n465BTO>!ZbT zfA*v|xKzXy(sk}*PvjW#jK}`^(xHOAcV}5$k2V{lfIx7BCM!GJXUgaEU7;Pxf){x+ z^4ZEU^3l-N};&se#lhxq%Fhqvh?L2Z5JX7AH65k`+}y z=j$4kEq*`So47@lq$D8RX2RyrROe}D7GNb|I->)+#u*S}2K{bJg7f~P=FK8&4rC+8 zVGEM*CeN}?Sjd`wL}4M<@}In@a)&YP(F!{@htcxwmRp<`HLd%8pd=4}r%34Ms3jkp zSvT0}l!xiQ$cyK267SbP25(4pL(c{FWEXot|Hz^vnn8-1>m{RK6UF>dFd|#Ijz4;vD}u zxmvp$yFr_PW%TJjUfrlL!utBiZ(=_NPAuQ{tj$zwywZ5IKznq&T8w>J!sO$5PzgA- zl)Bsx?)LPj*PgNDg3GIae~A*jvmv3jSxBi$M@tLOHRpwvq)rb^WktZY8*i>OGIATG z@RD?&{#bp#Ry*`nztg3ao~@MuS&v)llhx4BSSzl!IM|rrcbSnZzLQ2N{qqB5++cyu zn6XD(!+iBJkSJA6RjNEiUV+-@C&6Xh<)yY2S|zVr<2s+Eo+{~j_z!4$&d;##8XX72 z-iW#_oE{oJGHoVv8a9ZwDLHi#sf=6Yu~mZ^lFE3U-}CDu!-(32lLwF|MMoYfz6%^c z2=Iu$#Y`@!!ao7SP-L=wxaI^8rPD#tOen{6>w0lYo+l2hGsVfReLmb0SF zx2C^OUaLy0+GY#ch^G~{k3R_bB2UB}uNWZ-liUs~dc8>|?|S#Rpm_dk*q)_?6@tI7 zJ?I|!)5q-wQ13-5%3tp{`Zz^%0{LIs4wG$k+u<5HFWn5v)$>Q4XC*fexty$*BHlw$^J?E>YA$HQmF96 z-<*%<_hsKdoo)HNUJPW;tHYg&3;}buTJ{8NyDp4Xd|`Bw?AO0L$Z2H-A-Rkz5PI}S zuUB=D#-=mvBM)m>?1-et`c_gNS{dUdy%njC2LB@L7o@1CPVFwul=VpJ7uBvju$huP zgriON=HrYzBo4-zHyiN{Ze3BnUU?6GrmS&-J}m$|%4(Yxl8d!`rQrQCv02H0I^_2Y z)NLHONi!2_Z_fsjA!fx;oaZI$rM}nfUx2*ZsJ=M&f9l~=m;VpKl}3LV6kl-1tGe;A z+7V4(gH^S>-ylEkMh1xWAC&jht5aLYnHu7Z-TUN?sLj?=9pBY2B-)w(2qO1vxk`9t z2Q;G=CqNUIHNB#z?p`ua)ue>dWhUoIsD5RR$4t}mDj_z zc_ZN>Dx-h*oceTM-)HYk-m?W!`^&gz*Vwh4b<)cBu~_DeNoy$m&XB%S``N)1uUX5d zk}}8s1_-b~h|;p^2UO>GKnDmXx+hTlW#!9fIL`l%9JzOsI%#z356T`sAkFr#g9yuKjX?K{$8gO6bciQ$=FVzm;64ozlm?R1JQ}8=c zM*Px?HgNug0+J=AP_8NWsv9$I3@R$(zIn2ax6teTXccii!n*mu^7nVKt}H#K6+TSk z&8Yf|%EfPbwYk17>s+C@e((zIYe}C8>K1v`usR+w-M1}Ipgn;6?*VP_s`I$HsvtH& z)Ou-h=Dz)u#CD!^D9*UaDf;Yj@S<#rDNojo46e9Bnb-$nq#hvu?;|^p+a=^=i6S#B~ zFqT2LQV^1`zo3_e^1G;QUINj*V@~>(O9w8PC;8fHGwKd#$AZAb;%RvePCzH^HPB=) zWcx;hht2x&NO&aUAt7?hp%S3%Y5*o*tPU-8|AERs9S}eD+k2F7drEXfpDSG<;#;Q5 z;ZvHXP(f^u)cM9UJ4;MeRcLu@>V9Cy_#u3Q3i)oQMO;uo{dDP5yIbMiO`#8T4 z!VYyCd~^cjt^Z`ORO(LEfqm8l7y2D1zk@EBd$+PZkD-JRr77JHH%GntbDggyoX)qS z>tlll7!c;wf@iLdx;3z?%N_&#*wWD5#D~rs%6t7a4Q2PPAuaH5*Wtnw{5zvGGOr2@ZOwOT z5UQzt^mI42{mL~3t4{Uob1knWxs6-ifAfcUbtpGBNp{qt=Z7@5Z|wL<{b;Rg1*`k6 zb}2tw{^ zW~`%}3>?jYcE3QaW#;Rcc)Utx0Ikv<{6{_yc060n>e%m*7P7RobR*r}1J$RO_BGq- z`pNE@Wh>3DsO@RCTV**feF*g*;PW2hA5-z0mnV!oU?)e#8u{srm%COIxh>RP9HFNK zsA}Aul{y;_Hg&b6yvk7{Lu|Z!0=Gn2cIrE{ML6tkzUV4Vt0bCvb-tP?jmyll4@DgA zP?j*%pNxnN+{Iq}SXK_AV*WFSqkHvxN-#_Rw9I{&9;s*6HGNkh2`&*6iVGF0fBo+> zOUs<}g4F2(!}b&6P@LSk)rM8o&OiH%LgAmwHZvbWJel!|Oq->jsPUfGPr8a~+cqvv z=ZKQPs=WmRgWICbDJsl(g~dxh(f0MdcwX?;>9k4X243s0x;L&iM_om~y9i9Wf3llTBuI1R!3gTfz1C^kXjs>s(Yy7f zPTrs*GTllOaO3tE|FTs!VihsU@ zp0gB}v1iO;xAiC6__cCwZP)IHC%R90f=E$jH~DB==iTXE)7A%toD5?SK(s!Qc;y%bHr@-cty$CH0E?RiPxao zWtaWSeG!pOsyWkK>(EGp0Pe7hKJt1W%#le<;-sYN%1ADb) z5=+C&%lics+(FikO0oqJ>|!6cm0~8CTZY`LKK=BC&cKBz^aVk~ST_!V`jr&3`-^QJ z*eEQ};^N8Z<}W zWmT2fm1mapDt%xz*xJq6X;-reJKgz9Ya1BVyR9Voy!zkdQ@dxv2J4tbuPpDMm?^(U zLU%}zf0Ylwn@D4Obs?dBn1BjX8=ZSvAPwrS@0xkKL*1Oo$-YNAo;!_X;B#~f-{tC` z)ok04aw=?$71XyeXBQV$pfC-0aWis({2_PF4laF{^q z_7~`hCOll`Wk>@3!6zAH4G5J|z^vGt^Oel7>tao49trvz+rfcN`d)$KC zoVjVB*tDZ^ODH_EXY7`%_2RRZ;Y54lKaWG56_U;~Y^JU!+iikqT!J}9;STHP>HNTr zbSEC0XT*^|MK73u^2ps7*Wy(;NaN#to~an%w0B_L&2G>ovg>8DlF#cCAS&kg)@J1! zL&R?ZK;SFV@v-fi-b@;K9s{n$dg1Ub@q`;w`L%n`w;M@!o$;DZz^#N4eR+1q=7woD zYFl%i9ku(Ngyrx%d2`w#jeTQ2fGkE%ihmKqzO$aR68T8CGWH9 zFLzF`@=yT^b{*nC#R9VC_}!w?_nROX@M`NjeoT5#S%0?1zn#J|XWzpf8ES`lW!9Bt z5gv8g(zRJv_9ejO7;{+kqefqvP#xVA&7F9DF7Ez}sA_J58x|^5w~wo3so%|4q0k-3 zR9QXQP6xMd4?SGf-%*a-VY^t6D;Kk+Y;1`yFRqR(cd#9+mFc`)Tv%v5-$wO>!&){b zss_%@%e@Y*z-h(vHt|eJro{f%nmKCTaKIgW2b9@A_zoDO>l=p-&F8kXg>H|yL;B`7 zxnM2|Ub-Qe-4R@9YH0s+14cMNB@@NwIIDYBSbgkdKN zPPM;+a6T7Lxc$v9%aTj44$f{@1qw~mvx8MwfDxYLT!AwW=|av`f$L4wlG*OFX(vMl zmg=@eLZsVz^<9_?zB%;m=^Jpw< z(j8pu0^(mA5lFTYz?r6V2TfvPlQk#ObDy|C$7!zsaS_U0MbMviioa^E!)`6LTUKtC z8&!M@bK!|iubw>J8|{_MXM~ECEeIMO#wPaxkfpU;$43&oryuu#>qMwdx6G<-K0?|~ z`Bi+N$vh;2y+aYpHpKvd34PQhqyy((tD*+weYQ{@`ZRNQtagY zVelJQwN{vt*ER$pTheUn7t58pQ03?cXM^hdA}<#Hb?AXm?ozX7v_*?$eHC}#bfJwW zC8qr=;Fq1uf~)-3KYt!4+#sLQK(e5KwLsNXXlG#JiQeNPk7sIXu_H!^q-NA#kk;A9 zYU8>Of_Ii$lpbp|1$f(46@SgGfpQYWLDwcU6onq&HA=ueRG7w&08UoPN%eM|PBGv` zB#g{}46uYIaXo(#KnzD_i|@{~qCl5t$HepPQQY9^UQx6|*SPr(!b zT5hEIV?`!G$Bzkvoi|fj+Az(*RsPqvaNr>p|K-~n z#78DW3z5nde0Bq%=NMIR>c?6ogxISiVg#iq>v2$S(k4moMAN;ZA8>t4QvI;sqs6EJ z7Nv`EQ(HzOcgg#bH@dXhy;g4gQvc!KHA7~5rm&1_>aPh$xURcP^g@oA_7wIt-cg{; zYv(VBNwBMKGjkaFODk=o@a$F{{S4^wj8cIz9k{NQsa62O+ND}CvY2;A)h{igUmDEAu5BdaQE&^>F7$o9ml}&lMf_-ZPwg ztmjp1kiRK(1C7nu0%`n@G2l;fenAA1%k}eB;@JAboH%h1SAyybQ$~_%jhoYN%)u<^xO(>Ce`A7`P}>$fRLx+cEiG`;6$-? zyZ>0+J`{S)CVd1Hi3}J~$kMTPSrkwJ!JI>;NK-2a#TzpS8@_TU)>!YR4O~|Yc!aIJ zz#Y#In6`8qC;Y`H@cgQOzi!5s)MNIP3doT=9y?WI_g#;kCJ^unZe}>neh%I|ONox= zvPWH%YvyKdU!8d*$t~t~*S$BWuK*_@HQVAd1EN)XTa{em1j!f6x&BdMj6NTTEo+P5 zwUzoWj-YX(z^Q5J*zP*LzgN>_ zdoi+(J+ROM2-_x={&6#tQ7e0OX<$E46*APL$+k`>zvkT<>N0@NipPU(P4llnKr8}Q zwS^f#5`nH4p67_;c@aYCbHqkbOZ&YRarc3bCA`pw0YGJq2A*PXM4%^SmK}np7pTkV zLmbG?b%5>Nch72&7ykH~NMfiJT!KQj%4>MM#YS$0Fc#tV zVo{@18nVYYhk+#Lk!e{unJYa4+T~IARqY7BtZ$fSSE)n2*0}1Xz%6lLuhS1&bD;D9 zKuJuOUV!g80_+N<1emQ2De7W}$mrV}nBy5KxQk6VQ(YOtoAu2u2cU}G3c#(k9fv)Z zeuP!%Th_xZml}a&!Ox3&`L7VMUKo-70dU`X*Hn>f!&obgJMF=@ zA<$GB_pOKT!XJ;~X&B;cxUX=XOPItCqmoD@TM#!>ljhqA%BzcOpBT zW+C#Ga}|wSFF;M1;Z>kwnE4-;?>Sls^wSVhc2)e2V&(^%=n-p7kVU*?0!fC_RwW|G zJ@l%&=7>@MpYX&q>yc2CzYY}61e^yb&+~yaM^C(!556ws`ql>U0~^LL{ME;5Xm5F& z>|9%2U4}?TVI@Ghm--<=tVN{FSbNh0Z0?JU5Iihv$e<@bQ0v)f6ciTqUt|B&&MLM2|5ORLC*Di^C$_^EQ>X&kGFDRW0OK0{o2r2t?F9P%!JHJ~CyT$gs|t z)?et!vyvJx*=YF}ILoj^*l_h|fZ5e*p*4Q3+*hN1#g47j2 z*!(e=lWYrqkedL@FrkbI1JPSKcYSw7=eB3Z%M}L%B@MmokZhkowAiEz*k?fE3$kl4 zb1^uZ2?XCkeJpWm^|stxJ5N_@tNz?B;YRx?!Az$1G&sg)JHG?c+B%;_8X5I=nyE6s zOS81OCiq4G(zm?XS;OFmw*=I8vEoW-j` zvpM|)g38S;ML@88n~(7VTdeLTlqMuCJsmz11?2gLax$sez)_K2SD zR#>puHKOGd1UNydx*~il{ce4-x3yoYd00mbA6LbMQ)m}m>1P(Pae95H?P!sA=i!72 z1Z~!Poz#1K-k~;(cLfY-*&Yp=eX1^fu*^tYIEC>jQvUvrhUSV2UIdNp4FtRS33u_^ zzdPh61VEmkYH*fKeJB4>4~mUlt^V=`tertkLIWfVYY}1iVD)@}2M9JIaDFykj~S(8Ui(~JSe+W5+)p4` z2W`5qYeTtcsx~Ac5*_GUuUVfU&!uohWq`EGNf{LnbAG-uT>ge1Di!OK1eq`8GatJ* z->7r*=6G53RGl}PDp1zqno9h^B*{#w6!+t@vPYnS(Xp8&GX9_=+HW#v3`CQn(cEn^@9;fs)Y(_I8yX&| zj%XCSQYRt#vr=H0&_VV$7JXOT%esq*R-$3|*DX!Z-hlh(N%IT zMx!=MF-n`}ydQo+snnOdKT%IQPTm&ilnYXs+xU9S2!ddv=+%y{uCFECk2=hrT?vOO zH}N)2K<;i!dMg5$p?&?rC5*rOxUBRxnmyyb0;S0o^pk|N_taZUJwnTy*?jkSa|f~r z_z+U`g9WeD7Ktb@FFEnX=8?Ev2?KzSQG2YAy~wQ)soov#4qIADF?2X`q43Rss^`(? zQy8TxXyY#&q3yw|Rq2XXO(y%-Q+w3-HV&MbE@S|Fdm->ep3}-$wX^9bKmbcc4-=V6m|B~D##2?Tvv7gps9 zWbOu=hWOHPnfU-^Ic1C-t znSs3QR7V9|n|e^9)7IgaW3U(J|K+YgdF$cMHg29mw~f0oJVw+x{sP}E3Ed1Y zg&I&qeTAOucAj%rDVBNTfgyltE2l4Fkll9 zMJl)Iv&`E;TMb4;)7_IaD%EcSU9vy83KSO744+Np#7b>m7<#$pTY(L*n`AtVP4V8U zib_tl6Zt^EmX|JD@qkbxvw2MjrDIw$gPG^R6>{BzOH%k-yawFC5jr~F5YUHFT+j4` zIR_fP=@o$~CkkJI60NcBi&ZMVHD_uSRO@c-oJr2Ayl&~n9FpEgF&j3B@b8@BfJ`{M zNerc}wP`Pr_#IB~ws34;gb`~_*ZXdgYwR?LJRp+qGjIxgx&7t%`N_Cg9|=^qqhEk? zep;3I;NkKuTzP`;L$i$Ex-9y~4wBLUdx9t2xC<(WpWeq?uz*~bHwXF+Bc6n^eg20f z_N3m2Daps!4kPSa5zgVJYkj+$AIk0n8{Y&1GIwk*>5JjtU98U4mkfw_l+!37^Y71A z>}4F2T;KY(bsXB)VAcujefw*0^r}RpRE1oX8^9rw#4DYQF2-G2?+)JbP)E31e4Z-& z*(}ciX}Z-=&KQ-(PUm_fBjap7US)A$l2k*ZIZ`iY^0(=k3&9NG`D$24_d&tzhpB+G z?yN>p>H#r6^JLG=E;^V3tK$c@MpLh+Uchw3mQ=olf#A5Zcl$TmuBOXOMj2cldA}=k z;->UDfx_K}lyzd!96m;+Zr`BvT^<-A^#;BSI;nR5eHW;IN_8;4^gBn0Ntkw4xU;s! ztA zA$jA0$x#&0B#(76`zqhspoj+F(|td3$C@r=KBT0~F&!~=B0S3B_5ep&QCj-IzJ`Ay z%iwdx(snuXj8t77_sS9ib+Ku`Nq+LnBtr6)larH8Aa5^ex{QW_U%7J^7^t%di@#lO?xOVmGs<4sp$WTEqD@EvF+BQFoIR`dbiy{%2POb25* zGQe4ZTYuGFE3v%Qt^5cenUUqT=y57RZ+zFCxY8DYGxq_o4YGG;f?Y9}F-{e}`2hY{ zvzLnQNrrW4iP~>_5%U#M8{4;kjc?~bGy2O_`)ZaJ3Rn+ZsPQySvS0O-3<1-Bm@9Ft zueeW~%|va78M4OE13sE7P}Xri(->e&x{lMf_ZAdE>3rc-X>x?UK z(;oAm4YS;VL$5cUB2vbjX1o}`c^`w;0%1UP#EMP#bmaFWk9h()hEkB81nI|}XeEhw z-(!zs0w#-XA^~XChZen@e0t@-yXf3Rtp>k@3O--4z43`5oS?xKm~MUp-sa<-h_k1$ zY}ULv9fCDsZ#9}~{~dFOG;WK6R`|r@h*4qjU|vYP>_VDfL`zjVKe-EYZ)1NZ9>*t% zqXkt;hd@;1tiCe?rZbC68}EJ%e!%kAH;qzZMjo>tH;|7bS4AT)&Q`s-<(u&)!jr5- zGW0;d))^+B+Kg90&3$rtYk+zQ_6?X#;&3 zH^MnM4eJw1Uw!-CS`mjMrD5e&Nr06xt3RmoKK=#(rYW;e{v%UHHVO?tPV3zEWwjWH zO#T`Z8$jefp>mc$NXJFwrCAXtjYpBoZG{TsaQH?xC{ghJ`{fiweb-Em+EYP5EpEJs zBb!^Cgr>2%9YUyya6rTcNULn4#A8s={paHWd(s%~733WwEb<1EYogOG~f40Gr~r;xgbeuPP97=Mle;q2># z!(VDxz&fx=mp`igfwjMJdayp@m+HDP zs&O3#{Mf0EG^(6T3&^1g>hmxu$e|hg2l}GE%%@+ zEhJg6yrjrnZojKIV)J=;n-+3DFB_zrnr4BHwl0EbjoL6rD`EcX+v3f z3ipbu*!in*(&C7h5r8?zmCu4cIJ#A!Hy~>^N<3WFae=XZ5JQ0eD`93iF z!Zd>$09#Dzx-18DWfE^7 zZ5H?vjy$;507wle|1$$et8<(qk(HuhKF(en1#G(rWb;D>$~kJF(7zeULgeA`qCvm?i+-94Qy?d=PTB4e(U4IQd+BO7?HS7DvvX_ngc zh2F7RZwnlss7J0CL=nRi(~fUn-~0`1i&S8@G{+hl8YWNpoo~YSrxmzXQUcO)zsKHC zIL@^cF()oIqe+7R;4=#9J@e!l#BwX%$_U`4HcKek0Maq~w8xZ>Kr1)xn31K~84!ZJ z?J?HFm{q0#SXedoY~6OL0Pno{udBv}^uvfb0W5b2%YAo{&i+f=bn=(iLdoC5W^0(;#U~qZ#&&@N_~7@yR~<7JRpLd}_S= z##QBmQ0fZ?UeaRx@9SkYCoa4&BB>&p$xaV1 z=92#kBc?ig&K|o`HFv&Mv;8F$SLgpI0Mqu2{T-bh2ifTF7Zo_NvQIL5)qu^o*ht2) z;MG9X%O#Fyi2{PsSe#r(XJ-t61A5G`XAUP$`9$(X#Ezx#Zq@s<>u3uU;C+kDc2; z|L!)b9Ws+SA+Azdteg)o;c74%$7n?mv99pGF18vi{84dYKUSQP#8<5(RQKRZ($(CN zP4WZg>su_NHG!H1ms>Zzc3(`&7mW6D_({(`D!agi&95n1<=~`i|AiUNCz((@2JQKg zP1e3Fqgt=YJ`juAEPhu{+d1(PgZF+{|{6Du$58YA=$5uar-TPJ(*WrXT0$sm14?b0x_!!8Y z>p;D@4F*uzCwjgW)|2=9&Zxim0N5LTrRIFWk*{=Rh~_xY&qny8g#a6|*KmiYOj znZQ&TjV@pZ<&9<}HgF*xQe+)n|G=QS!&oOfWyb2?ApDw!3v61geu)rgcP5k(Fw4@o z)w&0tS~_lCY*Ei$lldZ@Ku0b4=v8E(YsYx!B`*%UtWjYs<}T!Lj3emO1J;Y!ff66U zC5^#DcZvfW&mw9qN>HF33TDpSF_0C?mWH2wb3M(mg%SPXYOS!b!71Fd9667WI(vr2 z9;0}==&PT8zV6vW4zQt@6O5aHF?*b^cYp?NCB=Yi1<=_6M6;Uoa1~3Rk9FAma!$i9 z0GpWj!&&YNrDlTt^INaN4%{UB$>O9hR@L?Z4&t5Uo332Ie_hN@zbO8KEzItt&3Ngr zE=;rM(+?Wz_G3=Y&Z}Ah9|j&*O=YLo5Ik~OF*Z+EU^*Ws)^Kk--3`mGf&7Kzw^gnT zv=QVV&NWrUuKHbt4%{mjyzPmGfy;#DNSc2arVOz6?iw;G4;5RFxpOus!oV{ewn-U<1vUh(@Ms0 z9q#sf!gDz(^1OaAgnlRg(jtM;>WVidqG$WW+bgOm0)%jb;>*JD5y z`$mGJC1_Usl5~SfVDe6l8`F5FQLRQ}{ejOkXkXsNbM&^>spk@tQw}!fK);9IdAa*= z>MSesAFf?=IXGW*N1Z2i#)~Qm)@w6=BE){#7~{6n(d$ANZ6gOC3qu=^Zs9UXuKj*z_yWp}MSOABc>Y@+2@T&#k*0j-|#0j*8&dMioGU4KL%l5@1kg(HD#Z$Z1ubPxP3_G!0dX&q!YQ*<1*Uuy){9RMc5gW;fEUVc69Eoy9xq+AO1DFlrVH0~&I(XiKq}B*sgvxbX_QUz9sKvT zh1EGsV!uA{E3*3wXfX`q`LlqcUIx*LG}+uB@E1z`=j#}qb~}m z`CD7o>mI130bK19?nK@^&4Q@gO8*iq{EW$t%l84TItCPurE!R41lAK@LfDeVOaOs| zDPW=-zL7a6wIG}Q%rbwRkxTnx+}4PNVU;MxkK3vUB@Naz;=gixzXue8z@Xv6wj1D3 z06}bXWb!o?L99wH_=(*n@-KGV)CI)I6Y!klaATmOInq`tc>IxKDxc`H-^H2dbliBM zaF`$p$Z0ci;u>if~6VC9Zbc$G(;*~>;osq)|x^*q-U|FkOiKhSJC z&Mir34RBwfrfjr)xZ@kg+@NDrcUd5tzw?Ve1tq2F?p*Vfj?ZR8K}vjqUWMx3{(jwP z;hTWtk=0>C+x{$jzS0(?+f6%=J|9^2H&IGvy{_|eJx^g(Jrib%{=#h)7@-W^Hb!SJ z%DfI#nNF&0jVR~KtVh0q*1aO2MgM#f$~eVxcy_oBcMj;m*Zi^iPQhK6$k+tIb#5x^ z174xD**zea#3OrVsutF7pWD3l+Qo7?Cj_IKz;9mN59*+o3^4GCvETRNG*CKJdV+ce z#dK$?B-BbPdHd8esPtsmCO*(?Uvgl-9_+gIH61@21>Gs_)wGgeUth$G9fM58VXbF zoB(W}3CvWeE~Worn6VB5*2$FouIbOL*3|mO$vaoc>Xe%{v3(wlP$^}QNJdN}k=2-n z!#5I5C|hr&YU`U!Hc6R!94wTK`PT3RxXZb}N3|Fn0UvjHJ-WJV@L7t~QW7U+#(eL- zuQ0HDGnAl&6vlgc2l8;Ptf2#SmEc63Rag4v(OM|3gdo)Uq%*v#&`c*9;i~VCeN&)cG8v9}0tVX%P3BJk!Ck_sIWTzE zsge5r_(q|$>?V{7)+xo$9R?(1X4p4O(DOSQ)_-2~CVOXm1J48++KL_ASRsU>!;TOtf6;FNVCd#>JP#RpsoZ+szy;E*~+ zz&X5((l|Bvg}Y!F_z0oX- zZ)@$%y-CpYGU<`yYK_w~5S29;Ne;G?17nmXKuUIn-$-&l0& zw^{n|qM$B12LiqdYf+Kx5vavEv_YXlxg8pyF*`>bI|_iLyxD5B0m-Bhml1J(3XF&T z{CIv9Cw2*3bu8bR{d@+jEWyWoV1f?pBG`-GoZx*q*+}exS7Nbwgo=;mkH;i%LONg< zL+1(>-~X(Hm~*%N+aaoPJV$gQIiR2UdxPlPO9%HfayF+j1Rxi&$g3=Y(~-WHk7?Zp zmW;P}sprT9zJ%py7qGYP#($h3L=x+;_M*IH*|G?}cfF7+VeQ;3(e$`Ul?Brwcg) ze3O9_Eh=Ke#-QNDd^3>kg*sgWOi%usA-z#W=1c4B`lAy0=XwzD5W@2d>7j~Ekb-YE zEf~JhAgz^4Iw1exQ{az=p1jv=oe#%3DJ-Es>G}^5Mh&ZTMxkJ(s5UaFkCGw_31zA; zNfOR{cwfhGPOIe&5x)m>N>WlO@M)BOm{ijv2Z%Y@2%eG3N+dW@K^daL?-iPrQY(Tl z8yTp_6C#r(9p)N5t*n?I)_T9Cg!IbK5JX52B~Exg+5NNH^a7sL9@Zmd`;ft}7q{Vr zU#R?7)XLtx6o%W<8;p~2-foWLw{li-`9vLB@PGs|@5Z@X>M7pQdbTlnf-Wa>&eTrG zgLcT^5S^Bnvh8yIV{rZRwsg>={-W>C3)zf4Cx9|u0xPvr`_;kYm%$MPF`3|)p5laE z+$I>+Ulsrm1H5ee_gA4&0wj{;@JL!nsOLC*2KsaR3!#i3{xNLBCrRl%2xPS509}5E z^6zB$DSBm%x>7K^qyucT-eF=*h!<4EU=bK)^?%fOQj$P1hS!F}ntNO1gNOHS3qUJz z0kxyT6TrVkv4SSpUcA)TTUh%0LG@J#+uAk6rH$2Vp@E3+W*hLZz9foLqpNv2Mu7BK znal`Tcn2>@&}CUG4WE%Mu2G+8Wy6Edm2`wkUHTx-? ze_{njV(rb6L;C#%2`}MLDXmP=*Pl@Dp!^)|5R)uu)U1Muj9r&=Qyn7MPJ!V%*gbC0 z#^@>zd~ZE~ynsS@j2`)pzLf@P3l1HJ0b1Mb7cJ6Fz~&kYqk$p?G`VTMa=4-Yq<;V- z?XG#{qeKr;H>DMs4Bc@N(fK%|diwa|daCz70R&=vyk%@ydgJVV|I6B&NV)2o12m_stVvY=luj(XiNstQR@=#u&6-vrO30v=duJXddPPSO(mAH1mI;$bOxT;BD4CosFeO;32f&^FEr~m#{oRCh&mR%7)sH@b{u+i+t zNU!|o(VgCq{g#ZN3o1!l=o&1Jxe@d3Eg2ZIlzQE^7^85dkC^Viefvx)gAjQ_IUm2r z8B$A`Dvr>t2KC1J_zfRi;eEn85P0`H+FMLOqq%FFfq?qlR zj0dq3G3fj5ywB%jdqkRQeK=k;+j#X*O-bRL+3Sr2uAnsf9i2EKbi)6M49w>y*M0R$ zcv>UG0FrUKv_K9?LNn2Y`2w`5&#aGHuR?!3uttktY0b{b;Shu%(STCPAb-Y%p{fKi z`*JR*O~xI2@)>-al32xc0&XJ$nF>qR;rytsKevU%K!H;R6jS*+gg#eVX&~@jug6${ zX)DFphUAVl%H0eI%pTYfJ5-rne(!6u~x zm)2PPYO|F++3PDa1zHRkWnJY}sPuGNZKAS zy=vjIJOL+90wIJMs7q(4+xZIHl3s8-Ns};TJbnINex0>{6L&{QtTO2m07S~TpxvML zxPUPN83L#=NB_n^I7WdM;`Cf88szkXh;^Fe_f{{@i7xf5y!j|+IB67#lv{bOM3ZT2~vLta~Hj-tSS^u2QGK;Jy7*GaLC%4{=PEfX(MA9 zS?yD0Je3NjTSX~2z#-tOZ$!gJWMMa#8rm&)71E98Z5Mq1VST1SItGky2DOz<0WpSt zFhF@R1k6*|@qa`LR`MA;NQIg%7K_TD?i9A5pFz@;Roa_M%M~!dA-__avHSR%N>)^e za!&#V;;w=k?+cXR|>au|gH#eU$>O z3V3YV`_De;4xNS*(is8CYdIG}tKf2{66}@`lOb%I>x=|R>e~MN30bw7wmAq7rKxAu z#GX9xLT%tL>9$a?iLU4X+8|p}&UO%@014%}AUF*oz7Jn!Jr$mgc2v(Hl(9Z%wqj#@ z7_Cuad>8+4-|x1ACWrrieef;kM>qC zCyYvr={l5CO;BerR%Xp4MgU2Xr$MO-@prURA>_L!(Uu{3c|8rzGr{%|`fR=DGI2e~11x(oO)u0Wa1N_-~jfVEZ<4Ubk83 zN>jrj-ZOmN{%KnU9KGk~-?}%TUXgoUZ~q!S8hx3`fW~Q|mpt9B*m+UPsE@>eR5WJ~ zEIUB$9q1b1O-Z@I-%1F^;mJ6Wzax{m8V+$@vt`wB)+N*L;P?&o_WK3K3uXsQl)u)c zUMDS~G{#~;asX?=2N74iBp=oOoHG^`N8h={P@j%mXrKoJ%(?c%z9oIND(QvK=uVxo zL3!~@p5|r-eU(yRyFCF?^7jt}2lP~sz(245tSsR4*nX5OQ2o{w`x-Jm%u}R&&m`@?{YSRPeGu)G!}`%4-e5LJuof- zsMT1VcLlg;YN(I+zJO%D=)zphHpd2={8-+xW<#k4L`mW^GvhQ!fL4H=KW=OZ(iWg! z>jJ#8j0*wI?@f}jGL3>ZB3`?@H(eW6x(=O3Q>*w_vY_}pL{1xeZA(-w=L9sYZp7dLc%cX6*+g&$&c%)l zrM$H0E3nf^2-{afBmH0y8H45xtNx#gU~7bcM+3nuxVJwkw)Y7TjJOiGzWwV! zx%d06L>cYM4FtIB-8b)M1!VDGpeF*}?0CgbaAl`ldoD1^-f1f;^B~W#P8Vi7{!Rq< z;HRVD<%Se|aNkV7x02E!F#hj$vhie_=Z!>A#66E}|O8M2iIY_$GNPhr}# zX`lViY~Y~)o{Ek{9Jql&bpBO0h{dmMIRg)ZDR+Cj?9C9|`_nJq-L-_v zuc+L;91H*mM30yrZfG`#V-7+05v*j?C@AK@cKK(91uk(FIPxIQ1PxHPrty7Bat4N~ z?w0g{A(FSh89}#vZ44Vx*xI2I z8$aalS@?sVlTJnU3N-$}xrA9ft?`EmgnEsJ*AS&D5>@nI4%YOC!ZF8v3LKzBpQEs1 zq80-)gvMdCOO5etV1};EbiL#kX>=YqQSVvrK=S5fGU_TJn6wK@f3LrLfFKBd>6{l3 zNNoIfCI3!*XlC0LFN%kYf%L<_FI}Mn0N9|b{T^4&xjSCU2*rqj{9S8;jt;fH`sRJifGFt}>skhFT-JK=!=ib2MfSwk52K zH<8QH-{!mC^TW;M2tPw8+wP_ah|WRf2CV84-HlhEqm3~xHR+4eF6hb-SA@h{1+3M8 zgxdZ7?eig-0v_Ed4dDn;6HxGd$z z@)Gvzwy$;$y}sHphs%HEswTu(>o zp%Fj{5%`kpjIQ`SV{nP)R)~pEI->ZWJHEPJF56qK1_g)WVo0M+Wv}zM{|LPhYy07d zkr2K#i$jI48g+yt9eVl0L4C;tnJFxf|T`Le+O;9fUp-Vj6(t9%IpNV!h6Q7^fHIE5g=iSpTBY(aRY=Fvm-(^KTfxw}!@XcSsnuq4Yu`N&1IIc~&9hI3Z zWW1rYWwgRQ*sU4ei6ud5DZ#t!PVXHjWH7*=+@U3T%Q)+ zFTM-8b4}C^7^H}Px0{00HG%uZyRMVy3@QPDL88u=<*R_B8hCOhl|(j{N8?1Ps$DBC z7lVGJc?qN=w#+gfD}Ouse62y}1&nUK^ITWvGenS@U0s^jG{^B{rZ@h8M-l@2?xam9 zP-~L>Dm2)4@ftt-!*-wIMa|RlF3{R~iUfu)70kdpyn^NdBriUgyVx0jfUJNl^n7`? z(sI_j?cE}&08(-*!{lt=oMZN101R2UI#8CwMi9K%t7KVccSjV4u_Tz`LD89Mpb2_` ziu9LmTb%VRsOi!O?#70fOFp>ZeDUzQ`^#IwPw)mSs~2CA-CPUP#W`TviQ`iAU@Ch~ zhub$E9R!Vm$?X8UewR9eJ6vf(Ykkn`QI)VDB9G{{<@dBE;H1jY>^unXcsSQmhOH-o zcwVJyps}bCzJR1`6!a+?IAG3fAwC#_yG&HT7W4!IA+t&_(edP6)8U$%l2{J8D1@r z6`D$BkZq_XI5`*TvV09f()F4zP8oohqtB|F@Tw$b>ng-rZC^C1%DmAe5C%Q}1=3Tb zfZ0v2vz>p`GUP76E&R@s=_nO>P<2_-SQ2(#H-%8<)4{gkXA>}OcIHgFQoexDCA*Ct z3{Dt~-B$e&At&T=qz9ci>dAZ`%3+)&DV`#AI0`-lG`v_8mtj|JNk0V1 zSr1$sK@M-pWCkfAyz!4=(nI@8@>fT2jZeJfsiDiJ>eJVwqW-Yl2hY+{fhZzM^$+ef zI>@#Em#Z0q^=*nx>>CY-BBU`J9GOr|r>YA%ygM$3@$+T{)PdDM zI?~Gj5_iSETb1p9JUw+lRL}E|Qb6evX^@aqPdcRyx*H^=OFGUVqy!YDk?wA3kdin$ zB?P29CGWR~pYLz~lJmWNyVEl}JEQ-VJ6l59>5@)AW3C~{C0m9NMK4+zmm(b_Meb0BgF4yIq@f$MSiXr9Nm-}Q8 zs&kA3x4k_K;s$V&?q2J3QQ(67HP})ZR+4@VSV+`p6K@cEkPBq>_^n>gZtUMNCjp52b~Gv(x)ctUiW{McUpkx z9B;nGK`VhGs>Zgu)k$EVq}{s@A-d4O{*)p~s(m?%FQzS%Ev^gPL&VQ*epB`tM!kJI zK+%3n!OV(80)%kZ+na_g=uy?kq4xo^n&52CzFc{@(*Lvp3qFvqJo{lz9rb?_#JQ{e zJ?psrq!9aZDXJVl=&{0$zomJC3(fk%O+d&LH6$hkDm0I9Aw~THB~;54)Gswv81#qZ z$or_)8K5^riPLJ0mktPzISpnpC6S>DzC%xcq*hv2kweZe7af5~UCxC15gZlR>R68D z3E~43Zh^ia;AO|Iepidabuu6^(Sf6j8iaiKg&H(M=#lN9pDY*>H3F{sGf6biq|WB` z577HiDI1Ni@enZ5P~-jc2w^M;{Oyct+JM?d5e=GS8U!3C)VSWl>s4|PL_rOB4!|1t|5^;q3Xo?;X+P7$l$0y;3dxvzKo@=ba0`9*QjCAjt>W!3D^IoFBk%LNhytA%Q25@;4|bybkHFrCHD0=#HLQ2wLzy!FW@-JAO0%&| z3Q^Z;U85^fn)-WXKWNnzk`G-MvJF|ZR%x2U;7n&fa3ZlZg{`Hz5`o_Tlge|rG>2#u% z#`VwgjHlB{I{98eM!G>u<^NVp_y`{ek>y@|=pUY;)<7mmZ2muR+EVN0e^-Y<)Dks- z!W1e}{KqBGv!AIn5cd)Ezen};H_+z6zWGEfilk@z3MPrKDzJ8+<-kb&FvE+qtP zXV%BKAVpBGwI=Q#2I%$2K%dTkVvU$8XgW~ge>7QU@(G3oF9f?JqzX!qYRW||2Y&DO zK+pOSU~uXw&Fe486I1aCmut?<5B&c+TH8~(9NSsm(<0D;EA3QKxTT;=RmAs&WvXs=0|2SW}KzE~; z`zYoNxy1!}Je^Sa_vrfoDI4{jO!0r@06l9?|{0dy{Y zV>68$1qR(+wCn9FV;1p3>iuNUm%=F!SU`oHr?nO|Ehfd>lV-YCw4t}A+ zNgeR%X`$y|)UKa|?v1?!k8abVIj-_wcbgzo!QX-=c8K6)$FLE$e-9DcMZJ@QeoiQj z4*$pjTI%PYjP9)a*%P;AQfKaeE%@dL76La$CDMQ^p%b@*L8+78sfr_;GSJ2$38!+K zrE#>S5expEg&gFlRN1$70}m&#&yfvzIcIx%l+rng0>-Y^^fH;*LXg}vm|%3nP5y}6 zVUqLEI!?Rjo$p786ILn={J{t`Ql(SpwQIEgbpV$mA1&yydWB&yFASU_4Oot+F_lla zJv82wxU3e)uz{Ms*C3I2u8dW^vZlur?UmBa~gYA^J9hAMaUWTGB|&)dy@*Gu-ZOYVcO z#x_mF@vc1qqS*;=J~1GXnLXkcD3}LEYI25d`DI{*NA)Eo{dWQ{f?k zDGZ6{F4N#{azeRdIZZ0t&*AH5aQ4rFLe0TZKOe>~KFc>SEXYp1M2s)&)DulYX+>c0 zrZKlK6cRv#nos^SFm#pz+{q_yJ0?i+q*F{G=(cu@Ye>#rfM^y0KIkNhR|xA_4VU>K z>332#gM!Bga_1b-#R3m56Y%NZOB&jPBzN#oO9Hq!?{m+0V(Bx zY|_dxzLhwB>@_M-z*8XwXJ)7;hZ7)BJi8S>`J6**k8T6UN7s^67E%S^2zGfl%+f_Skk=5clFJaAG!VA`utW$ORh8OG4z8E=y&S{1$;K=D7BswKe;ALpFmNqR;U#}x_CC{T zun+2k_)#rSbPkf%q_heE)a0OjV1~A4p7OP|d`2i-n@9Aet_(+@5v$#enkQ7yIuW@X z=>~SQEjV}@Ud2$p_YcIe^`}uJzMXIA6q5doOqFY6L6@OPY<2?21Pmsi`Ny|woTt}8@u(mHxN5O~*z|q!woE19{Aynr@$m$@ zdhbq9c{d(0Q|7l=0m%;TvBI%1(EI;&aDMH~@{a1ODz&e`-tQGcen|d{i+c2hpicV8Oo@0TnU@eBbp+ZoMYS?g$ir{f!B- zjtA#~Yc71}Nt*?rmh@K!&=D{O1duab5}pm!J=dU|$o~#X+GbtmgDE3>#jdOa zY`f>|VD$CWb!hW&8_$(n&>4p5U>S%sgD2*3XC9t)caecaWt-UruGt6+Iq7QX8w(sBK@N-y5UbiR1LY+)+;mh!=DH0xNCn@mp5* z|9L3X+)--cfk&w>H|tST4Hof)0$R^n>5kF(K6M*T_fjTt9@o^6h?Nr835~CIFk0nX zfIxBVT}KA)`Bu925s=}*Ixij+aE5r$EFQ#gsuLe9o?<#D_zVPE%wY6B5lw$NJdw{? z-ST@M3_jw!$OOk#v%F3|0H0G<_E%j4NW^YlXi`3KB^CkMYB?jpjpx_0B&#d z%#5cA1=f7GZumlk7N7)_yZK%*{`Wv}b+qOm?hQVD3=^0bY9Hf9r7fWtX$u=0)J;s(v9Uhs*D3vdMUw|b zIxKucqM7Wy6GkdW8+z%M+;a9py+vx2-v3*matHZ?8`{ zQT{2UDC)KH|?T1!LCV6vgeFw=|1)0Q}YI*M3M(LL z@o~aw=rz8W#|6^UXs}F|CPq>3@so_zxbm9o-+~rFc8g$z3{$QrKiXs5r|&-=e;kEVY8v*x!PbBQB+$Sr;Z?sX=U!qGhH8C?gjhEzcW@ zjDJ!3msbz$M(q{Sc_la6ez~353YWW=-%`on;(;mmjbVRR!@Q) z(18&xk*1KcsQ&Jhac6=@5zROeWM~xOJ4r?Iqy2MEzWrFwx><)YzpTY(DUYCTa8qcpUk{SjrBzCFw z)W|&x4HMCutB+xaH8{<9=0Afgtf*h`#{C1XR>wut-L)28boKO3!J`}GM2i#g&IOku z84IM)mjrE6KmZ^}VrC;|FlR8p@=Xl7R#t3DcVC%vT!g| z4{`(;lm}3d1hpN(p$8Ri)wTKZEDbOrxV=7Ub0lQAK<^Pz$Q)WJ`*8^$IwI2w5=b@8 z4*>_D;DATRsALA^fhiLMS$Rl%{BTk527Er;(S~bn10TKzPOP%BECYcBP(sJXg=%hmMWV*Ghk9>Rdcl=W`5>2iI*H8|Jw05%I0`p(_ zZj4&_7MlRLEv)Ri*JM`*3t#%0p#5Ju+0?Vk4NxR%6=Ob-{w`Wk8tNO$>{A%cNZ$l; zkhA-Qa<5R$rZeieT>F_Sq-+FoGF>+5UJl`=5q;Z+<{2eW%+|y9iQOEd5jRo{!VWPN zU3yl*!t)Bv_&vxTD()4R7VUZCcq@%)zF!nM4q{cEI{G%y)rfbc*`)JMgXu5f8`H+` zT1YXnPVP#w;Y32InDW@~iNCKkOv0VOVcD$NbIlzC-KXyTO^Z}P%C^tz{3mz*2qOB! zdVQV4GiC&ly@I@KW(*RHjEresAMmh8nRetAR&b#G`jDS@IgK3hF`Hj)bu00smB`4v z@wT$dH0LguH4BZ#5>voMyWy2eNMDRbNJK<5sucE>1dd<3gcOG=HyNa;{vJR4<{Ou+BVXrU%yV7KBF}B>WorcxW2xjwLvZ~rh z*vD~*(_xh|xfr9FJ4tfWuR`7pB&aDR+WX@$Z_wTEj1i?^dpKd z1>HlivOEUE91O>rpPIg6AA*B}Ihj#_(8hp^n6&(S@i`{N?TIg}ceFUWkbxLVP&ca! z8arugHT9C#YC*@r2f*F*rrWx43vku;?4rL zQ;+%*Q?lgDG}_KsTM5&fJ&Dxy3%ph;n77r-nKqd(` zP(I(E-8)h3p6Mjb3}a~5u;7|4LZko5uJ0wSEosGlh5uERJ)D1s(|C!_d&d~9l6@dGmn6soQc;BHyY zL?>NCy-dX1&G3z>J#xr;3f^BbeS6o3YH#}IPaO>SxJh2%>QG)tZsaIW-H3Xad+{zB>e4b2oQL% zqCrWDdofzc4QOIm8@=>w=_Q20)9A3OwU>5f_g$iuz$ogEFAU1@;1N)u;?ua z#Zihf^h!`H!T_ast!7573~X*q4x~|tsUz)%2YZ7 zuSGj+A7c12{=?l!yb~YVTHT#)Ot_t!n$cC44W!BU&7DbF?#5c2rMz%+`l!r5iP_XX z%?e#y#yNf)X7X-uSf0o`)+-F5n`sODlZ)B>kSqZpK{!!2q)+(ZD6D|xK{Elb~a<1sI7K#!LJq@XvDaF5Z{sYpA2N}o92f+7U7Ct z`;xD2eSAH3V20kTxdn+|ZetgtX-{bLjG%S7#?TjcXa+n%q3Xrr&S2pWg`~^@apQt zUsuCx+NBKBHT!l<&X0Y|y(e}>;-JV@tht(?UmDVPvto$7WE^@QT}#f4!YgTBD#oZ~aYUjg|1HV@wD^O@sH;Ukh>HqohS# zT7S=p>BM=-19e8}C@;L&B0H_VHQj$jonSTxBht_yj1#Od0JTC-gTm@! z$PuZO{jH$>#$11;!MXUyouq~+Nc$V(oM=U3TC{bqu5H3HmY8L0S60VUNNKwNAEujr z0zzSI#W)IHtp?IOF#6Hgt*0JbHMDNkvPxN2!Wo+(n54u%Y)8 zcK8n)is8UeQ6YK(4Rn{A1XVj|c!-$(4#$L-fCbZ$(?~s0iAjzc^;oP8ndPx2x^YCv%aegKc6OWHBkJiHiapAe zk4HY}61MCs|9HqqRYNOZZkDJI6H3b9v7nLBsB~`<-^i5+&`&xMaA1ZqPFg>1r;64i zM;SyJ^IwfWB2Jo;U5i_>JLtzc+^Un z`#$|$TDzUP;t4}jPLqt_o9L)BqLEYxZ)o)PX`gSzkq^E#2t!O7#F-{yqBKz&GUwPC zo&MegUCWc%U*98XYDZ?b+8hWr;_Q6V(|(B_iZ~!ieG8227`W1@Z)uj%)stai{xw;M zH4{BZjQ_4{n``|1+CU|V@jgYv`G_3uj%~O{)A^{O>Z?6l%Bof|h(H)uh3KCQ(9T}& zf1`~W)ps_0F1lY(X}*d51nj~!jhKfON=uK0zRewPQ&dgYndk35J3=njCqa>Oq(*=@ zc0r!p(z4b%rB3&1%TvSJ8<8poZ_`Hu{s|}ZO2hlwSI7zUNWaj{Nk46~I=aLb{1C%jJ{;?yw7|9N}Iuf-nH{8MNH0ZLU?*%)1-23!(l?pY`Ati}SZ@=uhU z_=kjyZl|8aMnREfXDM|-$0dGxv{jqz1F*8_D2%&I_gfAeuvORGT(I=eQ3l%wRURWW zt0{pOAA_&P1GlPv75nD2Zg8#i%0t+C1{j2#H>?zf@aEH;xDI>EJ@DXQpsXX8A!R|vUbJoUi?Mgr00?4BhBoq24fUjj>SVP~fSfQREzgWpm(YOT*Q z_@b-ppb~8Q%V>Ua!EbU^Ce{if;62qN#{h3D0-NfkyI3q;UiGi>nl-dSE6sDSk-6lJ zTkmsW^!Vu#4s>Q@!#C~hU!@)J%ALIT1r)q?Hp(~qcfyU5CFMbm61pyL=l~3%YY-n2 zFEo90hZ*eEn()+65&9vG#UR2dfe5}@Jo;$oDLFkNS~7_P37#o0&`!N0jqT!DB67^% zJHW^aJL%_zFHAkNE$;d!OXi|i{+uHfJT6lF0u^j3x@S+QP6ntgSfM4J`R9OyGf0}{ zU`xm7I(_XcUaV;JnlYt!P=3&9R)O`7Sd&fMW|CU68PATNXIF&5RL0+wT!Q94Mc`H# zVEjD^I#;&L-w$X^^+(8B&9SQ@L^v|@pxpnI?2_;!gV~uTS1X!KIcenHm}OOcyxrOC zI4%y{oZa@dr|I|=Bd1YHp`{s_Gv0h&=+tW5N+GAYo;=w1zNk)-s_1^iabd#&JZRH& zFpqQdmt|d$leY$Nh5AebFySK4{Sl%SAIHVao=b3iw|GW{iTM|3c{?t|gG^j*6g}%? zT&7g(F}ho|0WmaZt9!9SLzb1oR4s`v(WL^2a{ljrXS#eJ+c2ywWLcs0p|!E8=t05V zql3D)*XpS&$m9HP(>xg@G+5f}>;vm`v#T{=7jq=&GmpWPh!4R6RH|NQJxM%}R+K!X z`BB8dK<6&I^`DhE6v3Ph<%b@138Q0N415<1g@*j=_piiEc`RTq>eMdRQ@~3M9!dy$ zsT!Kp{rnW+{#&r`pLW#HS!s=bG;L0QZAyh!lR#ys1o&k`UjlmXc^C7GneR27l?*8j zSF`WWNO^bl=l|%0fkjxiN@6&EpwL6-=23W)-|wb_0%V$+Tkjv>D(oPKKg=S}2%Qf& zfF^O)@Y$h_M?AxUaUf&gw62U4)2(&MVyYUY_-9 zllfuU>DPO3{z^-9cm>d7TN7tW^_fOoE!H^jLUv6c4EKuHs(PgbXEh8(u;?9vg*nRp zecLm_^Az}`%NR6eRsIb&J)+~pWYaTJHYG1*!p9Hu?R=UZ;CbVFAt@Lyzvr`5%Symg zl09eNM!yrmkN{#hB4!&{Rmj4JT0S`oqRW?J1uJcDZFKT zJN32pLo|9U;#n7hEz>iKn<~ocnShpOId8C$lhci$95u$392pwb{0492ccp3pu#PW4 znG-Y$7Fbp>x+A&9%bG)QAck>pXF#hkqc@;hd}SG7&5|-q9bO+RD5h!Y=RE7Fzqq6heE*eH}NVeWU20ek8$KQ1q>(5Ia|;24U}&8Ayc zMdvNtJQ?sRNQ5FoX;lSUG&5-C!#t~c^aN(@hJ{w6+&A-k+n#^dVIcnjrm3IevlV|D$mHrqd)lG)Ukex_SSXBCRb;oG6Fic=ZB-z* z6l!~Mi1fU)e&zNQ^2uv;%RCZKd5wAVIvMi_4GI&v0Iu>_@cPo? zajavs(_gqWpzU(NV~uef28XsQu^~KwqROA6O9-LK3uEa;&Bh=m_}~4}<~N!({8(?P zxaT*rOO0}ACxfXkmGY1MsH##eAE3?WC;Po-wYIT4jXO0bn{D85iQkNd2$7m9ziFpB z>x^hyQ`0lOWHlYzPYa3Ys0u7+qZZlQ&K^zEykr2}M#s{L%*6i2|Jt4+9Qfa&Oh(T0 z@;G8=$0f;=&uVd}MGC6a(CPKsI{}*j!M)dOBMiJddJ#@Y%V0tlQE&NV?}_^8=qe>j zdsJ@m{r)SygFG){HJU}!3g=3>YASB!YiQc~3|6FWQnwSbW<=fkTRj=QBwDY?%Aq25 zFAJ#N0yfb!OB8*LF>6trXnFc8!8TveB2Q%Y50PySwx@?i$5}{cf~^*N9QYPHd2P@! zLpGG`n^XxEWSAzfy`Ua`0ao&=+vM};v!}OELS!R)7D-mgEqah_av2Bh$|y~qtRbn% zs)V-fssVOY0CuNSXqN6np|vL+{a#C=WUsk7*#$HHt)ik*MCxOG_VaVA3ah(CmP;I} zL(ldG!$?KTX{NUnwql{bhV6bQ17b_0%`PHX#`mL5Xh-;;fF-=1R9PCAFGTs}j2lj) zqc7#mza)Epq|AnO`zX~{V1NLMmjy>z?I#kb>SZS;+{gW*O|;@`wYa#ISxuKvuKq^qQoMk7GT^tF2(hV}^fY;_VV8(!N+Y^#8CtGt z>~wF69u!TspNG8=MiCoXtkZ>2@O-) z(4f&?uJD22PK~ebP3gd!>v1QfW9r79Yed)KrAJCScD%>x%d2rGx3^M)YhsN73+(RB zRZms!G;NA9qZ#YDDM9if`xpRBVwYFgd^?|qnar^Evs`AF?T&iuzI>*ZRSExGiE^YK zAv*R=Wvdw-OppBM%1pdz8{>V|5wSis$blOz*Jk%T*~X$|1AoT<_^C?b^X}85XK8TU6x8+vORsFz-WpUb}Hz zvISlOqoZ>_SFZj*lo}hK>TV7Eh783!eH z?aMJG@_4>F!;8Q2k+)bOoOD&EWM{ecMc?O4l;2l#@9&Ox^W0oLsL}Hr|85c*SV{1DI{9mDRWEHsJqMrA z9SSAmBmZn~>$;_CRHnPBorh5$suEVYgbk{-iu%5D}eg6#SnZt~5$NbyUa%?Gm;*AC2AkxK;MT54;h+7!J=SUWPYpg!$}KWNn>TF&gb&(naU=~=^p_TYY% zNh4v8E)J4Jy<{!CH%~+f35{p(!eA*@YmwppNeKud0x!zC41Vj zMr3_mi{lc_Frw+9AclI>dz}cjR&ur6=$XUinhL+^1>MnichVDubtW$OcIDq6M@6X8 zK3Rg=XHghb=`M9vfbxZ$3vT=SyHo>&zaZlPfZ7KRW~3Ta#hK3CX+`^Ynp~8rT=cAN zI~bc_1h(_ZP1J~Db4dmQQwcF!h+R@#+cjOe+kUq1l8SkH3$^_%JqYm_TSk16p7Eq1}Quxz(?P;u5+wtwK`6rMQ)i~ty#?$jvXl2faUZc11z_oRu$as@-r`b2ljA`<)LE>{7Xcg1J)TT16iAEUh75 zlB8TAuLAqAAD2cDdunRD=GyjheV06`S{Wh-Hg`~&@}pW^e}{GoaX@g=vQq>*S%cL! zTPGsd|-$yb%euR$p<_&iEUnTT0ggDe zm_c6^9v|T{dybVC@3g){Oq~5b&1biFCrCn!6GKCtb}6_nbaTmBt6lV)@!~Cq*$B*$ zk&)Ds_mdN)#mdF8i$7}&SIMWe!Rl{?S3N&UbE| zPRC8_e;v!fR~MfT~@`A_9t$LN_0w)s;9<5^aUfSnMEp_@VGo1;bdF|@s->$ zF6oH+D%pHXA?H#SjoRXirEGJ09?>=da|yGXq0jND4Bq$BD_s(B zB?P~(A#9LTgl#2j{I!w|0kjHZr*$RpeaFKB5vx-xoAij?+karz{?+thi4s$Tq=nE% zh>Ug(yBea`N!A0VPdcmtU{;S1e*lH+RY|B$Ys%Y>qKE78%Sj%XP1_zSX&BgX?Ze6< z8<7tl;Zv&?-;8p3j91B9Fe){ss;d{?WXUBblO%}4QTv@;OS?Nx)3&Q+_xW;L$IZ+o zw&KIT{V)r$Rtg~Kc-U?AR)2@l?rC54Y5zSxU-uW=(5~hsz+gr29#N$DQ0RlE&NPy- zq>PL^k;12aLX~pPL2g!4sd&*-6|S2WNHYyB4TggOBC}0#t9v&hwuIqBJ}$){i269E z!ZkAGfT8jh{j$`3u-0mnQ?fYnQ`hT>s})51a?+_+`xjc(z0cmO!alEWQf9s^4q@r& z4K(yBptDaIdxGeC@pZ0&7!)v*M)VR<5C+aM<4cwp!3`N0tp%JD#Y@dyDz=76?9Pv+ z7st3nyu|13Chiwz2USX=cwNH_#^KRA;*3TXj+2Z3v2Q zN6Ps|Hc_G?CB1b^pVz<8b*2J22Z+P@al`EyL~_7_s@}L{t!%sZcB;c3aE3+LV=6QV zOf94+l&sz$3~AFW&9cNrYH<)v4RVNH7Y%&}R~&^A7!ZsU#N;7-9VAz@R+ELdTGN@1 zw7N)shK7_n%vfw1)J1he){a4W`+KrQo?9V2DI>UajSu7zU9VW?>TkoAyD?m+qmu)C zHU2K~7#fK;@$0sup9~~h)C=Aw>6CYudi1G6Y|VL$GsE&JE75`U*-a`fG{zAQqCp27 zJOa5MT7xhJ|DY>hLekdk3yT~!uC^8hy;Aoh*$mt1ZhQ=N9^f=JB9w#S#L`{>m?N7V0Rp5JC*RQTuO65eHM8FbE1a|KI1p} za%yw}njV(N0du5Azw(aoNY&A8nslwN0>IP_`25(#=7U+?&#*Lj_mkdwlBk4Vb32dB zpg;pRv<`h3*%o5+T@6e4?;VGpYB9N6w{aiga{u%%rcDo|=X``NQ+t}jwpS~p;d(b? zYcA+iPuajyae?jeJt(E=5#~dcCtd-GOD&dsrnCFE0W}I5^2!7Vwm;;2%L@IPRdz}T zUWN0Bbzik*ItWVvNihk6c)o}vJgx9rJe5qUW&<3ILAD3_#e`8j(Z@cbvIKG~Etf>p zKNm0ESh#LfbLof7Vu7!4{6mNQF z`OC{or#)#qb>}5|b75(SFMC>e9Ny6m<;NJKa2W{*`6+&DyOVDr^CY7uqgEHIb0JIC zvN=V3Sz1iS>(HH22ojaWmq}k%IARx+(U4(2E-?JqS5CojZk1=qjE=w2>Th9(SQ1*l zVqQsgN>>IfprWT2#Gtlp+PAFV@bY;(dq$8~*24pndh~ggmGvd!22!kwOw+hO^3QI| zAd|xnapTOvp`FK0ejD4vTHa?S3IR^U-(EWw0GW^-oqPd=Qz!fui}9~MEy~NYwscRt zt`F$w7*QkNnZ1&Df93pEHc4oK@Uc?Yl~;PPVCeN zU>Q^F`1mO!Y&VF=TxO)^I^^CV8SswW^Q~U>+RAV1*`>hH`r^Bg9A+%}q?osN4?4HP z`=!;!FzB7{ddm;JBHv!qajr(-Y44cFCJOhPw6F!X+?>O9xv7 zN|r=^w3pn98m1Q145VF|Xk0-ukQr#G-+|F}l0Zm?xlvDc7AhE3UO`@aGDdsN#f7F< zKrbT}8bv`@^PsCse#SEn`ZcUy;`Njlm$5rc&LO+8R*8aZ)Gullz0!S+tS~ox|2{h1 znforMbtvXd%DDJ4@>6_7o-+%&kIWMVEHSQ$z`MivNDQT!agy{K%xjc_-6M6HFIU7| zbf58k)6$q>OfA%Q#Kj4<3C-Ed}wnc3oU-QpmScwJpp=n@p7kB(?+aRZE!_d^P zhQ_w5854Aa-O)Q=W&1u2ZWIv7+Ya77?R=!7X?C7opYgPGj$psK8Oz1%D)gX}udU79 zbr|Lxoxfx74wb;rTV<&6a7Hv^5yW~t!BLABSyh92?TT@!)4hDXPM|g{uyODu(kBRe zW4~r;^{T0?NZKnOPmL+AkeoKEvvbVjlbW?U!1>l|41+>@lQJCeUG!zBrgj+)#)upGS1+A zk{8g0$f<199fw|9YvASRi)P|6wq~|5#L3{Ib6-oxgV~IaJ*sr1j2MywBEfP;!3Au( zcE>>q4cL}*j?BcmB}wGG4}2NFox~Ca+me0^le#ZU^`tBX9J^_o><=Q(?bpMe_gC5- zaGqfap2pP7cbKzrRlgqfcd8Egv(1ncY>DF^cTbKlXVtvdcD?Ka*WO;}JK&W$uUg&r zia-)&nlB?C#c)8?d_pAMws`Vplu>)7cLYvFTf^Re_sS`w`a+9)QxOB zuf#4*ntkwOmD)!CCjEiD(XU8j|6V-{KfGw&qe9;jc7n^+PdWtfF0&RBhC9yPr?HOb zM^;lsycZX7Us-o>uy0G3D{h7V_@s~1rTBEzh#^0kAUlQJmRIpGZf%scUjmt2pS<+g z+5CtH`76t^idme+`Ec%7=eE*s*6_LG(vVzmgoW}*?8-Zt#J+H)_o}O@p`@o?8V7up zYK)Yk(+xOG`&CuGUN3tq)rdyf6pIPto?ByZ`XON%Ta(3FUDldkGhfR7<^~)9V#L~K zr#jB*@2T(cHxbVuPKn|K$-LUDW;GZH9M>pzF`FGe%|`L(me>W5Wy-P-gi}TTVVq{MFWm{_9E)R{!+DGsFx3#ya7w zu&?%$OZVt}T<(Avr4+(;>m&PeZP4L!5090lu5GLQ5XFw(te5HUN``}X1y&zxPeCK891fFc;!t@t|ku$Hr({)dWakuDh)pHs*GPImDW3~^L3S5BYhW1OAzv?APrx5 z8;-{om;Ll*>>#mYdmweg%c6>P#(OtpzdcpXFl0zOTNn67$6d4IwkW_GRd{;JkgM?A zm5L9$z1Fi!;rX}`z0xNpBv1*d--|SXn^(B51#I-|?3=FN0h&%kTf2?~F|FkH=K1CQU>h~L9 zgh`0`AB+3r|KzZpV;5JGp&%Qp(X+u2L36aupe_QQMn++xZy8yv`s;~ z_-07_aE(03y-5UaBuE3`FvI_~)lProKR+;Web*hSb) zEsc51e<=c<^j_Co^}>P9GI@+Gy^Ds(Jnc~Tf@$OMuu8x^bjk2Ig>I77%#@UXH@?~g z%`ceExerC%`Hr)-v(4AqOJ@f7!%n_@P?QZVqIvkcY9ss}6ve~N!S>@Rw#E_laj4Si%}qOl=6v3EOl0sb)kQwIr@ z&8ih3vOsClYx@iU*dQSs72cPBp!YpiW@_#B`d4FptP)8rsI>UexBwO*S##bW>82lu zIXbDV7p)Ui*$e``W2eSxUO2dTGAmCk=|%W%$hmJmu?yeYAx)4yyHr9<2miY3^3j4y z1BnzRO`0x!o%@NDT+8&=XlRs{@%b131r;*bjFb?_z)Nh@V9&}aInsG`isi^zew!k3 zM!z_cGWt_%m=F}IWw}P?GMa{e&VDe!T5$jE#b2AZtJ$q->5f;T)!(LfpX{3U-DXTz zllW9sIFRb^kg+A$jQ%EjZ6hsdLhY4b&^PhfDY9{cAx%WreR~gME)Y5&gXvADE?ATG z@+IH!D--wK$E!97@Hv<~8=m`OfMqrPh`8ZTrryQdk%h9?J$jUgx zeXcq;|7)I*W;{)gRsW*1E@~k9gIZcj^pbWp(K}@gl@DQ;L5J)~VKJF-VqWupZKwQo*%sPSdRhrfQ_^Mu_~F!nSXpNl#@bA7EQYaw6X2g{%w7MA}rzs zkELnOeNw01$a^3ch(TRlmnk()P0iwB$C`>dTPxV->vuaaDNGC>To^q zJ$bi4nJ>jH%;fh>8UcSR91~Dd`L_vL?B%TdJ?35a#7~>T`eC8=IuUv6x(zRF9CAIn zMKgw5FeNUz)WAs`BZ?Gx-j7nbIQ4zz>-`%qkGs|N8t$-?wN7?GlgBP`FmH7aGjM~9 zagHR7?}B8f5obmY?@yF}+NXP4=)!nH7@fzok9BVhAvlT(TciXP1xpATmS1}Mm@;mZS3+_`Q7lfDiSdI#J;KON3c8{CdaDeZT>M9VFFp7_r3*7e_~ptQP9nBF&?jKZ2a zT8DM}XX;T`9#t=1U`=em+2;yCe~BKzG@@2>W}iQQS=6Wsm@hF)^j0DcCcqeuJ*U8i z)quy)e!JgQ6eX78ZVD)6hdwEKO&*Bn?jJa&?1#p=Hxk-<8im%k7*l`PBE#Ag zSERNX)z$TaD!Ws+AWwsCkxvZb!gIGZ z9L8v$8gd+7oV-X|0-%k>69V4&C@!?!iK0c>NSI?TUq_`B{+sl)a-EE9Jh#h7aH0oV zdaLLsk1sENFnLZ6Sm!OGlV#p+*1VTeKI?LUQP2q5yaDS*=Pj-|{gX4dbOE{!YPIa< zI;%5}YMjfP<&tdG!%mS z@g-EhSbPY8LRxJcoQZFCyE109O{388YcTSAx}5mMn}_8GVgs8p^qszHzc4NO8dQ78 zKseHpV`SE<_s~GaX$8~LT16SCqHP2IhjcGcl0!cENv#|mMN+6hJZ3e&1U+CO2z#(} zfj1j8FVS|`*&LZ4EoZ@iiHW)OUj-9DDdRpqk}sc|F3D-t)l4d~+OUbIMqDvPt!#>h z&Voi}!Ux@upHN$hhz}ToT8CTh_yU+qA5UNg-@49Y6pM>&EA8xs%bF^P1^YPT1O01l zPM#RjmdL)PQ6~GU^Pc=ug&>BWfyORn!GVaQx^Xa_ z&|+QNY(;MfU{R)*U&4W`faFHAME#YL*3{&ZBZ&D=9Blr?!2>z-Y3RkmA>dcX!dh%I zRVbB1PUiUjm+*x0#vcXkM24O{v#U!h>5nRzsQG7+Oarp{uU*EOrcHm3xk|ZwDuk}# z2z`7=jEC-gwQlhXGFCeB!golMc<&T zSJ3a-drzchl%deS@@RJ^M=kH$eAp`wk^#CDjKj8XP|iZ7r%5B_B(|V<;Za&!04c3;v!ZZH9 zA`Xe3f!NQ&{C_C0Y%NyD8L>DJ0i@`Mz|bKev%fJg%V1R=qcO?4HkklvE*5e|JidjC zUz9rGpq*Uxk1`U*>KtEDpj5LvjeRripa5%ZRK}RSg~ztfK21Y^eFpbIUcb4&n+2-k z$3~u3B&hRLH-w5XhFC8fFPeU`x8p$y27l;j!)GEp^QbS_gZfFLyn|>=B&!!BzG1e5{*j(5}9=rhL)61PcaMRHq+d^Y?G^0OcE1MAdOa3d_eWEzc z0$Pk==}>l7`hPcPT_~^v^|Z@$TDxu1;76y(R>wXV!)atXD@Q4XdAY%-6%Fs2PDy2% zqJ}?G><;Ew9g^`~X}Pldt)e;qxqXx;PAVJftpTw|@ndzta(jxU_dU@gG;ay}I1Fe{ ziNi&V!X>j8KUMRajCq3YNPYtQGtiZ(g#WcH=fKRHTvWu>=dHbJUq_rPVEL!ZOY5tQ zwj&?!55IX3S7??_sECw##|2)tSgXe!qEq1bBS`(RZ2f_w0QfZ4^zv*jk@P8_T0;>$ z(=MF(kjDetbDw2O?ds8gnu+^WTGtMoH`}h+Xxo zVwpQ8r{~21_VU5-qVH>s{I;I5t8Uss!^~_Q`sY=aR7EqEgZ&g>{R4l(z8X9Vi3^Rc*lhs7$Wp7DzYOStw%lRY_suhpKi5J%F_;;HlvD(u^S)Ii_&49M&j zMZ8Hmu;?u3KkG!VfCI;Z9=`bSswyny{eK7ZnBF?&ic!U;%?RGdVv$UU^PMk3r~9;_ zOp6Qw{b8WP{F)`}uiwRVl-CBe?cH4_?qTIaL5?xSEORdc29O>B3D#{D881h&xBsmt z7*zu?9L+SQbZT7Sj?M0ETdDfuaWk}fl|0rc^;eeU5pkF6VC zK}m6EK#}Arfi2>h45MM*xZqCFuMuk#84!h7X0%+i(}Ca`SZh=AO^(xtU03i1hJfcSva)hAPYPAJ8;Xh??aC+gr9XbFg?T=KSF6`dy<7`N=cx64 zwO3JejrmUAUpPzj;cdud8oh_CUq0dNwCKK4P=)}eiPSN6l`Oad=}!jEpS>g?j3^cW z^=PZ?0;l)xiGoOXbZhX8+gGLXw=axT#hw~)Y%wO`z^z<=u#*rH5_EnJ^$RVy1Zk(- z>B(JxI%}DK7GO5`ag*0oIJXft%DQ}8E`obEpf2nAmBHVr`;M1j{FAeiD-~3v;KYf*IsoE9E792$v zyaUumlt<G-#O*RUP#-2II3fobP z$1qz@a`WVNkaD&ufsrV(_lBCF6mv)GB^J@?Te_BFy#|%{Xl6mXL12xwI}xf)D_F&o ziBSb>R0Od(6myGZdA{T8NQJf1uWEyeQF(OVw- zdZ7KNC6mn?f9Q@@ma%GRg{3ARU)lsACMM6dCFS3K@_j1MVa>U|_B+2wIs7qHP*%Gzx|@e{vC*!`^#PSpFbA`xzcH>{o^Dp3`7C*s zOt}%w{eDUA6~;{H?Xw|G0af`=Jfc0>6B_j5ZvzK5Yxun;6U5fA6@BdP`5D-?Hpk)@ z9Y_8g{P)6>YHT{)cl^Kqb|;Jg(d2c^uk|itBTq69H8hDQ8mnJR6zc)4jzNf1fq00&Xx|+yJ_qq%?>#YAuViC#E z1c9R;fV4wdnSTOxq-XN2qFEKK0dPfWu29#3%r z{mTWlvVE$=jw4sbQqQ*cQUA%G(?KW3H%HmtO75LF{!+X;qw?}1<{Qk_$$7c&s_&g{ z|GZhSq!nx+I3_Bta91fas*Jv`L*Q)tbA+Qt;1(=cGCFNwNr7eqo5Rh`?eezyiPOTW zNNz!iT=`oSbmmdY#qy*7RPT>-ul>j5=S^KNRkh%JmPp&*2SqJ0#%F-@(6}1*N^7}6 z@I;13G6Wn*Q6-{ZUf2jk(v4CyUM1?*!^G9!s(dfT=_`sDHH^pLpjAE#bnDIIKp28=6- zsw{>GBAS$fK9UGz;@NSxvs^iv$(c!rN_Ybq@QgE6wryJ4+qm4R%?`UqGy&c=Li^ov zmwr-i(OjJT(Yj{=1@hNZOA84QwfX@FRVbg%azTzipAJLYmjs)6MYY!}qoURHYFMBS zHbDySa$tgQ9H#*b+c%9O-P$Z*!pss{TTy;LU7ru=;2ABrDwJjuK{_k%Uxye~W&+f{ z^JbVt*74zDb^1%j)K?Ae0zZG2PTN#rHfX)lV7VcwdJNw9;48uzVnrv2UszHk6pyhP zWgP;%H9+;psq7Cp%yDss(ZqjF&WhaDrazl|$VaPF^G{@{Y3}i~>1jqbpV3|d*jj)z zL(=-_;)gA~j~9(1Dl(Xoio5OLH9Q1Z+r)ur=3AE{kk&e#*YqebaD+A~_E_z+%mS_1 z?+T6ozjuN@!zN66TeJY-n&`)IX1aAW7v=*7-kzzy&WAtVg5;y3m6eD=mX>8KYmv*{ zZ6QO1Rrs+w*sgy1AG9es(c5@F4QmpuW>QtlaMW^}=hlP1$q(Pj=~bU}Zxs?b@u+0} z5*x$6V+9WUPQh@D7K8Q{DhXRY&@fneN{DUb^Fc*U;gFzQ@AmeVA7s%>Us&y^a)jc?IDe@)z%-&&u<8vCc9s zpE@2#)H4s~&pE{oY`5S!Zyn5nm5=$pYEaH8RL{^+_ulXJ3*qxyuABr7Ls&4^!Y%#> zudCsLf})AyDRL&dNWF1Iz(enoO3Nl!>dULt;p_eFEAQ3g^b{7UVcKf;#`SGOf-~i8 zT&WI=GDaV=eyG2!qh+1u6;{fi5>}*rv=n|srFqs_JO&_>WkB?LTZ8L5oxBoD?6x?l zzPhO~MsegUy0fo8>|OAHw7q2+mMkfLvHOVU+Ty;K$NEid0_6W&IF;6+XEy9Y4k$vU(vz4M0k1>{= z-O?4DNauW7R0}c-r@SVNzp}HOBSpX4oM=HsFb+6>XqbU+m@VpBLh@?1UER?9A9sH| zeb!O0BhVdhL!4Qyk31I6RCsQ#@&j&Fd#b^bAeZ_rWE|{@y9Jh1fR^&ki29>S@|<`p&lCxADr4&q;LlR;|U90GRbmaYvulBJ{N;Xat=pI5E;y{7kHhMD>H& zn93Od#mP+No(AK?A&@95Bxr@q$KgGD^9 z=k04*pYquneI{I#k|(>5PQV7Bq1`0Z?~PL-;=P_@YLKEzvilEQU)X(m zV4rjN_tw0iwAP67Yf~8|5ZOQ6VfsYbDd3ejIA%RUvnApe2~IL(--1-L(-c$L50T<^ z{@hySZ)k@F>S&c=+d`fG!lEKur%EuO=k4b*9s2DQ!jCS$g)G+qP%LU(iQs^iX_4{} zy>?kr3Re&N@bfZBfiiP&JXe$F+C8Sb-bHiSS?>t}Gdkj2dxN?D`+d;Aeb~{2@%<3g zc2^hkh>?4g{ouztHX3*?|J89C5Nv$q{BuY zt>Svgqi$H>-$b!3>alyKD4vNmdnrZY7s=UoaK=3)F`s)0C{L*cW6AD-Z^`Gee@O z=Ks$9`u_e`%OLGz%HP$kt2?!AaRQ&s;Eqa&df<_T*anf#K#tM&t`T7YK}LST{s%gT z#D6#lMzEfP|}?h9__)3iF5D(zqa0}cMSICT%ACyUpVTZb8gUt=@eaD!y4 zbM$j}%4PoUVQ>1AH~O|iknvuy1m0jCvs2+*`9pF?z z4(Yv7L|u|oD&X=J`ItkuQo8Ta{eEr9jhA600 z=4TOT7MzhTx_^cq$AcenH9^_WeO#XqPe=CiE*epHhm%i(bN(>pFaVDJl#6sU!nvL z$p#(V1cO6xmX4K6V|CuVtFf;s8k@7R<55%m1dmmx^QQ&bjm$+f-9viHBAu}QWGx-o z^0y1)?qIdQ;S&bl^Ji4!ob|oQ<$y;jnrU#1%VBdOvmNqHw?!Gxi!Q(ETMc_{Qf|j;tf8$Dn7@GH%1k=~DQYOtlSQ$ivAs?_ zlF9)As0*~3P5x2#$viB7BcL5jfyc$vqp0sH7d(#{zmOHJ>XdH-u-^XVgE#xf^xIzFxIkMH6RCvFCnmd{!7`3)mPJv=$>J?OFw=8NHmvSuw z)I@>HoGXwF$d$)_@$!i3E-LTKxg#W-sXB9*_#I)kWHgh?3r+1=m{|MyZOLrYO94{u zub=%L@)yQ2)l7|ydcjkrSI3%T7#o1?Lz@OKyi{vr0T?J7pP_$gGtns`_iyeB_eHD^ zjT**?(=JS@>Cb4)_FC99Rdw`A`m*Y11L@+@5G4c6r^1A{rsL5*^h%ZPW(spSKdt?H zpT{ZPv*6fDm~7G!0rcCq!(sh}27FWts9UL5UVD^#l1;|tb3rXszK6rlFOkFX38v9+#S6*9quZ$0g z4TShwZAc6y+LZ*=mQ=I(zctz0vj|ax&%sLd#vslJq>p z)Y#JB@tNfoRAlT?dw1$$p~%+GA3`mNpCBy5)GA6%l{pUREwgJ;S7Q>|tlZO#dRL6u z$_>oliMo(Mr-Te5*QqL_-g}Sku;B_1!8bo??UYx5?>vv2z$2RLlbWvgI0gX=vm6TO7Q~UA(x262p4h zRkSPbg5fDINc#V+9IMPLq7=;rm)Kin+g1R3 zF6o!Jk3m)Bc=}s6It|{wM7?7E!qM`{`E6}rN&)k$<_aO8Uo-v1x2v;J8b5~tDIs>u zF|vx1yYY+WXF6V-lS6HAN25&l)g|c zRWk^Z`6^pb zP9iSn@vbw=Qfc{^^J~A+mw(FS-?OX9J)AX^|A=dJdnjaZ`h#I6vr4UePQ}8LdBM*e z1LdRRfrDF#UH2qz*dqpQD}lP)qikLA?9=N1+nn1QxdmkW&)PYRgD0>Xw1QVPHvrJL z;%Gzgn;~OR`x)hExuPWAnaYp!Z#00A>qkr{MqE+LihU&nc@LqFFkzA;9;tI6Ic1d4 z0NX8kXU9gscGNENX1q3~(Gt_f=HExFCWi~r+@I)z=OxDnws%R=Utin9F>miP#Ogqt zvew+T!-D)99}({z$v3g zVmKq5fGy#S@GAba#5vsqsbM$A5~}MPcq*8lXBG`sCcgcokiUZkF)la^A6w9f^P4&) zy>ntIkA(nAcZltQzWIb*1(D2|T2=+8s7O;nu0Gy_rY*_)O^1?Tk4Zf0|J6Ny(Ecg8 z4c%X-h>Vf2RU%us>3ep5excrcPG#Cvs}dWp4az(2T*x?O1F2I-&XP7SZR0&q| zLI3E%Sp~fiQ(d%NtE_g#fn(+9ujvmos7Gj%$e8`G1yj|G3+%Pc&|UPAs9F4FwX6kx zVj?63^)KIZ3km=FBLsP?#@=_E)z(h6-&RKsW-k{^ra23MJGy_Me1bcg3wggBcGM=o zCaw&u0rSlh5Lys&m?+A=E~g4FYw*?cK4JB(DFKD@_*vSXkfqee_UZ;2QMLNcXGLeq zaUBd0JyG4nn2pZSuv3yL9Ucgif(r4XmI4_n(?=|4$#kCK17 zRuG+tT@}~T`GxJ2 z#hiwg(7FY&p1}(2Af=~@9%%{uP7zC|$VjlM(lFSduidstd$zkZAxwKGY^JMQXWlzAQzqUp!j7meMnqh99|nDKAXZGHJpWT7mN9{o;y!GX`{padK%zgv zD8s0@sCK#TqCCm`{|LP_q+jg+em`s2jN+>NFs$8b29MwJJ&$RPGDoj9KVW>^nMapw zw%Nwox+2J^e)P#($@Tj3GM@IQ8Dxysu|ePj6L_8~YiryB}IFW?XkPf&G|P{e=YUx=*xZAueWC4gCBc%-sUDd z@`z}b&n*IPyUU~2IBJO+(0n(pu`z$}JF|{DRMYqS(4te!E_0Cx`0p&d8l*X9Dc}J# zTwjahC}Dlop}LVQJ;Qb8(jvrpIocasT=hm{XUw=VhF*teD(PEJ&eXdJ{CmgM8|8at zFYt5h7c;?eXE{^jLR5D3hv{(h!swZedy`SEOry=kw72)lQ2wx7uIN?R^R}JpgP5fp z{p8$=87nA~>7oT8(-bOq)`x>3RD%o6qd0hIzJZ;0Zz%GkO@g*w&}dNnCB=jp=zMlT zHtT!Kkn(jcWN#wG=?nRO*eK&nD6@8WF<^%nIZK<(-emRf<~1xr+vU zvfbTIC$_^Z%>Iifato~gZYfLkqC*$G8R^dL#*2Tm@ERSoRthpA4M{%XRCl#M;pQ&y zti31|UTF7?^+{^e2_AkY)n9#QihNy+2nFgc7B!wWKl?{-MXU@lxPJop3Ea@#gID`g z-tHW3m>~)^T}xq<3K|_ee?v%qh=~~ITFI>`PfkvzG+GFwZc#)tSzfy3;wl%>0-l?V zn4-CgEg%3AQ`JN6SgUUOC#MLtkFR@>`z-3+D{OSf?1co(%HQL92NnDAPI4VIN6M`v z=W4tv%+`LhXa`!tJLjEN6y@kKbyk7Khq4K$1F{yc#l=f-u7hn$yBar{3EQUWw zpRwDc7{T7BWhBQ~vmRskD%+z6p5t>N(6tL{Jfx@h&$c@JPy>$^*^F9KNi)Ep~EHFpU6-R!%>vqrQPw{45g_!)_<4y@Tno&y{$yQ zt9~$N$Lg56F>qz!S+DSN3`<7B{IU!Z({^yb>+qRdks8?Ez z>CFTyhH!E9cQD=$a0{u0tCg=tM`C*?YDS&F#5zGs0NG5Kb&z9>TJ&tcp;bgv|L!XJ6Eez)Bh1*bNz||aC-C$ zYsJIq3*#SdYKsx&$*8bBtW6#9?@rK9!Y7g0J@d{;gwK$mnv zkEg{Y?v|{E$Dum6s4^ol!P-<0zw4;|zN1wD_AKaRBU2{^%{@;-)sFUKjG5U)RLZYI zfIV)MY{Q)wE9BVi<^k2}${%f$-4C0!)so6EXxW>iZvt$`pbgh=m#a1SW;r42ehw?v z(Uc%IaLl3f3N9)7T0j`nR>4s2Q$Zpm^xMmZYZ`Md_{TDkmjlCunF6{)TLAn!?$V*3 zy%&vmONNY3O{E<)YNmGZyCB#o(ObF9@soj&LoBo|+D~>K9A0epsXC_#JRzQ5sPFYu zJ5$#%*^-60MA5_%T+)3h`PzQeYn*h_OuHJ2Td{E7RP(t-#wkHo@zDh~{G)~RCr6#><#AtV0u__dJg>^gORMRCvyoZ$m~Ajxrq@<@`@)3q}w!M8y)Cjm+KladW z2h>5+nL9C^zT|-AKG9LoP5$Jd!`OH(;BE88cxhG!)Z!$|axjd?znQ2iM-(DcKy5MT zGg0t7>Ifx+m#s&>kX=JmNW>iOzrLFRKd67_L$;shN&G#s9pfm9ZWqRCHRtXqB5lY| zp2d6l0Eh>+VT+WsWd6XTDez94!qQn{`~uCQc2nNwA+*7lpm`u|p8-nubiMz=>>|J$ zLmTg*YY7KZuX!5D4u88~J1ziEVp;b?V7YN;xNvu=aQ|LE*o--9#ar!l;p<6>Gk-|m zwc5URzggpawYaDHuxEv-LO*I-23Xxi!eyNB#OuxJ^sgNNc)NU!KgBAt=u!UTeT0@$ z^RAh>f(6UnuT~~F-iG=#&85tjw(O=xU;e>vz53jvaL3<9^|0KoXZ`Wn*f+P6`|kt2 z5KVG)X4nTQ@6N@n#c~U#vF{FN(W;l0+jc(9b`PR0j(_Z{_4MJb7T$O*15p~yfvQ{| zOnY(#X!KW&5BAU2q3AwkINBO4cZjJ+1@}#u9Jts5C~D0I2D8Hfm7u7bBnzhCo=Jtc z4tQ%l6_||Cy5pm|Lgs|rB61qoIvFW+p{mte`a`UzruVK&sK7J6&YGOcf)C^yeiM?s2bQ zJwhjvCN5Y2^O^foy&grVPaAaM^AYc1?>Lf_j<=q@|24o!mIsnOmNJG(x?ro`o9yeU z9f>b`S98E5sB#DRr%#pr93PWvHuYQGtG^yMaNscPgIV-gfzM$Gf%V zIVKzaEuUR_S|W^&tThgoppYy&LDmR!EdM=))EIbv*X9jztu0MB2}Xw0r&$V4d-ZS*G^zm;poM%kIq}Uxpnh7T zKU1zM?)LO^-#6aCgQl09t|}|z25|_C3ZwELjfu=A*V<1|XyT^_RU$)kt;!d|!rrwK zUE6E4I7}}ib?e-ClkD{(deXE3ZsrlECa)#OT-Msa%h*!OErL3YSer;qH%UOD+E9Mq z7_?dJQJy7XcA%24D*g`oS)+X^R}G9AW=EYe_&HXax?qF394ZlH`uZjh8xda?V#vTF z_eb!_Zz4PV_kBm(Y=QLS=qAk{sv{)4Hye`pXUhDl&aL+db-N;qgo#5mMh<|1=+eHe zI4fU>&d;r6<*Mr)wxx-8 zFJ{TixpU}XXL`NEm(GmqlItKK!|E2ufsv$7x-eh6??xqMNU#EGG72g-J6d9{TW?D= z^-nw7X!>9las1MKx+q3_h}E%gk|A;x<-C`*m8_)F4w5wgMVU&{ck6Aij&q$s9j7=1 zE7`v~aA_>ipLBh+@0eOsM&Z?V8rSOwFYH-uyxT3;@FkXFBlR%+Lz*!B z;e!DS+qIBlbmDT;&6EkS>1>t#nFjr!qYe-SyRdfTz~j8LpC?dHQls#2!Y)CQ^IlcW zKiiP1V$kMkIqsBhxm_??gJ^BT#N*g;Qt{-oeU&7CzaPUyil2`VV0vgUvqmdp78e#5 zzj&ztt)rg_w#Rp;OT4Nts-!qbC|B=}yXFT(w~$4{X@a2ZYagV6xU%t}I6|MvsP7^J z9c%n5rr2{Lg>;r0Wb^D)MVHTTnS5=NGpEVQCH?SL%srEzw~_Y~ut zUMkG~#>4a`TdFrCEAdK-8|1J+oH*66y4XUnNV&ZRjL}y>7E1>Vv*E|Tc3wgjX$4u+ zzz$9So=j*>KGTc!yBG-i(Ee){SL_Uqj&K@Pb7o$~fLr+F& z(@2iH&hhOLH?O>tlK`O#VVu4}SV^E=-p0FClh?Fklf@K^D#Jsv?$7y(3~$@2avqfC z@}J#mH=yF{cy3?j%ST@?qWZmnDaQ_EtW8@$75891ZWRE^uQ?YN;iJ#FL_M^gqL6De zEmS-FSHUoaY5!@NxftKjh%=}VyO~7`-^sB}{r0DsAuN6Yy}Cy47)dEZ zDk|>DDJrmNMFn-slI9o%-C{pvWc*OGfBykdXqFc-F-cI?wTGOlVX6AZmpH+nn3*3v zTwj0Z*l081m2vIG@RIvicrjKmHPU&b85~F|#->U?w)K4(}eFrE9gya$P1PytO7kT!$|I!Mh6F z)wx4k~sqtFDcm_1D^}jgtBXgN61Sz!bP2BszEp!y%la))&GVvyS zr0odLOBu&sr0@N8a3KE{k4S=aHoV3)@dKa8Yx~eUTW9F#cH9N_?By^RUZk^y^99PM zT{)OSDt31&sH7IhN*E%6e!o5J@R_8KqR1U6c<3u%^C*!S4ZXjF;;mrCwFM6z|#dLgwDjj}68;nA{EINZ`m?R(upc>%*?>Stl&W;Srw&&z=$!JNTH{me5^n&nu>bKswUp+ z_CksjLRz3R5N>0$!*8W0zHqhobrh}ln&4rahtj{q=e}f)+dq;TbnS1`*qz7Ki1Weh z5V)_k1PZnBVbwzlNz!Ml1Ns zNz8J$!vh*z96k2Te!;wyEXnP5knP`D=9iNR(JODg3`Vo{&2;Wl^*e>Bz z!-ES;0zyadpy2;>HBr?xB0I}ISHg7M+Y#T<3frBTdIgDR-)vlXI%SD<{S{1h|Ruk zU($q37cY(8JGQ=MS6kzhAtEWdEGA4cK>be;WgK(fGa@*nIy&9YFa}F=K3p%!y|l0- zOu&?riF-&Dc*K@gL<|>97i=NOVbe1Uk#zVwB{_eUOw4n9vCGbVtO$XNfJcrdz{o)h zYxHnIQ>g{7nVi#MJ6N~KZ0Yx-pVex8;wXsHn|ZLa;BcHpUZoeN0)Jk{IqZx+rreBr zeMc!@9yB^q5Q4KhC=EmV08C?MFljcFWSm&KLI^pDq#*@K=c!<7;pMBIhvo`0+^mu# z4P=K)rG}(rd*wO<(^}e-X3IZc`ACRBfNGkByIvL!Y!~vR7D4SJ03;P)L{uQKtUc8U zYssC1MZn<=OmIIxI`3e2GA=G5cSvJbRNgN*`8*5BlzqrBL}kep_H1 zG8_NPXu{%S=W`U$`*wPEJrMw%3Y@CL$6t7aa_^I%eYwGOfC{5*WA!{)fc9GthHb$Q z4WqW|-QFApBaC_@{nUY2kFnt!_kMF>wE(9$A6GjWAj+Cg<(Q{b|A!3@<7;lYjqv*< zU>p>!sbf3NeK9u7?rRa@rp$+W&-FeMgFKQe)7xAGOl3QTA`wHt@IcfAEMow*)^Vg=ja1oa-?3dB~EJ(To{Ib`Z{8=-uH(b6BZrL(={J#+z3&qq-O3TZI<{!nN!Goawk0EPVcd6fC3D^7PC-%mXy~^X6ivowcV4!m%T_7J?M7a`N3lB`z&Hr#kj>6|=}Tr}gWU98 zqIPC2?%d6KIomb$YeyaEN=0tf6)L5*SGl-~N=v)B78yFTvKA396t51iDM~d3V7SG% z0JbalWvgHpaV_a!w8p|K<)@_(}*GZP>l!M(u ztv25e{1nxSEth-8Z<7W{-^bCHZIHg7hK|jgo`^rZB5VA~o`@9r2RJRR@oSOODp1td zoVS!=Tc;q1Cn^+*W4)~hdLv)Ybx21Yt}%I;PB6RQ{~E&OVOziVv+vse<4lbdQq!`FU}y%@F^vPl_v{7cKJ z79rLf8MYGzzFtUj!HOY-Jd!4ZY@7P<6uUK&G7wlHbK#!T^E`cxZwI@R7y<9-X9{%= z?hLW{XU^@?EYdSGGa}0Og)&FDeuNfGKQtH2D%c3O;WJ(}^vN-U;BDJzldaURCZU1bg$L$ntG;Vp$lO_Utp_;UGmfS~mo z*}Bl3%T{Qg~DVcXQY(0 zCU#Tn$&vHfRL;gmnf>@Umq>K4{bnVdWhjj;$T^r2H#SsUNU z67zx)tIfT!pQqb89!uR0nvoJ{*PrY}H(wF(!X;y7b!<>A^j9|(#&To&{_iU-`~K^( zKN?~98_CuKY@@Z^L~AEJSJ0^G(4CB}6VpF1gF>^2yEg{ithmjA)rL?lrMPEFwbI03 z^jL~`AqQh)ar3AYm|D~i)fjsjC5CYC^s(}≈i)lMcRw(xLHObb~3b^=$^j`oVN4=3~Cb5tIi+gho)Dw~rjt zS?JkyFgY*>kDxz;JBf>+h~>(Z7s_&Gz#Y;D;>#lX2)dv>I*~GJZORIsq(xpB1;4p? zeK#p<;iz0@JNyyJ3rlKX*G0>T5jMz1e6aY&sZAhvoHN{?`^o!YnT3TY9c(Zk>iIt@ zb{NuQ1*lzJRu}pGw47QIu3%>x^?;YZT2npz2pI)IoYvmX^qV0uQs(ah1lvnwtQdT_ z{0}F;eO9BWBHmi5-dqt|U~(IXrU1-UgP|7Y2(2$!a#i z{urpAY`I0=X;s)+Y4+{_BS%`}$8d5z%Fy3#9 zGJnMZng@Fz^mQZ_!@&6i(LCHm$Ms*wb%Y#X$H89`7=4640pkdmU;Z-iN~9g3g65UO z>jcBs0m@&8%Y~UagSb5lmMUv;kel0Ve~tmrtJq*-?O7}ky0!#g6(eG zg)QYQ^ll4OESQdIxMtXfcQyQ#6Cp zhSFdb8MItVv=~OZl_kNUWUl?LC*5MjU)4S84H==1k9c@;HQctdFYc%zSm{28>~gbL z+uf+#P4ZjoUjW000!JZP|>1IvWf;5P2gJ3sp3x72zTaul=4A%!B?FUAz5Jt5N`WQP`*X$Ui z0^F_*_oAn5FRpYgTT94eb~kS92b|}bi<9@to=@a6I|n4(`~}Wm ztrI5yh)iW)JYM9PdDxAGSPKopAIZcRCeM-&H&^)Le+-zF7_r#{HP91*x&Dd6t!|!P zxyPSGruk65((lQq&^eev{KnPOV2Zd~k7QQ6^seNG;ehUH`uwcuRA88ZsVg>NZpFJ( z9?B`^@ydA9=+<+Vei@q~Ss;V49apOXrAJelWT)<+r``Qcd34@nzvc&TV>(&304~hz%lEtbuN+=-`BqoQPme(%|y`&Eicr$ESJ+Ds(F5vR|4cz z;rA`AOrGc1X|iSF?7l&Dd>8QwQw^kLlSuX1Q40TVpFJF4K72&bhV3T_QO1^d&6YG> zx!WE9frxus^WI?Xs>~J*^3FX0S6LA6$P?8N$zza74OIQ>fzl(C?U1N0E`pLqHgFPJ zO6a7785bVnLITl56Zg66?gOifO`NwaY^{zk5_z>8Yd&U6(v5!N`0O_$ zJRhRrW1~}OeFfBrwRV3n@EdJ_QZCcxhd-FI=?dqrO_V}9ZwaW=0@4FO=KL?+ll*2o zLaKK8rL#ysJQOi-EZ4^3yT@OOe?nRL;DImJ_o0#K>_1C2PXz5KF27hNjdlMT;mss9 zPriHZOl-3PtIqx?zQF<9g)Ov{rQAER!IZZ<%p)Fg%~$Q-XPc9oSb0=t#SiZ*W`s0V zO#!*nXkK6aPR8Mqem}~uRvZ`tOv6U&Rw69_dtIsa-8Z-K{m~5-Gl4l^+GWzCPdYsr zuU=wsfqmAV4<_E7b1zCB?Do59zHCy__}U!Sm_8Sr5PvsgLBADp<)mTH`Ox_W2i?gS zsuYnjHhj`!kItsqx*;?uZvZ?aDPX$=-!h-s+P&jSU0U+r3TRcVWy3>ZDaWMvj)CoL-Xj{<^<#7 zK00&8XQvI3uioYd-?5tUa#+9k*yKn-wlNEQp=2{t^6p$TW!}s+t!H_18coZnOzYT1 zUU;Z6SK&fk;OL=?oThJmSxA+yQa_yS5PZX?S@6Qo$<2OBm0WWdrjZ22erTk=m)4p~ z{cE%_RWy4OwG6v$FJ zyc*HtMZw`GUz^QTayi$a;h`KDsxVTQdI;1c9(8STCRQXm>$J>?z!6n$4I|pev_5G1 zlJwDT@$I?#O3Q}bx$2R<^fkhW+NC49)(F+TTlQtlpkx=YcnswTU% zMFm^iN|Us3xRGD9_qnqtp*>jMGIge46mHc1b)3mP?_a9cBJv~TOObYDM*$`;v*;3h zyk7`L3OL4$^cGxOCLg&RN*Z`U+DJP|d`;iC(^%T5k>q_as8&g5eW<4{q`5mf3ETaC zKtAii2Z#@R(4DR|er5f*!M>>ZS-T1%o`&4Hxf>CZ z2>{7A=0}{g?Rm5qu2>+0TQ%-tbKvu`6rI#?G44#o`=Jef1>YXX>7531!`$ z0l`Vef@RXTodKO)szRrNGv0_qnCMDNg=vRCDF+%kzkTIb3KC#(DR-%komQ0B^9lx^ zX9BTbHuS|m6@_SR{B!oo_c0-VTSJwfC-hqSBUVu*O{?tggvJi72UaJo+L&t|v!nAl z4dRgvw9}YiLrm~W)uHH!|Ds|8&uU@9Hi1Q&~E^*rB6;5PL#A zkblHAuC$t^I_GE=x;44eNPoZbw3yx*Eq5n11y_({MVVQ1Y@6D6u48>aeBkc5DQ+T- zJRF_;7t;Q$)M8taCppn{ z+0PXHYM03UfYj@5>ym?uTWu45G7ELpx&J_usIQwMW4m1Wo?4C&?{(X>ZH=+KwKbAYPt&eawWW6p3C-zkU4pYWTypEgQd5Lz1$T^m+l`XgTw06;Bc* zfmkw+*>%Z#@^-b*x&B^$_hI?7weL3{s^-l?Bx13iNq%Etj5rjw^hGRZszL&ak6h`E_k3v zU0(aWdV|!MQt$sZz!gZvFd^1E%lh4sZlPx$2!V1fiqi&C^TgbTuzw|(T|XRB$@$KZ z2~%t+BaS?_tkzz@Et{+!wa$mq~cJ-PrV@z{cVxo;Blf^kslksI?_LJqkerP0?A0YO)6%(w;mfER% z>6L5~;<>BdQYunCn15(83REk*kF9cbkqe^9@U(q{Lkj4ifd_vf0d;`As^}K0~OOEQWxRC7d-Va6XA(AKZuBD5))U+Lg1LCxXPniSs z1$3LR<1rH_KqOCd$c+R~*ag{8Gt4y21%Fqhu?Nr|3dCEp8|a_&2~!XfxnKhxoikd{ z7bp!zc8Iy3aPbkak5A0|wy89tFgunKTvY^)F*YK_ZT{JCYs3wP?c%H5C^RAOu=20k znDOdeSOwKyy|mW@RUE}8S1iZT_XWkRFO=LEH^&G2*aaZL?M~ z+csx(>KIf|eF=a1yP{n=71Y9y>t&m#7E%)VMQrxe3~X8UQF0itq06+)NG9?P2&!1P z967|0`6nmcYH-cC{$73z9!=K69;Gi1@VJ!A(bL0irVmZREs3n-NRMK1-;jjqe;L+r zAWx}j1YXN^GenaMuFA%&4 zLwt?VTXMf&OB9dm7s&IOzzcQi?F?%;UhA)O3MttZT4t_K-Ds5ra+$!t%0sI3sgD^c zSrT>=q>&{Na1z+%W}Y+;Bxly&Vyf>fS@11ly1=BGV)Zxyl`1Vpo5K=>mVv-79tpi< zuubNe**v96mFaggC7-D-y^UE;K&7dVO)_#v{N_sqoL{T6PrbWwgD@E2vQH#p@CDB$ z{VB<77tLVMVQI+K3;z=%gwjs06un$rvmU;BQEA%Y{54md%ljLBrj*Q{iPqd?7#?M&L?ZP*WwV@7N@r+1km6o%I)CN&Cf+Mnc@#3QOotSWyA-00V$u%s* zhkiYvE%cOe(a?DTRd>4W9HuMa#(gQVPCiBu7PI<;?SDj36oYrlpNZhbVm)jpp>!6J zVrIkTQ*||)o2#5NYl5NIZ7yrsPH!uwU49gc0ZH-h_Mic%v2?@aIMIMJuS`Y5xO+Y_=i* literal 0 HcmV?d00001 diff --git a/Sandbox/res/fragment.fragment b/Sandbox/res/fragment.fragment deleted file mode 100644 index 7d9b08c..0000000 --- a/Sandbox/res/fragment.fragment +++ /dev/null @@ -1,18 +0,0 @@ -+GLSL -#version 440 core - -in vec4 fragColor; - -out vec4 FragmentColor; - -void main() -{ - FragmentColor = fragColor; -} --GLSL -+HLSL -float4 main(float4 Color : COLOR) : SV_Target -{ - return Color; -} --HLSL \ No newline at end of file diff --git a/Sandbox/res/vertex.vertex b/Sandbox/res/vertex.vertex deleted file mode 100644 index 8eade4c..0000000 --- a/Sandbox/res/vertex.vertex +++ /dev/null @@ -1,29 +0,0 @@ -+GLSL -#version 440 core - -layout(location = 0) in vec3 a_Position; -layout(location = 1) in vec4 a_Color; - -out vec4 fragColor; - -void main() -{ - gl_Position = vec4(a_Position, 1.0); - fragColor = a_Color; -} --GLSL -+HLSL -struct VertexOut -{ - float4 Color : COLOR; - float4 Position : SV_Position; -}; - -VertexOut main(float3 InPosition : POSITION, float4 InColor : COLOR) -{ - VertexOut vso; - vso.Position = float4(InPosition.x, InPosition.y, InPosition.z, 1.0); - vso.Color = InColor; - return vso; -} --HLSL \ No newline at end of file diff --git a/Sandbox/src/SandboxLayer.h b/Sandbox/src/SandboxLayer.h index 24b35ce..b2803ab 100644 --- a/Sandbox/src/SandboxLayer.h +++ b/Sandbox/src/SandboxLayer.h @@ -2,15 +2,23 @@ class SandboxLayer : public Light::Layer { +private: + std::shared_ptr m_AwesomefaceTexture; + public: - SandboxLayer(const std::string& name): Light::Layer(name) {} + SandboxLayer(const std::string& name) : Light::Layer(name) + { + Light::ResourceManager::LoadTexture("awesomeface", "res/Textures/awesomeface.png"); + m_AwesomefaceTexture = Light::ResourceManager::GetTexture("awesomeface"); + } void OnRender() override { - Light::Renderer::DrawQuad(glm::vec3(-0.5f, -0.5f, 0.0f), glm::vec2(1.f, 1.f), glm::vec4(1.12f, 1.41f, 1.72f, 1.0f)); Light::Renderer::DrawQuad(glm::vec3(-0.75f, -0.75f, 0.0f), glm::vec2(0.7f, 0.1f), glm::vec4(1.f, .2f, .2f, 1.0f)); Light::Renderer::DrawQuad(glm::vec3(0.2f, 0.5f, 0.0f), glm::vec2(0.6f, 0.6f), glm::vec4(.2f, 1.f, .2f, 1.0f)); Light::Renderer::DrawQuad(glm::vec3(-0.3f, 0.2f, 0.0f), glm::vec2(.4f, .4f), glm::vec4(.2f, 2.f, 1.f, 1.0f)); + + Light::Renderer::DrawQuad(glm::vec3(-0.5f, -0.5f, 0.0f), glm::vec2(1.0f, 1.0f), m_AwesomefaceTexture); } };