diff --git a/README.md b/README.md index f044c82..e411638 100644 --- a/README.md +++ b/README.md @@ -3,11 +3,76 @@ CUDA Denoiser For CUDA Path Tracer **University of Pennsylvania, CIS 565: GPU Programming and Architecture, Project 4** -* (TODO) YOUR NAME HERE -* Tested on: (TODO) Windows 22, i7-2222 @ 2.22GHz 22GB, GTX 222 222MB (Moore 2222 Lab) +* Stephen Lee + * [LinkedIn](https://www.linkedin.com/in/stephen-lee-bb5a40163/) +* Tested on: Windows 10, i7-9750H @2.60GHz, RTX 2060 6GB (personal laptop) +* 2 late days used for this project -### (TODO: Your README) +# Project Overview +The goal of this project was to build off of [Project 3](https://github.com/StephenLee129/Project3-CUDA-Path-Tracer), which implemented a path tracer to render photorealistic scenes by using physically-based rendering techniques. While path tracing produces scenes which are more photorealistic than other rendering techniques such as rasterization, a few key drawbacks are that path tracing can be incredibly computationally expensive and can produce noisy scenes even when left running for a long time due to diminishing returns for computation spent. In this project a CUDA denoiser was implemented as a post-processor for path traced scenes to cut down on the amount of compute that has to be done to get nice looking renders. -*DO NOT* leave the README to the last minute! It is a crucial part of the -project, and we will not be able to grade you without a good README. +Here's an example of a scene rendered for 100 path tracing iterations + +And here's an example of the same scene rendered for 2000 path tracing iterations + +Clearly the 2000 iteration example is much better, but it cost 20 times more compute to get it and it's still a little grainy. By applying the denoiser from this project we can transform the much cheaper 100 iteration render to make it look more like the 2000 iteration render without paying all the compute costs. +# Performance Analysis +### Cost of Denoising a Render +For the first part of my analysis, we examine the cost of running our denoiser on our renders. A total of 3 different scenes were analyzed for this section, where the time it took to path trace 100 iterations was measured, and then the denoiser was run with a filter size of 80 and a resolution of 800x800 to determine how much overhead it added to producing the final image. + +In this first graph, only 2 out of the 3 scenes are shown for the sake of clarity: + +It is clear that denoising had a minimal impact on the total time it took to produce the final scene. It only added a fairly constant amount of overhead of about 6-7ms to the overall render time. This overhead is made comically meaningless for scenes that are more computationally expensive in the path trace section. + +We can see clearly see here that the time it takes to path trace the scene heavily dominates the amount of time it takes to denoise the scene. Since the Bunny scene took so much longer to path trace than the simpler scenes, the additional time added to denoise the scene is partically negligible. +### Iterations Need to Get a Smooth Image +Having shown that the denoiser hardly affects overall performance for rendering a scene, we now want to analyze how many iterations we can cut out and still get a smooth image by applying our denoiser. To test this, the Cornell Ceiling Light scene was used with an 800x800 resolution. + +For our baseline in this analysis a standard path traced render without denoising was done for 2000 iterations. + +We can see that the image has rendered well and is pretty smooth overall with some small amounts of noise on the walls of the room. Compare this to the same scene rendered for only 100 iterations without denoising + +Clearly the extra 1900 iterations removed a lot of noise from the side walls, and we can even see some noise in the reflective ball in the center of the room now. However when we apply our denoiser with a FilterSize of 88, a color weight of 0.309, a normal weight of 0.052, and a position weight of 0.103, we can see that the image is incredibly smooth now + +When looking at the walls of the scene, they're even less noisy than the first 2000 iteration image. One downside is that the image gets slightly blurred when geometries meet. This can be attributed to the normal weights. Finer tuning of the normal weights may be able to get sharper edges between geometries while still denoising objects. The main issue in this scene is that if I decreased normal weights any more, the sphere in the center would get noisy since its surface is completely comprise of unique normals which weren't being weight heavily enough. However when normal weights were increased, the scene became a little blurred. +### Impact of Resolution +The next paramter that we want to analyze is image resolution. Since the path tracing algorithm shoots out a ray per pixel, we would expect this operation to take more time as resolution increases since more pixels will be in the scene, consequently making more rays to trace through the scene. Further since denoising similarly does per-pixel operatoins, we would expect denoising time to also increase with respect to pixel count. + + +We can see an exponential growth in both path tracing time (no denoising) in the first graph and denoising time in the second graph. Since we're scaling pixel count in the image exponentially as well, we can assert that this is a direct positive correlation between both the time it takes to path trace a scene and denoise a scene with respect to resolution. The first graph also further reinforces the findings from earlier where the cost of denoising hardly affects the overall image render time. Regardless of resolution, the first graph shows that No Denoising and With Denoising have virtually the same render times. +### Impact of Filter Size +Finally we test the impact of filter size on performance. Increasing filter size will increase the run time of denoising, because it increases the number of iterations needed to complete the A-Trous denoising algorithm that has been implemented. This number of iterations is discretized based on the `ceil(logbase2(filtersize))` so we would expect to see the time it takes to denoise the image increase in discrete steps with respect to filter size. For this section, we once again use the Cornell Ceiling Light test scene with a resolution of 800x800 to gather data. + +We observe the phenomenon that was discussed having discrete step ups in compute time for denoising as filter size increased. + +### How Filter Size Impacts Visuals +Since filter size impacts how wide of a range we are gather samples from we would expect larger filters to blur out finer local details. We can observe this when comparing two Bunny scenes where one on top has filter size of 16 and the other has a filter size of 58 + + +The finer details of the bunny's surface get completely smoothed out when going from a size 16 filter to a size 58 filter. While the walls of the scene become much less noisy at size 58, we lose a lot of details in the scene that makes this filter size not too practical for denoising this particular scene. + +### Denoising Effectiveness on Different Materials +Taking the example from earlier of a 100 path trace iteration render of Cornell Ceiling Light, we can compare the effectiveness on diffuse materials and reflective materials + +We can see that the geometry of the diffuse materials on the walls get maintained much better than that of the reflective sphere in the center of the scene. This is especially prominent at the center of the sphere where it is reflecting light out of the front of box. The edges of the walls in the reflection with respect to the empty front panel of the box is pretty hazy here and the wall colors kind of just fade into the black box opening. While there is some blurring on the edges of the actual walls, there are still very pronounced. +### Scene Result Comparison +The biggest difference across scenes that I encountered were between the simple Cornell box scenes and the Bunny scene. + + +Both of these renders were done for 100 path trace iterations, but the bunny scene comes out a lot worse. This can largely be attributed to the fact that the process of denoising an image will inherently take away information from the image by smoothing out values. Since the Bunny scene has a much more complex surface texture than the Cornell scene, denoising was not only a lot more finicky with regards to paramter tuning, but also just less effective overall with respect to visual fidelity. + +# Bloopers +A summary of a few of the issues that I encountered + + + +At first, I was doing the denoise computations in the same buffer that the pathtrace computations were done in. When the max iterations was reached, pathtracing stopped producing new data, and my denoising algorithm slowly made all the colors approach 0 leading to a black screen. + + + +In my first attempt to solve the issue from the previous blooper, I forgot to normalize the data I was copying over based on iteration number. The accumulated values would all flow past 1 leading to the entire screen except for the part that was black for all iterations to become white. + + + +Here I accidentally added each of the weight contributations together rather than multiplying them. This caused the weight to be much bigger than it should have been leading to a very blurry scene. \ No newline at end of file diff --git a/data.xlsx b/data.xlsx new file mode 100644 index 0000000..5de6539 Binary files /dev/null and b/data.xlsx differ diff --git a/external/include/tiny_obj_loader.h b/external/include/tiny_obj_loader.h new file mode 100644 index 0000000..6969a31 --- /dev/null +++ b/external/include/tiny_obj_loader.h @@ -0,0 +1,3369 @@ +/* +The MIT License (MIT) + +Copyright (c) 2012-Present, Syoyo Fujita and many contributors. + +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. +*/ + +// +// version 2.0.0 : Add new object oriented API. 1.x API is still provided. +// * Support line primitive. +// * Support points primitive. +// * Support multiple search path for .mtl(v1 API). +// * Support vertex weight `vw`(as an tinyobj extension) +// * Support escaped whitespece in mtllib +// * Add robust triangulation using Mapbox earcut(TINYOBJLOADER_USE_MAPBOX_EARCUT). +// version 1.4.0 : Modifed ParseTextureNameAndOption API +// version 1.3.1 : Make ParseTextureNameAndOption API public +// version 1.3.0 : Separate warning and error message(breaking API of LoadObj) +// version 1.2.3 : Added color space extension('-colorspace') to tex opts. +// version 1.2.2 : Parse multiple group names. +// version 1.2.1 : Added initial support for line('l') primitive(PR #178) +// version 1.2.0 : Hardened implementation(#175) +// version 1.1.1 : Support smoothing groups(#162) +// version 1.1.0 : Support parsing vertex color(#144) +// version 1.0.8 : Fix parsing `g` tag just after `usemtl`(#138) +// version 1.0.7 : Support multiple tex options(#126) +// version 1.0.6 : Add TINYOBJLOADER_USE_DOUBLE option(#124) +// version 1.0.5 : Ignore `Tr` when `d` exists in MTL(#43) +// version 1.0.4 : Support multiple filenames for 'mtllib'(#112) +// version 1.0.3 : Support parsing texture options(#85) +// version 1.0.2 : Improve parsing speed by about a factor of 2 for large +// files(#105) +// version 1.0.1 : Fixes a shape is lost if obj ends with a 'usemtl'(#104) +// version 1.0.0 : Change data structure. Change license from BSD to MIT. +// + +// +// Use this in *one* .cc +// #define TINYOBJLOADER_IMPLEMENTATION +// #include "tiny_obj_loader.h" +// + +#ifndef TINY_OBJ_LOADER_H_ +#define TINY_OBJ_LOADER_H_ + +#include +#include +#include + +namespace tinyobj { + + // TODO(syoyo): Better C++11 detection for older compiler +#if __cplusplus > 199711L +#define TINYOBJ_OVERRIDE override +#else +#define TINYOBJ_OVERRIDE +#endif + +#ifdef __clang__ +#pragma clang diagnostic push +#if __has_warning("-Wzero-as-null-pointer-constant") +#pragma clang diagnostic ignored "-Wzero-as-null-pointer-constant" +#endif + +#pragma clang diagnostic ignored "-Wpadded" + +#endif + +// https://en.wikipedia.org/wiki/Wavefront_.obj_file says ... +// +// -blendu on | off # set horizontal texture blending +// (default on) +// -blendv on | off # set vertical texture blending +// (default on) +// -boost real_value # boost mip-map sharpness +// -mm base_value gain_value # modify texture map values (default +// 0 1) +// # base_value = brightness, +// gain_value = contrast +// -o u [v [w]] # Origin offset (default +// 0 0 0) +// -s u [v [w]] # Scale (default +// 1 1 1) +// -t u [v [w]] # Turbulence (default +// 0 0 0) +// -texres resolution # texture resolution to create +// -clamp on | off # only render texels in the clamped +// 0-1 range (default off) +// # When unclamped, textures are +// repeated across a surface, +// # when clamped, only texels which +// fall within the 0-1 +// # range are rendered. +// -bm mult_value # bump multiplier (for bump maps +// only) +// +// -imfchan r | g | b | m | l | z # specifies which channel of the file +// is used to +// # create a scalar or bump texture. +// r:red, g:green, +// # b:blue, m:matte, l:luminance, +// z:z-depth.. +// # (the default for bump is 'l' and +// for decal is 'm') +// bump -imfchan r bumpmap.tga # says to use the red channel of +// bumpmap.tga as the bumpmap +// +// For reflection maps... +// +// -type sphere # specifies a sphere for a "refl" +// reflection map +// -type cube_top | cube_bottom | # when using a cube map, the texture +// file for each +// cube_front | cube_back | # side of the cube is specified +// separately +// cube_left | cube_right +// +// TinyObjLoader extension. +// +// -colorspace SPACE # Color space of the texture. e.g. +// 'sRGB` or 'linear' +// + +#ifdef TINYOBJLOADER_USE_DOUBLE +//#pragma message "using double" + typedef double real_t; +#else +//#pragma message "using float" + typedef float real_t; +#endif + + typedef enum { + TEXTURE_TYPE_NONE, // default + TEXTURE_TYPE_SPHERE, + TEXTURE_TYPE_CUBE_TOP, + TEXTURE_TYPE_CUBE_BOTTOM, + TEXTURE_TYPE_CUBE_FRONT, + TEXTURE_TYPE_CUBE_BACK, + TEXTURE_TYPE_CUBE_LEFT, + TEXTURE_TYPE_CUBE_RIGHT + } texture_type_t; + + struct texture_option_t { + texture_type_t type; // -type (default TEXTURE_TYPE_NONE) + real_t sharpness; // -boost (default 1.0?) + real_t brightness; // base_value in -mm option (default 0) + real_t contrast; // gain_value in -mm option (default 1) + real_t origin_offset[3]; // -o u [v [w]] (default 0 0 0) + real_t scale[3]; // -s u [v [w]] (default 1 1 1) + real_t turbulence[3]; // -t u [v [w]] (default 0 0 0) + int texture_resolution; // -texres resolution (No default value in the spec. + // We'll use -1) + bool clamp; // -clamp (default false) + char imfchan; // -imfchan (the default for bump is 'l' and for decal is 'm') + bool blendu; // -blendu (default on) + bool blendv; // -blendv (default on) + real_t bump_multiplier; // -bm (for bump maps only, default 1.0) + + // extension + std::string colorspace; // Explicitly specify color space of stored texel + // value. Usually `sRGB` or `linear` (default empty). + }; + + struct material_t { + std::string name; + + real_t ambient[3]; + real_t diffuse[3]; + real_t specular[3]; + real_t transmittance[3]; + real_t emission[3]; + real_t shininess; + real_t ior; // index of refraction + real_t dissolve; // 1 == opaque; 0 == fully transparent + // illumination model (see http://www.fileformat.info/format/material/) + int illum; + + int dummy; // Suppress padding warning. + + std::string ambient_texname; // map_Ka + std::string diffuse_texname; // map_Kd + std::string specular_texname; // map_Ks + std::string specular_highlight_texname; // map_Ns + std::string bump_texname; // map_bump, map_Bump, bump + std::string displacement_texname; // disp + std::string alpha_texname; // map_d + std::string reflection_texname; // refl + + texture_option_t ambient_texopt; + texture_option_t diffuse_texopt; + texture_option_t specular_texopt; + texture_option_t specular_highlight_texopt; + texture_option_t bump_texopt; + texture_option_t displacement_texopt; + texture_option_t alpha_texopt; + texture_option_t reflection_texopt; + + // PBR extension + // http://exocortex.com/blog/extending_wavefront_mtl_to_support_pbr + real_t roughness; // [0, 1] default 0 + real_t metallic; // [0, 1] default 0 + real_t sheen; // [0, 1] default 0 + real_t clearcoat_thickness; // [0, 1] default 0 + real_t clearcoat_roughness; // [0, 1] default 0 + real_t anisotropy; // aniso. [0, 1] default 0 + real_t anisotropy_rotation; // anisor. [0, 1] default 0 + real_t pad0; + std::string roughness_texname; // map_Pr + std::string metallic_texname; // map_Pm + std::string sheen_texname; // map_Ps + std::string emissive_texname; // map_Ke + std::string normal_texname; // norm. For normal mapping. + + texture_option_t roughness_texopt; + texture_option_t metallic_texopt; + texture_option_t sheen_texopt; + texture_option_t emissive_texopt; + texture_option_t normal_texopt; + + int pad2; + + std::map unknown_parameter; + +#ifdef TINY_OBJ_LOADER_PYTHON_BINDING + // For pybind11 + std::array GetDiffuse() { + std::array values; + values[0] = double(diffuse[0]); + values[1] = double(diffuse[1]); + values[2] = double(diffuse[2]); + + return values; + } + + std::array GetSpecular() { + std::array values; + values[0] = double(specular[0]); + values[1] = double(specular[1]); + values[2] = double(specular[2]); + + return values; + } + + std::array GetTransmittance() { + std::array values; + values[0] = double(transmittance[0]); + values[1] = double(transmittance[1]); + values[2] = double(transmittance[2]); + + return values; + } + + std::array GetEmission() { + std::array values; + values[0] = double(emission[0]); + values[1] = double(emission[1]); + values[2] = double(emission[2]); + + return values; + } + + std::array GetAmbient() { + std::array values; + values[0] = double(ambient[0]); + values[1] = double(ambient[1]); + values[2] = double(ambient[2]); + + return values; + } + + void SetDiffuse(std::array& a) { + diffuse[0] = real_t(a[0]); + diffuse[1] = real_t(a[1]); + diffuse[2] = real_t(a[2]); + } + + void SetAmbient(std::array& a) { + ambient[0] = real_t(a[0]); + ambient[1] = real_t(a[1]); + ambient[2] = real_t(a[2]); + } + + void SetSpecular(std::array& a) { + specular[0] = real_t(a[0]); + specular[1] = real_t(a[1]); + specular[2] = real_t(a[2]); + } + + void SetTransmittance(std::array& a) { + transmittance[0] = real_t(a[0]); + transmittance[1] = real_t(a[1]); + transmittance[2] = real_t(a[2]); + } + + std::string GetCustomParameter(const std::string& key) { + std::map::const_iterator it = + unknown_parameter.find(key); + + if (it != unknown_parameter.end()) { + return it->second; + } + return std::string(); + } + +#endif + }; + + struct tag_t { + std::string name; + + std::vector intValues; + std::vector floatValues; + std::vector stringValues; + }; + + struct joint_and_weight_t { + int joint_id; + real_t weight; + }; + + struct skin_weight_t { + int vertex_id; // Corresponding vertex index in `attrib_t::vertices`. + // Compared to `index_t`, this index must be positive and + // start with 0(does not allow relative indexing) + std::vector weightValues; + }; + + // Index struct to support different indices for vtx/normal/texcoord. + // -1 means not used. + struct index_t { + int vertex_index; + int normal_index; + int texcoord_index; + }; + + struct mesh_t { + std::vector indices; + std::vector + num_face_vertices; // The number of vertices per + // face. 3 = triangle, 4 = quad, + // ... Up to 255 vertices per face. + std::vector material_ids; // per-face material ID + std::vector smoothing_group_ids; // per-face smoothing group + // ID(0 = off. positive value + // = group id) + std::vector tags; // SubD tag + }; + + // struct path_t { + // std::vector indices; // pairs of indices for lines + //}; + + struct lines_t { + // Linear flattened indices. + std::vector indices; // indices for vertices(poly lines) + std::vector num_line_vertices; // The number of vertices per line. + }; + + struct points_t { + std::vector indices; // indices for points + }; + + struct shape_t { + std::string name; + mesh_t mesh; + lines_t lines; + points_t points; + }; + + // Vertex attributes + struct attrib_t { + std::vector vertices; // 'v'(xyz) + + // For backward compatibility, we store vertex weight in separate array. + std::vector vertex_weights; // 'v'(w) + std::vector normals; // 'vn' + std::vector texcoords; // 'vt'(uv) + + // For backward compatibility, we store texture coordinate 'w' in separate + // array. + std::vector texcoord_ws; // 'vt'(w) + std::vector colors; // extension: vertex colors + + // + // TinyObj extension. + // + + // NOTE(syoyo): array index is based on the appearance order. + // To get a corresponding skin weight for a specific vertex id `vid`, + // Need to reconstruct a look up table: `skin_weight_t::vertex_id` == `vid` + // (e.g. using std::map, std::unordered_map) + std::vector skin_weights; + + attrib_t() {} + + // + // For pybind11 + // + const std::vector& GetVertices() const { return vertices; } + + const std::vector& GetVertexWeights() const { return vertex_weights; } + }; + + struct callback_t { + // W is optional and set to 1 if there is no `w` item in `v` line + void (*vertex_cb)(void* user_data, real_t x, real_t y, real_t z, real_t w); + void (*normal_cb)(void* user_data, real_t x, real_t y, real_t z); + + // y and z are optional and set to 0 if there is no `y` and/or `z` item(s) in + // `vt` line. + void (*texcoord_cb)(void* user_data, real_t x, real_t y, real_t z); + + // called per 'f' line. num_indices is the number of face indices(e.g. 3 for + // triangle, 4 for quad) + // 0 will be passed for undefined index in index_t members. + void (*index_cb)(void* user_data, index_t* indices, int num_indices); + // `name` material name, `material_id` = the array index of material_t[]. -1 + // if + // a material not found in .mtl + void (*usemtl_cb)(void* user_data, const char* name, int material_id); + // `materials` = parsed material data. + void (*mtllib_cb)(void* user_data, const material_t* materials, + int num_materials); + // There may be multiple group names + void (*group_cb)(void* user_data, const char** names, int num_names); + void (*object_cb)(void* user_data, const char* name); + + callback_t() + : vertex_cb(NULL), + normal_cb(NULL), + texcoord_cb(NULL), + index_cb(NULL), + usemtl_cb(NULL), + mtllib_cb(NULL), + group_cb(NULL), + object_cb(NULL) {} + }; + + class MaterialReader { + public: + MaterialReader() {} + virtual ~MaterialReader(); + + virtual bool operator()(const std::string& matId, + std::vector* materials, + std::map* matMap, std::string* warn, + std::string* err) = 0; + }; + + /// + /// Read .mtl from a file. + /// + class MaterialFileReader : public MaterialReader { + public: + // Path could contain separator(';' in Windows, ':' in Posix) + explicit MaterialFileReader(const std::string& mtl_basedir) + : m_mtlBaseDir(mtl_basedir) {} + virtual ~MaterialFileReader() TINYOBJ_OVERRIDE {} + virtual bool operator()(const std::string& matId, + std::vector* materials, + std::map* matMap, std::string* warn, + std::string* err) TINYOBJ_OVERRIDE; + + private: + std::string m_mtlBaseDir; + }; + + /// + /// Read .mtl from a stream. + /// + class MaterialStreamReader : public MaterialReader { + public: + explicit MaterialStreamReader(std::istream& inStream) + : m_inStream(inStream) {} + virtual ~MaterialStreamReader() TINYOBJ_OVERRIDE {} + virtual bool operator()(const std::string& matId, + std::vector* materials, + std::map* matMap, std::string* warn, + std::string* err) TINYOBJ_OVERRIDE; + + private: + std::istream& m_inStream; + }; + + // v2 API + struct ObjReaderConfig { + bool triangulate; // triangulate polygon? + + // Currently not used. + // "simple" or empty: Create triangle fan + // "earcut": Use the algorithm based on Ear clipping + std::string triangulation_method; + + /// Parse vertex color. + /// If vertex color is not present, its filled with default value. + /// false = no vertex color + /// This will increase memory of parsed .obj + bool vertex_color; + + /// + /// Search path to .mtl file. + /// Default = "" = search from the same directory of .obj file. + /// Valid only when loading .obj from a file. + /// + std::string mtl_search_path; + + ObjReaderConfig() + : triangulate(true), triangulation_method("simple"), vertex_color(true) {} + }; + + /// + /// Wavefront .obj reader class(v2 API) + /// + class ObjReader { + public: + ObjReader() : valid_(false) {} + + /// + /// Load .obj and .mtl from a file. + /// + /// @param[in] filename wavefront .obj filename + /// @param[in] config Reader configuration + /// + bool ParseFromFile(const std::string& filename, + const ObjReaderConfig& config = ObjReaderConfig()); + + /// + /// Parse .obj from a text string. + /// Need to supply .mtl text string by `mtl_text`. + /// This function ignores `mtllib` line in .obj text. + /// + /// @param[in] obj_text wavefront .obj filename + /// @param[in] mtl_text wavefront .mtl filename + /// @param[in] config Reader configuration + /// + bool ParseFromString(const std::string& obj_text, const std::string& mtl_text, + const ObjReaderConfig& config = ObjReaderConfig()); + + /// + /// .obj was loaded or parsed correctly. + /// + bool Valid() const { return valid_; } + + const attrib_t& GetAttrib() const { return attrib_; } + + const std::vector& GetShapes() const { return shapes_; } + + const std::vector& GetMaterials() const { return materials_; } + + /// + /// Warning message(may be filled after `Load` or `Parse`) + /// + const std::string& Warning() const { return warning_; } + + /// + /// Error message(filled when `Load` or `Parse` failed) + /// + const std::string& Error() const { return error_; } + + private: + bool valid_; + + attrib_t attrib_; + std::vector shapes_; + std::vector materials_; + + std::string warning_; + std::string error_; + }; + + /// ==>>========= Legacy v1 API ============================================= + + /// Loads .obj from a file. + /// 'attrib', 'shapes' and 'materials' will be filled with parsed shape data + /// 'shapes' will be filled with parsed shape data + /// Returns true when loading .obj become success. + /// Returns warning message into `warn`, and error message into `err` + /// 'mtl_basedir' is optional, and used for base directory for .mtl file. + /// In default(`NULL'), .mtl file is searched from an application's working + /// directory. + /// 'triangulate' is optional, and used whether triangulate polygon face in .obj + /// or not. + /// Option 'default_vcols_fallback' specifies whether vertex colors should + /// always be defined, even if no colors are given (fallback to white). + bool LoadObj(attrib_t* attrib, std::vector* shapes, + std::vector* materials, std::string* warn, + std::string* err, const char* filename, + const char* mtl_basedir = NULL, bool triangulate = true, + bool default_vcols_fallback = true); + + /// Loads .obj from a file with custom user callback. + /// .mtl is loaded as usual and parsed material_t data will be passed to + /// `callback.mtllib_cb`. + /// Returns true when loading .obj/.mtl become success. + /// Returns warning message into `warn`, and error message into `err` + /// See `examples/callback_api/` for how to use this function. + bool LoadObjWithCallback(std::istream& inStream, const callback_t& callback, + void* user_data = NULL, + MaterialReader* readMatFn = NULL, + std::string* warn = NULL, std::string* err = NULL); + + /// Loads object from a std::istream, uses `readMatFn` to retrieve + /// std::istream for materials. + /// Returns true when loading .obj become success. + /// Returns warning and error message into `err` + bool LoadObj(attrib_t* attrib, std::vector* shapes, + std::vector* materials, std::string* warn, + std::string* err, std::istream* inStream, + MaterialReader* readMatFn = NULL, bool triangulate = true, + bool default_vcols_fallback = true); + + /// Loads materials into std::map + void LoadMtl(std::map* material_map, + std::vector* materials, std::istream* inStream, + std::string* warning, std::string* err); + + /// + /// Parse texture name and texture option for custom texture parameter through + /// material::unknown_parameter + /// + /// @param[out] texname Parsed texture name + /// @param[out] texopt Parsed texopt + /// @param[in] linebuf Input string + /// + bool ParseTextureNameAndOption(std::string* texname, texture_option_t* texopt, + const char* linebuf); + + /// =<<========== Legacy v1 API ============================================= + +} // namespace tinyobj + +#endif // TINY_OBJ_LOADER_H_ + +#ifdef TINYOBJLOADER_IMPLEMENTATION +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#ifdef TINYOBJLOADER_USE_MAPBOX_EARCUT + +#ifdef TINYOBJLOADER_DONOT_INCLUDE_MAPBOX_EARCUT +// Assume earcut.hpp is included outside of tiny_obj_loader.h +#else + +#ifdef __clang__ +#pragma clang diagnostic push +#pragma clang diagnostic ignored "-Weverything" +#endif + +#include +#include "mapbox/earcut.hpp" + +#ifdef __clang__ +#pragma clang diagnostic pop +#endif + +#endif + +#endif // TINYOBJLOADER_USE_MAPBOX_EARCUT + +namespace tinyobj { + + MaterialReader::~MaterialReader() {} + + struct vertex_index_t { + int v_idx, vt_idx, vn_idx; + vertex_index_t() : v_idx(-1), vt_idx(-1), vn_idx(-1) {} + explicit vertex_index_t(int idx) : v_idx(idx), vt_idx(idx), vn_idx(idx) {} + vertex_index_t(int vidx, int vtidx, int vnidx) + : v_idx(vidx), vt_idx(vtidx), vn_idx(vnidx) {} + }; + + // Internal data structure for face representation + // index + smoothing group. + struct face_t { + unsigned int + smoothing_group_id; // smoothing group id. 0 = smoothing groupd is off. + int pad_; + std::vector vertex_indices; // face vertex indices. + + face_t() : smoothing_group_id(0), pad_(0) {} + }; + + // Internal data structure for line representation + struct __line_t { + // l v1/vt1 v2/vt2 ... + // In the specification, line primitrive does not have normal index, but + // TinyObjLoader allow it + std::vector vertex_indices; + }; + + // Internal data structure for points representation + struct __points_t { + // p v1 v2 ... + // In the specification, point primitrive does not have normal index and + // texture coord index, but TinyObjLoader allow it. + std::vector vertex_indices; + }; + + struct tag_sizes { + tag_sizes() : num_ints(0), num_reals(0), num_strings(0) {} + int num_ints; + int num_reals; + int num_strings; + }; + + struct obj_shape { + std::vector v; + std::vector vn; + std::vector vt; + }; + + // + // Manages group of primitives(face, line, points, ...) + struct PrimGroup { + std::vector faceGroup; + std::vector<__line_t> lineGroup; + std::vector<__points_t> pointsGroup; + + void clear() { + faceGroup.clear(); + lineGroup.clear(); + pointsGroup.clear(); + } + + bool IsEmpty() const { + return faceGroup.empty() && lineGroup.empty() && pointsGroup.empty(); + } + + // TODO(syoyo): bspline, surface, ... + }; + + // See + // http://stackoverflow.com/questions/6089231/getting-std-ifstream-to-handle-lf-cr-and-crlf + static std::istream& safeGetline(std::istream& is, std::string& t) { + t.clear(); + + // The characters in the stream are read one-by-one using a std::streambuf. + // That is faster than reading them one-by-one using the std::istream. + // Code that uses streambuf this way must be guarded by a sentry object. + // The sentry object performs various tasks, + // such as thread synchronization and updating the stream state. + + std::istream::sentry se(is, true); + std::streambuf* sb = is.rdbuf(); + + if (se) { + for (;;) { + int c = sb->sbumpc(); + switch (c) { + case '\n': + return is; + case '\r': + if (sb->sgetc() == '\n') sb->sbumpc(); + return is; + case EOF: + // Also handle the case when the last line has no line ending + if (t.empty()) is.setstate(std::ios::eofbit); + return is; + default: + t += static_cast(c); + } + } + } + + return is; + } + +#define IS_SPACE(x) (((x) == ' ') || ((x) == '\t')) +#define IS_DIGIT(x) \ + (static_cast((x) - '0') < static_cast(10)) +#define IS_NEW_LINE(x) (((x) == '\r') || ((x) == '\n') || ((x) == '\0')) + + // Make index zero-base, and also support relative index. + static inline bool fixIndex(int idx, int n, int* ret) { + if (!ret) { + return false; + } + + if (idx > 0) { + (*ret) = idx - 1; + return true; + } + + if (idx == 0) { + // zero is not allowed according to the spec. + return false; + } + + if (idx < 0) { + (*ret) = n + idx; // negative value = relative + return true; + } + + return false; // never reach here. + } + + static inline std::string parseString(const char** token) { + std::string s; + (*token) += strspn((*token), " \t"); + size_t e = strcspn((*token), " \t\r"); + s = std::string((*token), &(*token)[e]); + (*token) += e; + return s; + } + + static inline int parseInt(const char** token) { + (*token) += strspn((*token), " \t"); + int i = atoi((*token)); + (*token) += strcspn((*token), " \t\r"); + return i; + } + + // Tries to parse a floating point number located at s. + // + // s_end should be a location in the string where reading should absolutely + // stop. For example at the end of the string, to prevent buffer overflows. + // + // Parses the following EBNF grammar: + // sign = "+" | "-" ; + // END = ? anything not in digit ? + // digit = "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" | "8" | "9" ; + // integer = [sign] , digit , {digit} ; + // decimal = integer , ["." , integer] ; + // float = ( decimal , END ) | ( decimal , ("E" | "e") , integer , END ) ; + // + // Valid strings are for example: + // -0 +3.1417e+2 -0.0E-3 1.0324 -1.41 11e2 + // + // If the parsing is a success, result is set to the parsed value and true + // is returned. + // + // The function is greedy and will parse until any of the following happens: + // - a non-conforming character is encountered. + // - s_end is reached. + // + // The following situations triggers a failure: + // - s >= s_end. + // - parse failure. + // + static bool tryParseDouble(const char* s, const char* s_end, double* result) { + if (s >= s_end) { + return false; + } + + double mantissa = 0.0; + // This exponent is base 2 rather than 10. + // However the exponent we parse is supposed to be one of ten, + // thus we must take care to convert the exponent/and or the + // mantissa to a * 2^E, where a is the mantissa and E is the + // exponent. + // To get the final double we will use ldexp, it requires the + // exponent to be in base 2. + int exponent = 0; + + // NOTE: THESE MUST BE DECLARED HERE SINCE WE ARE NOT ALLOWED + // TO JUMP OVER DEFINITIONS. + char sign = '+'; + char exp_sign = '+'; + char const* curr = s; + + // How many characters were read in a loop. + int read = 0; + // Tells whether a loop terminated due to reaching s_end. + bool end_not_reached = false; + bool leading_decimal_dots = false; + + /* + BEGIN PARSING. + */ + + // Find out what sign we've got. + if (*curr == '+' || *curr == '-') { + sign = *curr; + curr++; + if ((curr != s_end) && (*curr == '.')) { + // accept. Somethig like `.7e+2`, `-.5234` + leading_decimal_dots = true; + } + } + else if (IS_DIGIT(*curr)) { /* Pass through. */ + } + else if (*curr == '.') { + // accept. Somethig like `.7e+2`, `-.5234` + leading_decimal_dots = true; + } + else { + goto fail; + } + + // Read the integer part. + end_not_reached = (curr != s_end); + if (!leading_decimal_dots) { + while (end_not_reached && IS_DIGIT(*curr)) { + mantissa *= 10; + mantissa += static_cast(*curr - 0x30); + curr++; + read++; + end_not_reached = (curr != s_end); + } + + // We must make sure we actually got something. + if (read == 0) goto fail; + } + + // We allow numbers of form "#", "###" etc. + if (!end_not_reached) goto assemble; + + // Read the decimal part. + if (*curr == '.') { + curr++; + read = 1; + end_not_reached = (curr != s_end); + while (end_not_reached && IS_DIGIT(*curr)) { + static const double pow_lut[] = { + 1.0, 0.1, 0.01, 0.001, 0.0001, 0.00001, 0.000001, 0.0000001, + }; + const int lut_entries = sizeof pow_lut / sizeof pow_lut[0]; + + // NOTE: Don't use powf here, it will absolutely murder precision. + mantissa += static_cast(*curr - 0x30) * + (read < lut_entries ? pow_lut[read] : std::pow(10.0, -read)); + read++; + curr++; + end_not_reached = (curr != s_end); + } + } + else if (*curr == 'e' || *curr == 'E') { + } + else { + goto assemble; + } + + if (!end_not_reached) goto assemble; + + // Read the exponent part. + if (*curr == 'e' || *curr == 'E') { + curr++; + // Figure out if a sign is present and if it is. + end_not_reached = (curr != s_end); + if (end_not_reached && (*curr == '+' || *curr == '-')) { + exp_sign = *curr; + curr++; + } + else if (IS_DIGIT(*curr)) { /* Pass through. */ + } + else { + // Empty E is not allowed. + goto fail; + } + + read = 0; + end_not_reached = (curr != s_end); + while (end_not_reached && IS_DIGIT(*curr)) { + // To avoid annoying MSVC's min/max macro definiton, + // Use hardcoded int max value + if (exponent > (2147483647 / 10)) { // 2147483647 = std::numeric_limits::max() + // Integer overflow + goto fail; + } + exponent *= 10; + exponent += static_cast(*curr - 0x30); + curr++; + read++; + end_not_reached = (curr != s_end); + } + exponent *= (exp_sign == '+' ? 1 : -1); + if (read == 0) goto fail; + } + + assemble: + *result = (sign == '+' ? 1 : -1) * + (exponent ? std::ldexp(mantissa * std::pow(5.0, exponent), exponent) + : mantissa); + return true; + fail: + return false; + } + + static inline real_t parseReal(const char** token, double default_value = 0.0) { + (*token) += strspn((*token), " \t"); + const char* end = (*token) + strcspn((*token), " \t\r"); + double val = default_value; + tryParseDouble((*token), end, &val); + real_t f = static_cast(val); + (*token) = end; + return f; + } + + static inline bool parseReal(const char** token, real_t* out) { + (*token) += strspn((*token), " \t"); + const char* end = (*token) + strcspn((*token), " \t\r"); + double val; + bool ret = tryParseDouble((*token), end, &val); + if (ret) { + real_t f = static_cast(val); + (*out) = f; + } + (*token) = end; + return ret; + } + + static inline void parseReal2(real_t* x, real_t* y, const char** token, + const double default_x = 0.0, + const double default_y = 0.0) { + (*x) = parseReal(token, default_x); + (*y) = parseReal(token, default_y); + } + + static inline void parseReal3(real_t* x, real_t* y, real_t* z, + const char** token, const double default_x = 0.0, + const double default_y = 0.0, + const double default_z = 0.0) { + (*x) = parseReal(token, default_x); + (*y) = parseReal(token, default_y); + (*z) = parseReal(token, default_z); + } + + static inline void parseV(real_t* x, real_t* y, real_t* z, real_t* w, + const char** token, const double default_x = 0.0, + const double default_y = 0.0, + const double default_z = 0.0, + const double default_w = 1.0) { + (*x) = parseReal(token, default_x); + (*y) = parseReal(token, default_y); + (*z) = parseReal(token, default_z); + (*w) = parseReal(token, default_w); + } + + // Extension: parse vertex with colors(6 items) + static inline bool parseVertexWithColor(real_t* x, real_t* y, real_t* z, + real_t* r, real_t* g, real_t* b, + const char** token, + const double default_x = 0.0, + const double default_y = 0.0, + const double default_z = 0.0) { + (*x) = parseReal(token, default_x); + (*y) = parseReal(token, default_y); + (*z) = parseReal(token, default_z); + + const bool found_color = + parseReal(token, r) && parseReal(token, g) && parseReal(token, b); + + if (!found_color) { + (*r) = (*g) = (*b) = 1.0; + } + + return found_color; + } + + static inline bool parseOnOff(const char** token, bool default_value = true) { + (*token) += strspn((*token), " \t"); + const char* end = (*token) + strcspn((*token), " \t\r"); + + bool ret = default_value; + if ((0 == strncmp((*token), "on", 2))) { + ret = true; + } + else if ((0 == strncmp((*token), "off", 3))) { + ret = false; + } + + (*token) = end; + return ret; + } + + static inline texture_type_t parseTextureType( + const char** token, texture_type_t default_value = TEXTURE_TYPE_NONE) { + (*token) += strspn((*token), " \t"); + const char* end = (*token) + strcspn((*token), " \t\r"); + texture_type_t ty = default_value; + + if ((0 == strncmp((*token), "cube_top", strlen("cube_top")))) { + ty = TEXTURE_TYPE_CUBE_TOP; + } + else if ((0 == strncmp((*token), "cube_bottom", strlen("cube_bottom")))) { + ty = TEXTURE_TYPE_CUBE_BOTTOM; + } + else if ((0 == strncmp((*token), "cube_left", strlen("cube_left")))) { + ty = TEXTURE_TYPE_CUBE_LEFT; + } + else if ((0 == strncmp((*token), "cube_right", strlen("cube_right")))) { + ty = TEXTURE_TYPE_CUBE_RIGHT; + } + else if ((0 == strncmp((*token), "cube_front", strlen("cube_front")))) { + ty = TEXTURE_TYPE_CUBE_FRONT; + } + else if ((0 == strncmp((*token), "cube_back", strlen("cube_back")))) { + ty = TEXTURE_TYPE_CUBE_BACK; + } + else if ((0 == strncmp((*token), "sphere", strlen("sphere")))) { + ty = TEXTURE_TYPE_SPHERE; + } + + (*token) = end; + return ty; + } + + static tag_sizes parseTagTriple(const char** token) { + tag_sizes ts; + + (*token) += strspn((*token), " \t"); + ts.num_ints = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + return ts; + } + + (*token)++; // Skip '/' + + (*token) += strspn((*token), " \t"); + ts.num_reals = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + return ts; + } + (*token)++; // Skip '/' + + ts.num_strings = parseInt(token); + + return ts; + } + + // Parse triples with index offsets: i, i/j/k, i//k, i/j + static bool parseTriple(const char** token, int vsize, int vnsize, int vtsize, + vertex_index_t* ret) { + if (!ret) { + return false; + } + + vertex_index_t vi(-1); + + if (!fixIndex(atoi((*token)), vsize, &(vi.v_idx))) { + return false; + } + + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + (*ret) = vi; + return true; + } + (*token)++; + + // i//k + if ((*token)[0] == '/') { + (*token)++; + if (!fixIndex(atoi((*token)), vnsize, &(vi.vn_idx))) { + return false; + } + (*token) += strcspn((*token), "/ \t\r"); + (*ret) = vi; + return true; + } + + // i/j/k or i/j + if (!fixIndex(atoi((*token)), vtsize, &(vi.vt_idx))) { + return false; + } + + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + (*ret) = vi; + return true; + } + + // i/j/k + (*token)++; // skip '/' + if (!fixIndex(atoi((*token)), vnsize, &(vi.vn_idx))) { + return false; + } + (*token) += strcspn((*token), "/ \t\r"); + + (*ret) = vi; + + return true; + } + + // Parse raw triples: i, i/j/k, i//k, i/j + static vertex_index_t parseRawTriple(const char** token) { + vertex_index_t vi(static_cast(0)); // 0 is an invalid index in OBJ + + vi.v_idx = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + return vi; + } + (*token)++; + + // i//k + if ((*token)[0] == '/') { + (*token)++; + vi.vn_idx = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + return vi; + } + + // i/j/k or i/j + vi.vt_idx = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + if ((*token)[0] != '/') { + return vi; + } + + // i/j/k + (*token)++; // skip '/' + vi.vn_idx = atoi((*token)); + (*token) += strcspn((*token), "/ \t\r"); + return vi; + } + + bool ParseTextureNameAndOption(std::string* texname, texture_option_t* texopt, + const char* linebuf) { + // @todo { write more robust lexer and parser. } + bool found_texname = false; + std::string texture_name; + + const char* token = linebuf; // Assume line ends with NULL + + while (!IS_NEW_LINE((*token))) { + token += strspn(token, " \t"); // skip space + if ((0 == strncmp(token, "-blendu", 7)) && IS_SPACE((token[7]))) { + token += 8; + texopt->blendu = parseOnOff(&token, /* default */ true); + } + else if ((0 == strncmp(token, "-blendv", 7)) && IS_SPACE((token[7]))) { + token += 8; + texopt->blendv = parseOnOff(&token, /* default */ true); + } + else if ((0 == strncmp(token, "-clamp", 6)) && IS_SPACE((token[6]))) { + token += 7; + texopt->clamp = parseOnOff(&token, /* default */ true); + } + else if ((0 == strncmp(token, "-boost", 6)) && IS_SPACE((token[6]))) { + token += 7; + texopt->sharpness = parseReal(&token, 1.0); + } + else if ((0 == strncmp(token, "-bm", 3)) && IS_SPACE((token[3]))) { + token += 4; + texopt->bump_multiplier = parseReal(&token, 1.0); + } + else if ((0 == strncmp(token, "-o", 2)) && IS_SPACE((token[2]))) { + token += 3; + parseReal3(&(texopt->origin_offset[0]), &(texopt->origin_offset[1]), + &(texopt->origin_offset[2]), &token); + } + else if ((0 == strncmp(token, "-s", 2)) && IS_SPACE((token[2]))) { + token += 3; + parseReal3(&(texopt->scale[0]), &(texopt->scale[1]), &(texopt->scale[2]), + &token, 1.0, 1.0, 1.0); + } + else if ((0 == strncmp(token, "-t", 2)) && IS_SPACE((token[2]))) { + token += 3; + parseReal3(&(texopt->turbulence[0]), &(texopt->turbulence[1]), + &(texopt->turbulence[2]), &token); + } + else if ((0 == strncmp(token, "-type", 5)) && IS_SPACE((token[5]))) { + token += 5; + texopt->type = parseTextureType((&token), TEXTURE_TYPE_NONE); + } + else if ((0 == strncmp(token, "-texres", 7)) && IS_SPACE((token[7]))) { + token += 7; + // TODO(syoyo): Check if arg is int type. + texopt->texture_resolution = parseInt(&token); + } + else if ((0 == strncmp(token, "-imfchan", 8)) && IS_SPACE((token[8]))) { + token += 9; + token += strspn(token, " \t"); + const char* end = token + strcspn(token, " \t\r"); + if ((end - token) == 1) { // Assume one char for -imfchan + texopt->imfchan = (*token); + } + token = end; + } + else if ((0 == strncmp(token, "-mm", 3)) && IS_SPACE((token[3]))) { + token += 4; + parseReal2(&(texopt->brightness), &(texopt->contrast), &token, 0.0, 1.0); + } + else if ((0 == strncmp(token, "-colorspace", 11)) && + IS_SPACE((token[11]))) { + token += 12; + texopt->colorspace = parseString(&token); + } + else { + // Assume texture filename +#if 0 + size_t len = strcspn(token, " \t\r"); // untile next space + texture_name = std::string(token, token + len); + token += len; + + token += strspn(token, " \t"); // skip space +#else + // Read filename until line end to parse filename containing whitespace + // TODO(syoyo): Support parsing texture option flag after the filename. + texture_name = std::string(token); + token += texture_name.length(); +#endif + + found_texname = true; + } + } + + if (found_texname) { + (*texname) = texture_name; + return true; + } + else { + return false; + } + } + + static void InitTexOpt(texture_option_t* texopt, const bool is_bump) { + if (is_bump) { + texopt->imfchan = 'l'; + } + else { + texopt->imfchan = 'm'; + } + texopt->bump_multiplier = static_cast(1.0); + texopt->clamp = false; + texopt->blendu = true; + texopt->blendv = true; + texopt->sharpness = static_cast(1.0); + texopt->brightness = static_cast(0.0); + texopt->contrast = static_cast(1.0); + texopt->origin_offset[0] = static_cast(0.0); + texopt->origin_offset[1] = static_cast(0.0); + texopt->origin_offset[2] = static_cast(0.0); + texopt->scale[0] = static_cast(1.0); + texopt->scale[1] = static_cast(1.0); + texopt->scale[2] = static_cast(1.0); + texopt->turbulence[0] = static_cast(0.0); + texopt->turbulence[1] = static_cast(0.0); + texopt->turbulence[2] = static_cast(0.0); + texopt->texture_resolution = -1; + texopt->type = TEXTURE_TYPE_NONE; + } + + static void InitMaterial(material_t* material) { + InitTexOpt(&material->ambient_texopt, /* is_bump */ false); + InitTexOpt(&material->diffuse_texopt, /* is_bump */ false); + InitTexOpt(&material->specular_texopt, /* is_bump */ false); + InitTexOpt(&material->specular_highlight_texopt, /* is_bump */ false); + InitTexOpt(&material->bump_texopt, /* is_bump */ true); + InitTexOpt(&material->displacement_texopt, /* is_bump */ false); + InitTexOpt(&material->alpha_texopt, /* is_bump */ false); + InitTexOpt(&material->reflection_texopt, /* is_bump */ false); + InitTexOpt(&material->roughness_texopt, /* is_bump */ false); + InitTexOpt(&material->metallic_texopt, /* is_bump */ false); + InitTexOpt(&material->sheen_texopt, /* is_bump */ false); + InitTexOpt(&material->emissive_texopt, /* is_bump */ false); + InitTexOpt(&material->normal_texopt, + /* is_bump */ false); // @fixme { is_bump will be true? } + material->name = ""; + material->ambient_texname = ""; + material->diffuse_texname = ""; + material->specular_texname = ""; + material->specular_highlight_texname = ""; + material->bump_texname = ""; + material->displacement_texname = ""; + material->reflection_texname = ""; + material->alpha_texname = ""; + for (int i = 0; i < 3; i++) { + material->ambient[i] = static_cast(0.0); + material->diffuse[i] = static_cast(0.0); + material->specular[i] = static_cast(0.0); + material->transmittance[i] = static_cast(0.0); + material->emission[i] = static_cast(0.0); + } + material->illum = 0; + material->dissolve = static_cast(1.0); + material->shininess = static_cast(1.0); + material->ior = static_cast(1.0); + + material->roughness = static_cast(0.0); + material->metallic = static_cast(0.0); + material->sheen = static_cast(0.0); + material->clearcoat_thickness = static_cast(0.0); + material->clearcoat_roughness = static_cast(0.0); + material->anisotropy_rotation = static_cast(0.0); + material->anisotropy = static_cast(0.0); + material->roughness_texname = ""; + material->metallic_texname = ""; + material->sheen_texname = ""; + material->emissive_texname = ""; + material->normal_texname = ""; + + material->unknown_parameter.clear(); + } + + // code from https://wrf.ecse.rpi.edu//Research/Short_Notes/pnpoly.html + template + static int pnpoly(int nvert, T* vertx, T* verty, T testx, T testy) { + int i, j, c = 0; + for (i = 0, j = nvert - 1; i < nvert; j = i++) { + if (((verty[i] > testy) != (verty[j] > testy)) && + (testx < + (vertx[j] - vertx[i]) * (testy - verty[i]) / (verty[j] - verty[i]) + + vertx[i])) + c = !c; + } + return c; + } + + // TODO(syoyo): refactor function. + static bool exportGroupsToShape(shape_t* shape, const PrimGroup& prim_group, + const std::vector& tags, + const int material_id, const std::string& name, + bool triangulate, const std::vector& v, + std::string* warn) { + if (prim_group.IsEmpty()) { + return false; + } + + shape->name = name; + + // polygon + if (!prim_group.faceGroup.empty()) { + // Flatten vertices and indices + for (size_t i = 0; i < prim_group.faceGroup.size(); i++) { + const face_t& face = prim_group.faceGroup[i]; + + size_t npolys = face.vertex_indices.size(); + + if (npolys < 3) { + // Face must have 3+ vertices. + if (warn) { + (*warn) += "Degenerated face found\n."; + } + continue; + } + + if (triangulate) { + if (npolys == 4) { + vertex_index_t i0 = face.vertex_indices[0]; + vertex_index_t i1 = face.vertex_indices[1]; + vertex_index_t i2 = face.vertex_indices[2]; + vertex_index_t i3 = face.vertex_indices[3]; + + size_t vi0 = size_t(i0.v_idx); + size_t vi1 = size_t(i1.v_idx); + size_t vi2 = size_t(i2.v_idx); + size_t vi3 = size_t(i3.v_idx); + + if (((3 * vi0 + 2) >= v.size()) || ((3 * vi1 + 2) >= v.size()) || + ((3 * vi2 + 2) >= v.size()) || ((3 * vi3 + 2) >= v.size())) { + // Invalid triangle. + // FIXME(syoyo): Is it ok to simply skip this invalid triangle? + if (warn) { + (*warn) += "Face with invalid vertex index found.\n"; + } + continue; + } + + real_t v0x = v[vi0 * 3 + 0]; + real_t v0y = v[vi0 * 3 + 1]; + real_t v0z = v[vi0 * 3 + 2]; + real_t v1x = v[vi1 * 3 + 0]; + real_t v1y = v[vi1 * 3 + 1]; + real_t v1z = v[vi1 * 3 + 2]; + real_t v2x = v[vi2 * 3 + 0]; + real_t v2y = v[vi2 * 3 + 1]; + real_t v2z = v[vi2 * 3 + 2]; + real_t v3x = v[vi3 * 3 + 0]; + real_t v3y = v[vi3 * 3 + 1]; + real_t v3z = v[vi3 * 3 + 2]; + + // There are two candidates to split the quad into two triangles. + // + // Choose the shortest edge. + // TODO: Is it better to determine the edge to split by calculating + // the area of each triangle? + // + // +---+ + // |\ | + // | \ | + // | \| + // +---+ + // + // +---+ + // | /| + // | / | + // |/ | + // +---+ + + real_t e02x = v2x - v0x; + real_t e02y = v2y - v0y; + real_t e02z = v2z - v0z; + real_t e13x = v3x - v1x; + real_t e13y = v3y - v1y; + real_t e13z = v3z - v1z; + + real_t sqr02 = e02x * e02x + e02y * e02y + e02z * e02z; + real_t sqr13 = e13x * e13x + e13y * e13y + e13z * e13z; + + index_t idx0, idx1, idx2, idx3; + + idx0.vertex_index = i0.v_idx; + idx0.normal_index = i0.vn_idx; + idx0.texcoord_index = i0.vt_idx; + idx1.vertex_index = i1.v_idx; + idx1.normal_index = i1.vn_idx; + idx1.texcoord_index = i1.vt_idx; + idx2.vertex_index = i2.v_idx; + idx2.normal_index = i2.vn_idx; + idx2.texcoord_index = i2.vt_idx; + idx3.vertex_index = i3.v_idx; + idx3.normal_index = i3.vn_idx; + idx3.texcoord_index = i3.vt_idx; + + if (sqr02 < sqr13) { + // [0, 1, 2], [0, 2, 3] + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx2); + + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx2); + shape->mesh.indices.push_back(idx3); + } + else { + // [0, 1, 3], [1, 2, 3] + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx3); + + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx2); + shape->mesh.indices.push_back(idx3); + } + + // Two triangle faces + shape->mesh.num_face_vertices.push_back(3); + shape->mesh.num_face_vertices.push_back(3); + + shape->mesh.material_ids.push_back(material_id); + shape->mesh.material_ids.push_back(material_id); + + shape->mesh.smoothing_group_ids.push_back(face.smoothing_group_id); + shape->mesh.smoothing_group_ids.push_back(face.smoothing_group_id); + + } + else { + vertex_index_t i0 = face.vertex_indices[0]; + vertex_index_t i1(-1); + vertex_index_t i2 = face.vertex_indices[1]; + + // find the two axes to work in + size_t axes[2] = { 1, 2 }; + for (size_t k = 0; k < npolys; ++k) { + i0 = face.vertex_indices[(k + 0) % npolys]; + i1 = face.vertex_indices[(k + 1) % npolys]; + i2 = face.vertex_indices[(k + 2) % npolys]; + size_t vi0 = size_t(i0.v_idx); + size_t vi1 = size_t(i1.v_idx); + size_t vi2 = size_t(i2.v_idx); + + if (((3 * vi0 + 2) >= v.size()) || ((3 * vi1 + 2) >= v.size()) || + ((3 * vi2 + 2) >= v.size())) { + // Invalid triangle. + // FIXME(syoyo): Is it ok to simply skip this invalid triangle? + continue; + } + real_t v0x = v[vi0 * 3 + 0]; + real_t v0y = v[vi0 * 3 + 1]; + real_t v0z = v[vi0 * 3 + 2]; + real_t v1x = v[vi1 * 3 + 0]; + real_t v1y = v[vi1 * 3 + 1]; + real_t v1z = v[vi1 * 3 + 2]; + real_t v2x = v[vi2 * 3 + 0]; + real_t v2y = v[vi2 * 3 + 1]; + real_t v2z = v[vi2 * 3 + 2]; + real_t e0x = v1x - v0x; + real_t e0y = v1y - v0y; + real_t e0z = v1z - v0z; + real_t e1x = v2x - v1x; + real_t e1y = v2y - v1y; + real_t e1z = v2z - v1z; + real_t cx = std::fabs(e0y * e1z - e0z * e1y); + real_t cy = std::fabs(e0z * e1x - e0x * e1z); + real_t cz = std::fabs(e0x * e1y - e0y * e1x); + const real_t epsilon = std::numeric_limits::epsilon(); + // std::cout << "cx " << cx << ", cy " << cy << ", cz " << cz << + // "\n"; + if (cx > epsilon || cy > epsilon || cz > epsilon) { + // std::cout << "corner\n"; + // found a corner + if (cx > cy && cx > cz) { + // std::cout << "pattern0\n"; + } + else { + // std::cout << "axes[0] = 0\n"; + axes[0] = 0; + if (cz > cx && cz > cy) { + // std::cout << "axes[1] = 1\n"; + axes[1] = 1; + } + } + break; + } + } + +#ifdef TINYOBJLOADER_USE_MAPBOX_EARCUT + using Point = std::array; + + // first polyline define the main polygon. + // following polylines define holes(not used in tinyobj). + std::vector > polygon; + + std::vector polyline; + + // Fill polygon data(facevarying vertices). + for (size_t k = 0; k < npolys; k++) { + i0 = face.vertex_indices[k]; + size_t vi0 = size_t(i0.v_idx); + + assert(((3 * vi0 + 2) < v.size())); + + real_t v0x = v[vi0 * 3 + axes[0]]; + real_t v0y = v[vi0 * 3 + axes[1]]; + + polyline.push_back({ v0x, v0y }); + } + + polygon.push_back(polyline); + std::vector indices = mapbox::earcut(polygon); + // => result = 3 * faces, clockwise + + assert(indices.size() % 3 == 0); + + // Reconstruct vertex_index_t + for (size_t k = 0; k < indices.size() / 3; k++) { + { + index_t idx0, idx1, idx2; + idx0.vertex_index = face.vertex_indices[indices[3 * k + 0]].v_idx; + idx0.normal_index = + face.vertex_indices[indices[3 * k + 0]].vn_idx; + idx0.texcoord_index = + face.vertex_indices[indices[3 * k + 0]].vt_idx; + idx1.vertex_index = face.vertex_indices[indices[3 * k + 1]].v_idx; + idx1.normal_index = + face.vertex_indices[indices[3 * k + 1]].vn_idx; + idx1.texcoord_index = + face.vertex_indices[indices[3 * k + 1]].vt_idx; + idx2.vertex_index = face.vertex_indices[indices[3 * k + 2]].v_idx; + idx2.normal_index = + face.vertex_indices[indices[3 * k + 2]].vn_idx; + idx2.texcoord_index = + face.vertex_indices[indices[3 * k + 2]].vt_idx; + + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx2); + + shape->mesh.num_face_vertices.push_back(3); + shape->mesh.material_ids.push_back(material_id); + shape->mesh.smoothing_group_ids.push_back( + face.smoothing_group_id); + } + } + +#else // Built-in ear clipping triangulation + + + face_t remainingFace = face; // copy + size_t guess_vert = 0; + vertex_index_t ind[3]; + real_t vx[3]; + real_t vy[3]; + + // How many iterations can we do without decreasing the remaining + // vertices. + size_t remainingIterations = face.vertex_indices.size(); + size_t previousRemainingVertices = + remainingFace.vertex_indices.size(); + + while (remainingFace.vertex_indices.size() > 3 && + remainingIterations > 0) { + // std::cout << "remainingIterations " << remainingIterations << + // "\n"; + + npolys = remainingFace.vertex_indices.size(); + if (guess_vert >= npolys) { + guess_vert -= npolys; + } + + if (previousRemainingVertices != npolys) { + // The number of remaining vertices decreased. Reset counters. + previousRemainingVertices = npolys; + remainingIterations = npolys; + } + else { + // We didn't consume a vertex on previous iteration, reduce the + // available iterations. + remainingIterations--; + } + + for (size_t k = 0; k < 3; k++) { + ind[k] = remainingFace.vertex_indices[(guess_vert + k) % npolys]; + size_t vi = size_t(ind[k].v_idx); + if (((vi * 3 + axes[0]) >= v.size()) || + ((vi * 3 + axes[1]) >= v.size())) { + // ??? + vx[k] = static_cast(0.0); + vy[k] = static_cast(0.0); + } + else { + vx[k] = v[vi * 3 + axes[0]]; + vy[k] = v[vi * 3 + axes[1]]; + } + } + + // + // area is calculated per face + // + real_t e0x = vx[1] - vx[0]; + real_t e0y = vy[1] - vy[0]; + real_t e1x = vx[2] - vx[1]; + real_t e1y = vy[2] - vy[1]; + real_t cross = e0x * e1y - e0y * e1x; + // std::cout << "axes = " << axes[0] << ", " << axes[1] << "\n"; + // std::cout << "e0x, e0y, e1x, e1y " << e0x << ", " << e0y << ", " + // << e1x << ", " << e1y << "\n"; + + real_t area = (vx[0] * vy[1] - vy[0] * vx[1]) * static_cast(0.5); + // std::cout << "cross " << cross << ", area " << area << "\n"; + // if an internal angle + if (cross * area < static_cast(0.0)) { + // std::cout << "internal \n"; + guess_vert += 1; + // std::cout << "guess vert : " << guess_vert << "\n"; + continue; + } + + // check all other verts in case they are inside this triangle + bool overlap = false; + for (size_t otherVert = 3; otherVert < npolys; ++otherVert) { + size_t idx = (guess_vert + otherVert) % npolys; + + if (idx >= remainingFace.vertex_indices.size()) { + // std::cout << "???0\n"; + // ??? + continue; + } + + size_t ovi = size_t(remainingFace.vertex_indices[idx].v_idx); + + if (((ovi * 3 + axes[0]) >= v.size()) || + ((ovi * 3 + axes[1]) >= v.size())) { + // std::cout << "???1\n"; + // ??? + continue; + } + real_t tx = v[ovi * 3 + axes[0]]; + real_t ty = v[ovi * 3 + axes[1]]; + if (pnpoly(3, vx, vy, tx, ty)) { + // std::cout << "overlap\n"; + overlap = true; + break; + } + } + + if (overlap) { + // std::cout << "overlap2\n"; + guess_vert += 1; + continue; + } + + // this triangle is an ear + { + index_t idx0, idx1, idx2; + idx0.vertex_index = ind[0].v_idx; + idx0.normal_index = ind[0].vn_idx; + idx0.texcoord_index = ind[0].vt_idx; + idx1.vertex_index = ind[1].v_idx; + idx1.normal_index = ind[1].vn_idx; + idx1.texcoord_index = ind[1].vt_idx; + idx2.vertex_index = ind[2].v_idx; + idx2.normal_index = ind[2].vn_idx; + idx2.texcoord_index = ind[2].vt_idx; + + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx2); + + shape->mesh.num_face_vertices.push_back(3); + shape->mesh.material_ids.push_back(material_id); + shape->mesh.smoothing_group_ids.push_back( + face.smoothing_group_id); + } + + // remove v1 from the list + size_t removed_vert_index = (guess_vert + 1) % npolys; + while (removed_vert_index + 1 < npolys) { + remainingFace.vertex_indices[removed_vert_index] = + remainingFace.vertex_indices[removed_vert_index + 1]; + removed_vert_index += 1; + } + remainingFace.vertex_indices.pop_back(); + } + + // std::cout << "remainingFace.vi.size = " << + // remainingFace.vertex_indices.size() << "\n"; + if (remainingFace.vertex_indices.size() == 3) { + i0 = remainingFace.vertex_indices[0]; + i1 = remainingFace.vertex_indices[1]; + i2 = remainingFace.vertex_indices[2]; + { + index_t idx0, idx1, idx2; + idx0.vertex_index = i0.v_idx; + idx0.normal_index = i0.vn_idx; + idx0.texcoord_index = i0.vt_idx; + idx1.vertex_index = i1.v_idx; + idx1.normal_index = i1.vn_idx; + idx1.texcoord_index = i1.vt_idx; + idx2.vertex_index = i2.v_idx; + idx2.normal_index = i2.vn_idx; + idx2.texcoord_index = i2.vt_idx; + + shape->mesh.indices.push_back(idx0); + shape->mesh.indices.push_back(idx1); + shape->mesh.indices.push_back(idx2); + + shape->mesh.num_face_vertices.push_back(3); + shape->mesh.material_ids.push_back(material_id); + shape->mesh.smoothing_group_ids.push_back( + face.smoothing_group_id); + } + } +#endif + } // npolys + } + else { + for (size_t k = 0; k < npolys; k++) { + index_t idx; + idx.vertex_index = face.vertex_indices[k].v_idx; + idx.normal_index = face.vertex_indices[k].vn_idx; + idx.texcoord_index = face.vertex_indices[k].vt_idx; + shape->mesh.indices.push_back(idx); + } + + shape->mesh.num_face_vertices.push_back( + static_cast(npolys)); + shape->mesh.material_ids.push_back(material_id); // per face + shape->mesh.smoothing_group_ids.push_back( + face.smoothing_group_id); // per face + } + } + + shape->mesh.tags = tags; + } + + // line + if (!prim_group.lineGroup.empty()) { + // Flatten indices + for (size_t i = 0; i < prim_group.lineGroup.size(); i++) { + for (size_t j = 0; j < prim_group.lineGroup[i].vertex_indices.size(); + j++) { + const vertex_index_t& vi = prim_group.lineGroup[i].vertex_indices[j]; + + index_t idx; + idx.vertex_index = vi.v_idx; + idx.normal_index = vi.vn_idx; + idx.texcoord_index = vi.vt_idx; + + shape->lines.indices.push_back(idx); + } + + shape->lines.num_line_vertices.push_back( + int(prim_group.lineGroup[i].vertex_indices.size())); + } + } + + // points + if (!prim_group.pointsGroup.empty()) { + // Flatten & convert indices + for (size_t i = 0; i < prim_group.pointsGroup.size(); i++) { + for (size_t j = 0; j < prim_group.pointsGroup[i].vertex_indices.size(); + j++) { + const vertex_index_t& vi = prim_group.pointsGroup[i].vertex_indices[j]; + + index_t idx; + idx.vertex_index = vi.v_idx; + idx.normal_index = vi.vn_idx; + idx.texcoord_index = vi.vt_idx; + + shape->points.indices.push_back(idx); + } + } + } + + return true; + } + + // Split a string with specified delimiter character and escape character. + // https://rosettacode.org/wiki/Tokenize_a_string_with_escaping#C.2B.2B + static void SplitString(const std::string& s, char delim, char escape, + std::vector& elems) { + std::string token; + + bool escaping = false; + for (size_t i = 0; i < s.size(); ++i) { + char ch = s[i]; + if (escaping) { + escaping = false; + } + else if (ch == escape) { + escaping = true; + continue; + } + else if (ch == delim) { + if (!token.empty()) { + elems.push_back(token); + } + token.clear(); + continue; + } + token += ch; + } + + elems.push_back(token); + } + + static std::string JoinPath(const std::string& dir, + const std::string& filename) { + if (dir.empty()) { + return filename; + } + else { + // check '/' + char lastChar = *dir.rbegin(); + if (lastChar != '/') { + return dir + std::string("/") + filename; + } + else { + return dir + filename; + } + } + } + + void LoadMtl(std::map* material_map, + std::vector* materials, std::istream* inStream, + std::string* warning, std::string* err) { + (void)err; + + // Create a default material anyway. + material_t material; + InitMaterial(&material); + + // Issue 43. `d` wins against `Tr` since `Tr` is not in the MTL specification. + bool has_d = false; + bool has_tr = false; + + // has_kd is used to set a default diffuse value when map_Kd is present + // and Kd is not. + bool has_kd = false; + + std::stringstream warn_ss; + + size_t line_no = 0; + std::string linebuf; + while (inStream->peek() != -1) { + safeGetline(*inStream, linebuf); + line_no++; + + // Trim trailing whitespace. + if (linebuf.size() > 0) { + linebuf = linebuf.substr(0, linebuf.find_last_not_of(" \t") + 1); + } + + // Trim newline '\r\n' or '\n' + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\n') + linebuf.erase(linebuf.size() - 1); + } + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\r') + linebuf.erase(linebuf.size() - 1); + } + + // Skip if empty line. + if (linebuf.empty()) { + continue; + } + + // Skip leading space. + const char* token = linebuf.c_str(); + token += strspn(token, " \t"); + + assert(token); + if (token[0] == '\0') continue; // empty line + + if (token[0] == '#') continue; // comment line + + // new mtl + if ((0 == strncmp(token, "newmtl", 6)) && IS_SPACE((token[6]))) { + // flush previous material. + if (!material.name.empty()) { + material_map->insert(std::pair( + material.name, static_cast(materials->size()))); + materials->push_back(material); + } + + // initial temporary material + InitMaterial(&material); + + has_d = false; + has_tr = false; + + // set new mtl name + token += 7; + { + std::stringstream sstr; + sstr << token; + material.name = sstr.str(); + } + continue; + } + + // ambient + if (token[0] == 'K' && token[1] == 'a' && IS_SPACE((token[2]))) { + token += 2; + real_t r, g, b; + parseReal3(&r, &g, &b, &token); + material.ambient[0] = r; + material.ambient[1] = g; + material.ambient[2] = b; + continue; + } + + // diffuse + if (token[0] == 'K' && token[1] == 'd' && IS_SPACE((token[2]))) { + token += 2; + real_t r, g, b; + parseReal3(&r, &g, &b, &token); + material.diffuse[0] = r; + material.diffuse[1] = g; + material.diffuse[2] = b; + has_kd = true; + continue; + } + + // specular + if (token[0] == 'K' && token[1] == 's' && IS_SPACE((token[2]))) { + token += 2; + real_t r, g, b; + parseReal3(&r, &g, &b, &token); + material.specular[0] = r; + material.specular[1] = g; + material.specular[2] = b; + continue; + } + + // transmittance + if ((token[0] == 'K' && token[1] == 't' && IS_SPACE((token[2]))) || + (token[0] == 'T' && token[1] == 'f' && IS_SPACE((token[2])))) { + token += 2; + real_t r, g, b; + parseReal3(&r, &g, &b, &token); + material.transmittance[0] = r; + material.transmittance[1] = g; + material.transmittance[2] = b; + continue; + } + + // ior(index of refraction) + if (token[0] == 'N' && token[1] == 'i' && IS_SPACE((token[2]))) { + token += 2; + material.ior = parseReal(&token); + continue; + } + + // emission + if (token[0] == 'K' && token[1] == 'e' && IS_SPACE(token[2])) { + token += 2; + real_t r, g, b; + parseReal3(&r, &g, &b, &token); + material.emission[0] = r; + material.emission[1] = g; + material.emission[2] = b; + continue; + } + + // shininess + if (token[0] == 'N' && token[1] == 's' && IS_SPACE(token[2])) { + token += 2; + material.shininess = parseReal(&token); + continue; + } + + // illum model + if (0 == strncmp(token, "illum", 5) && IS_SPACE(token[5])) { + token += 6; + material.illum = parseInt(&token); + continue; + } + + // dissolve + if ((token[0] == 'd' && IS_SPACE(token[1]))) { + token += 1; + material.dissolve = parseReal(&token); + + if (has_tr) { + warn_ss << "Both `d` and `Tr` parameters defined for \"" + << material.name + << "\". Use the value of `d` for dissolve (line " << line_no + << " in .mtl.)\n"; + } + has_d = true; + continue; + } + if (token[0] == 'T' && token[1] == 'r' && IS_SPACE(token[2])) { + token += 2; + if (has_d) { + // `d` wins. Ignore `Tr` value. + warn_ss << "Both `d` and `Tr` parameters defined for \"" + << material.name + << "\". Use the value of `d` for dissolve (line " << line_no + << " in .mtl.)\n"; + } + else { + // We invert value of Tr(assume Tr is in range [0, 1]) + // NOTE: Interpretation of Tr is application(exporter) dependent. For + // some application(e.g. 3ds max obj exporter), Tr = d(Issue 43) + material.dissolve = static_cast(1.0) - parseReal(&token); + } + has_tr = true; + continue; + } + + // PBR: roughness + if (token[0] == 'P' && token[1] == 'r' && IS_SPACE(token[2])) { + token += 2; + material.roughness = parseReal(&token); + continue; + } + + // PBR: metallic + if (token[0] == 'P' && token[1] == 'm' && IS_SPACE(token[2])) { + token += 2; + material.metallic = parseReal(&token); + continue; + } + + // PBR: sheen + if (token[0] == 'P' && token[1] == 's' && IS_SPACE(token[2])) { + token += 2; + material.sheen = parseReal(&token); + continue; + } + + // PBR: clearcoat thickness + if (token[0] == 'P' && token[1] == 'c' && IS_SPACE(token[2])) { + token += 2; + material.clearcoat_thickness = parseReal(&token); + continue; + } + + // PBR: clearcoat roughness + if ((0 == strncmp(token, "Pcr", 3)) && IS_SPACE(token[3])) { + token += 4; + material.clearcoat_roughness = parseReal(&token); + continue; + } + + // PBR: anisotropy + if ((0 == strncmp(token, "aniso", 5)) && IS_SPACE(token[5])) { + token += 6; + material.anisotropy = parseReal(&token); + continue; + } + + // PBR: anisotropy rotation + if ((0 == strncmp(token, "anisor", 6)) && IS_SPACE(token[6])) { + token += 7; + material.anisotropy_rotation = parseReal(&token); + continue; + } + + // ambient texture + if ((0 == strncmp(token, "map_Ka", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.ambient_texname), + &(material.ambient_texopt), token); + continue; + } + + // diffuse texture + if ((0 == strncmp(token, "map_Kd", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.diffuse_texname), + &(material.diffuse_texopt), token); + + // Set a decent diffuse default value if a diffuse texture is specified + // without a matching Kd value. + if (!has_kd) { + material.diffuse[0] = static_cast(0.6); + material.diffuse[1] = static_cast(0.6); + material.diffuse[2] = static_cast(0.6); + } + + continue; + } + + // specular texture + if ((0 == strncmp(token, "map_Ks", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.specular_texname), + &(material.specular_texopt), token); + continue; + } + + // specular highlight texture + if ((0 == strncmp(token, "map_Ns", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.specular_highlight_texname), + &(material.specular_highlight_texopt), token); + continue; + } + + // bump texture + if ((0 == strncmp(token, "map_bump", 8)) && IS_SPACE(token[8])) { + token += 9; + ParseTextureNameAndOption(&(material.bump_texname), + &(material.bump_texopt), token); + continue; + } + + // bump texture + if ((0 == strncmp(token, "map_Bump", 8)) && IS_SPACE(token[8])) { + token += 9; + ParseTextureNameAndOption(&(material.bump_texname), + &(material.bump_texopt), token); + continue; + } + + // bump texture + if ((0 == strncmp(token, "bump", 4)) && IS_SPACE(token[4])) { + token += 5; + ParseTextureNameAndOption(&(material.bump_texname), + &(material.bump_texopt), token); + continue; + } + + // alpha texture + if ((0 == strncmp(token, "map_d", 5)) && IS_SPACE(token[5])) { + token += 6; + material.alpha_texname = token; + ParseTextureNameAndOption(&(material.alpha_texname), + &(material.alpha_texopt), token); + continue; + } + + // displacement texture + if ((0 == strncmp(token, "disp", 4)) && IS_SPACE(token[4])) { + token += 5; + ParseTextureNameAndOption(&(material.displacement_texname), + &(material.displacement_texopt), token); + continue; + } + + // reflection map + if ((0 == strncmp(token, "refl", 4)) && IS_SPACE(token[4])) { + token += 5; + ParseTextureNameAndOption(&(material.reflection_texname), + &(material.reflection_texopt), token); + continue; + } + + // PBR: roughness texture + if ((0 == strncmp(token, "map_Pr", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.roughness_texname), + &(material.roughness_texopt), token); + continue; + } + + // PBR: metallic texture + if ((0 == strncmp(token, "map_Pm", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.metallic_texname), + &(material.metallic_texopt), token); + continue; + } + + // PBR: sheen texture + if ((0 == strncmp(token, "map_Ps", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.sheen_texname), + &(material.sheen_texopt), token); + continue; + } + + // PBR: emissive texture + if ((0 == strncmp(token, "map_Ke", 6)) && IS_SPACE(token[6])) { + token += 7; + ParseTextureNameAndOption(&(material.emissive_texname), + &(material.emissive_texopt), token); + continue; + } + + // PBR: normal map texture + if ((0 == strncmp(token, "norm", 4)) && IS_SPACE(token[4])) { + token += 5; + ParseTextureNameAndOption(&(material.normal_texname), + &(material.normal_texopt), token); + continue; + } + + // unknown parameter + const char* _space = strchr(token, ' '); + if (!_space) { + _space = strchr(token, '\t'); + } + if (_space) { + std::ptrdiff_t len = _space - token; + std::string key(token, static_cast(len)); + std::string value = _space + 1; + material.unknown_parameter.insert( + std::pair(key, value)); + } + } + // flush last material. + material_map->insert(std::pair( + material.name, static_cast(materials->size()))); + materials->push_back(material); + + if (warning) { + (*warning) = warn_ss.str(); + } + } + + bool MaterialFileReader::operator()(const std::string& matId, + std::vector* materials, + std::map* matMap, + std::string* warn, std::string* err) { + if (!m_mtlBaseDir.empty()) { +#ifdef _WIN32 + char sep = ';'; +#else + char sep = ':'; +#endif + + // https://stackoverflow.com/questions/5167625/splitting-a-c-stdstring-using-tokens-e-g + std::vector paths; + std::istringstream f(m_mtlBaseDir); + + std::string s; + while (getline(f, s, sep)) { + paths.push_back(s); + } + + for (size_t i = 0; i < paths.size(); i++) { + std::string filepath = JoinPath(paths[i], matId); + + std::ifstream matIStream(filepath.c_str()); + if (matIStream) { + LoadMtl(matMap, materials, &matIStream, warn, err); + + return true; + } + } + + std::stringstream ss; + ss << "Material file [ " << matId + << " ] not found in a path : " << m_mtlBaseDir << "\n"; + if (warn) { + (*warn) += ss.str(); + } + return false; + + } + else { + std::string filepath = matId; + std::ifstream matIStream(filepath.c_str()); + if (matIStream) { + LoadMtl(matMap, materials, &matIStream, warn, err); + + return true; + } + + std::stringstream ss; + ss << "Material file [ " << filepath + << " ] not found in a path : " << m_mtlBaseDir << "\n"; + if (warn) { + (*warn) += ss.str(); + } + + return false; + } + } + + bool MaterialStreamReader::operator()(const std::string& matId, + std::vector* materials, + std::map* matMap, + std::string* warn, std::string* err) { + (void)err; + (void)matId; + if (!m_inStream) { + std::stringstream ss; + ss << "Material stream in error state. \n"; + if (warn) { + (*warn) += ss.str(); + } + return false; + } + + LoadMtl(matMap, materials, &m_inStream, warn, err); + + return true; + } + + bool LoadObj(attrib_t* attrib, std::vector* shapes, + std::vector* materials, std::string* warn, + std::string* err, const char* filename, const char* mtl_basedir, + bool triangulate, bool default_vcols_fallback) { + attrib->vertices.clear(); + attrib->normals.clear(); + attrib->texcoords.clear(); + attrib->colors.clear(); + shapes->clear(); + + std::stringstream errss; + + std::ifstream ifs(filename); + if (!ifs) { + errss << "Cannot open file [" << filename << "]\n"; + if (err) { + (*err) = errss.str(); + } + return false; + } + + std::string baseDir = mtl_basedir ? mtl_basedir : ""; + if (!baseDir.empty()) { +#ifndef _WIN32 + const char dirsep = '/'; +#else + const char dirsep = '\\'; +#endif + if (baseDir[baseDir.length() - 1] != dirsep) baseDir += dirsep; + } + MaterialFileReader matFileReader(baseDir); + + return LoadObj(attrib, shapes, materials, warn, err, &ifs, &matFileReader, + triangulate, default_vcols_fallback); + } + + bool LoadObj(attrib_t* attrib, std::vector* shapes, + std::vector* materials, std::string* warn, + std::string* err, std::istream* inStream, + MaterialReader* readMatFn /*= NULL*/, bool triangulate, + bool default_vcols_fallback) { + std::stringstream errss; + + std::vector v; + std::vector vn; + std::vector vt; + std::vector vc; + std::vector vw; + std::vector tags; + PrimGroup prim_group; + std::string name; + + // material + std::map material_map; + int material = -1; + + // smoothing group id + unsigned int current_smoothing_id = + 0; // Initial value. 0 means no smoothing. + + int greatest_v_idx = -1; + int greatest_vn_idx = -1; + int greatest_vt_idx = -1; + + shape_t shape; + + bool found_all_colors = true; + + size_t line_num = 0; + std::string linebuf; + while (inStream->peek() != -1) { + safeGetline(*inStream, linebuf); + + line_num++; + + // Trim newline '\r\n' or '\n' + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\n') + linebuf.erase(linebuf.size() - 1); + } + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\r') + linebuf.erase(linebuf.size() - 1); + } + + // Skip if empty line. + if (linebuf.empty()) { + continue; + } + + // Skip leading space. + const char* token = linebuf.c_str(); + token += strspn(token, " \t"); + + assert(token); + if (token[0] == '\0') continue; // empty line + + if (token[0] == '#') continue; // comment line + + // vertex + if (token[0] == 'v' && IS_SPACE((token[1]))) { + token += 2; + real_t x, y, z; + real_t r, g, b; + + found_all_colors &= parseVertexWithColor(&x, &y, &z, &r, &g, &b, &token); + + v.push_back(x); + v.push_back(y); + v.push_back(z); + + if (found_all_colors || default_vcols_fallback) { + vc.push_back(r); + vc.push_back(g); + vc.push_back(b); + } + + continue; + } + + // normal + if (token[0] == 'v' && token[1] == 'n' && IS_SPACE((token[2]))) { + token += 3; + real_t x, y, z; + parseReal3(&x, &y, &z, &token); + vn.push_back(x); + vn.push_back(y); + vn.push_back(z); + continue; + } + + // texcoord + if (token[0] == 'v' && token[1] == 't' && IS_SPACE((token[2]))) { + token += 3; + real_t x, y; + parseReal2(&x, &y, &token); + vt.push_back(x); + vt.push_back(y); + continue; + } + + // skin weight. tinyobj extension + if (token[0] == 'v' && token[1] == 'w' && IS_SPACE((token[2]))) { + token += 3; + + // vw ... + // example: + // vw 0 0 0.25 1 0.25 2 0.5 + + // TODO(syoyo): Add syntax check + int vid = 0; + vid = parseInt(&token); + + skin_weight_t sw; + + sw.vertex_id = vid; + + while (!IS_NEW_LINE(token[0])) { + real_t j, w; + // joint_id should not be negative, weight may be negative + // TODO(syoyo): # of elements check + parseReal2(&j, &w, &token, -1.0); + + if (j < static_cast(0)) { + if (err) { + std::stringstream ss; + ss << "Failed parse `vw' line. joint_id is negative. " + "line " + << line_num << ".)\n"; + (*err) += ss.str(); + } + return false; + } + + joint_and_weight_t jw; + + jw.joint_id = int(j); + jw.weight = w; + + sw.weightValues.push_back(jw); + + size_t n = strspn(token, " \t\r"); + token += n; + } + + vw.push_back(sw); + } + + // line + if (token[0] == 'l' && IS_SPACE((token[1]))) { + token += 2; + + __line_t line; + + while (!IS_NEW_LINE(token[0])) { + vertex_index_t vi; + if (!parseTriple(&token, static_cast(v.size() / 3), + static_cast(vn.size() / 3), + static_cast(vt.size() / 2), &vi)) { + if (err) { + std::stringstream ss; + ss << "Failed parse `l' line(e.g. zero value for vertex index. " + "line " + << line_num << ".)\n"; + (*err) += ss.str(); + } + return false; + } + + line.vertex_indices.push_back(vi); + + size_t n = strspn(token, " \t\r"); + token += n; + } + + prim_group.lineGroup.push_back(line); + + continue; + } + + // points + if (token[0] == 'p' && IS_SPACE((token[1]))) { + token += 2; + + __points_t pts; + + while (!IS_NEW_LINE(token[0])) { + vertex_index_t vi; + if (!parseTriple(&token, static_cast(v.size() / 3), + static_cast(vn.size() / 3), + static_cast(vt.size() / 2), &vi)) { + if (err) { + std::stringstream ss; + ss << "Failed parse `p' line(e.g. zero value for vertex index. " + "line " + << line_num << ".)\n"; + (*err) += ss.str(); + } + return false; + } + + pts.vertex_indices.push_back(vi); + + size_t n = strspn(token, " \t\r"); + token += n; + } + + prim_group.pointsGroup.push_back(pts); + + continue; + } + + // face + if (token[0] == 'f' && IS_SPACE((token[1]))) { + token += 2; + token += strspn(token, " \t"); + + face_t face; + + face.smoothing_group_id = current_smoothing_id; + face.vertex_indices.reserve(3); + + while (!IS_NEW_LINE(token[0])) { + vertex_index_t vi; + if (!parseTriple(&token, static_cast(v.size() / 3), + static_cast(vn.size() / 3), + static_cast(vt.size() / 2), &vi)) { + if (err) { + std::stringstream ss; + ss << "Failed parse `f' line(e.g. zero value for face index. line " + << line_num << ".)\n"; + (*err) += ss.str(); + } + return false; + } + + greatest_v_idx = greatest_v_idx > vi.v_idx ? greatest_v_idx : vi.v_idx; + greatest_vn_idx = + greatest_vn_idx > vi.vn_idx ? greatest_vn_idx : vi.vn_idx; + greatest_vt_idx = + greatest_vt_idx > vi.vt_idx ? greatest_vt_idx : vi.vt_idx; + + face.vertex_indices.push_back(vi); + size_t n = strspn(token, " \t\r"); + token += n; + } + + // replace with emplace_back + std::move on C++11 + prim_group.faceGroup.push_back(face); + + continue; + } + + // use mtl + if ((0 == strncmp(token, "usemtl", 6))) { + token += 6; + std::string namebuf = parseString(&token); + + int newMaterialId = -1; + std::map::const_iterator it = + material_map.find(namebuf); + if (it != material_map.end()) { + newMaterialId = it->second; + } + else { + // { error!! material not found } + if (warn) { + (*warn) += "material [ '" + namebuf + "' ] not found in .mtl\n"; + } + } + + if (newMaterialId != material) { + // Create per-face material. Thus we don't add `shape` to `shapes` at + // this time. + // just clear `faceGroup` after `exportGroupsToShape()` call. + exportGroupsToShape(&shape, prim_group, tags, material, name, + triangulate, v, warn); + prim_group.faceGroup.clear(); + material = newMaterialId; + } + + continue; + } + + // load mtl + if ((0 == strncmp(token, "mtllib", 6)) && IS_SPACE((token[6]))) { + if (readMatFn) { + token += 7; + + std::vector filenames; + SplitString(std::string(token), ' ', '\\', filenames); + + if (filenames.empty()) { + if (warn) { + std::stringstream ss; + ss << "Looks like empty filename for mtllib. Use default " + "material (line " + << line_num << ".)\n"; + + (*warn) += ss.str(); + } + } + else { + bool found = false; + for (size_t s = 0; s < filenames.size(); s++) { + std::string warn_mtl; + std::string err_mtl; + bool ok = (*readMatFn)(filenames[s].c_str(), materials, + &material_map, &warn_mtl, &err_mtl); + if (warn && (!warn_mtl.empty())) { + (*warn) += warn_mtl; + } + + if (err && (!err_mtl.empty())) { + (*err) += err_mtl; + } + + if (ok) { + found = true; + break; + } + } + + if (!found) { + if (warn) { + (*warn) += + "Failed to load material file(s). Use default " + "material.\n"; + } + } + } + } + + continue; + } + + // group name + if (token[0] == 'g' && IS_SPACE((token[1]))) { + // flush previous face group. + bool ret = exportGroupsToShape(&shape, prim_group, tags, material, name, + triangulate, v, warn); + (void)ret; // return value not used. + + if (shape.mesh.indices.size() > 0) { + shapes->push_back(shape); + } + + shape = shape_t(); + + // material = -1; + prim_group.clear(); + + std::vector names; + + while (!IS_NEW_LINE(token[0])) { + std::string str = parseString(&token); + names.push_back(str); + token += strspn(token, " \t\r"); // skip tag + } + + // names[0] must be 'g' + + if (names.size() < 2) { + // 'g' with empty names + if (warn) { + std::stringstream ss; + ss << "Empty group name. line: " << line_num << "\n"; + (*warn) += ss.str(); + name = ""; + } + } + else { + std::stringstream ss; + ss << names[1]; + + // tinyobjloader does not support multiple groups for a primitive. + // Currently we concatinate multiple group names with a space to get + // single group name. + + for (size_t i = 2; i < names.size(); i++) { + ss << " " << names[i]; + } + + name = ss.str(); + } + + continue; + } + + // object name + if (token[0] == 'o' && IS_SPACE((token[1]))) { + // flush previous face group. + bool ret = exportGroupsToShape(&shape, prim_group, tags, material, name, + triangulate, v, warn); + (void)ret; // return value not used. + + if (shape.mesh.indices.size() > 0 || shape.lines.indices.size() > 0 || + shape.points.indices.size() > 0) { + shapes->push_back(shape); + } + + // material = -1; + prim_group.clear(); + shape = shape_t(); + + // @todo { multiple object name? } + token += 2; + std::stringstream ss; + ss << token; + name = ss.str(); + + continue; + } + + if (token[0] == 't' && IS_SPACE(token[1])) { + const int max_tag_nums = 8192; // FIXME(syoyo): Parameterize. + tag_t tag; + + token += 2; + + tag.name = parseString(&token); + + tag_sizes ts = parseTagTriple(&token); + + if (ts.num_ints < 0) { + ts.num_ints = 0; + } + if (ts.num_ints > max_tag_nums) { + ts.num_ints = max_tag_nums; + } + + if (ts.num_reals < 0) { + ts.num_reals = 0; + } + if (ts.num_reals > max_tag_nums) { + ts.num_reals = max_tag_nums; + } + + if (ts.num_strings < 0) { + ts.num_strings = 0; + } + if (ts.num_strings > max_tag_nums) { + ts.num_strings = max_tag_nums; + } + + tag.intValues.resize(static_cast(ts.num_ints)); + + for (size_t i = 0; i < static_cast(ts.num_ints); ++i) { + tag.intValues[i] = parseInt(&token); + } + + tag.floatValues.resize(static_cast(ts.num_reals)); + for (size_t i = 0; i < static_cast(ts.num_reals); ++i) { + tag.floatValues[i] = parseReal(&token); + } + + tag.stringValues.resize(static_cast(ts.num_strings)); + for (size_t i = 0; i < static_cast(ts.num_strings); ++i) { + tag.stringValues[i] = parseString(&token); + } + + tags.push_back(tag); + + continue; + } + + if (token[0] == 's' && IS_SPACE(token[1])) { + // smoothing group id + token += 2; + + // skip space. + token += strspn(token, " \t"); // skip space + + if (token[0] == '\0') { + continue; + } + + if (token[0] == '\r' || token[1] == '\n') { + continue; + } + + if (strlen(token) >= 3 && token[0] == 'o' && token[1] == 'f' && + token[2] == 'f') { + current_smoothing_id = 0; + } + else { + // assume number + int smGroupId = parseInt(&token); + if (smGroupId < 0) { + // parse error. force set to 0. + // FIXME(syoyo): Report warning. + current_smoothing_id = 0; + } + else { + current_smoothing_id = static_cast(smGroupId); + } + } + + continue; + } // smoothing group id + + // Ignore unknown command. + } + + // not all vertices have colors, no default colors desired? -> clear colors + if (!found_all_colors && !default_vcols_fallback) { + vc.clear(); + } + + if (greatest_v_idx >= static_cast(v.size() / 3)) { + if (warn) { + std::stringstream ss; + ss << "Vertex indices out of bounds (line " << line_num << ".)\n\n"; + (*warn) += ss.str(); + } + } + if (greatest_vn_idx >= static_cast(vn.size() / 3)) { + if (warn) { + std::stringstream ss; + ss << "Vertex normal indices out of bounds (line " << line_num << ".)\n\n"; + (*warn) += ss.str(); + } + } + if (greatest_vt_idx >= static_cast(vt.size() / 2)) { + if (warn) { + std::stringstream ss; + ss << "Vertex texcoord indices out of bounds (line " << line_num << ".)\n\n"; + (*warn) += ss.str(); + } + } + + bool ret = exportGroupsToShape(&shape, prim_group, tags, material, name, + triangulate, v, warn); + // exportGroupsToShape return false when `usemtl` is called in the last + // line. + // we also add `shape` to `shapes` when `shape.mesh` has already some + // faces(indices) + if (ret || shape.mesh.indices + .size()) { // FIXME(syoyo): Support other prims(e.g. lines) + shapes->push_back(shape); + } + prim_group.clear(); // for safety + + if (err) { + (*err) += errss.str(); + } + + attrib->vertices.swap(v); + attrib->vertex_weights.swap(v); + attrib->normals.swap(vn); + attrib->texcoords.swap(vt); + attrib->texcoord_ws.swap(vt); + attrib->colors.swap(vc); + attrib->skin_weights.swap(vw); + + return true; + } + + bool LoadObjWithCallback(std::istream& inStream, const callback_t& callback, + void* user_data /*= NULL*/, + MaterialReader* readMatFn /*= NULL*/, + std::string* warn, /* = NULL*/ + std::string* err /*= NULL*/) { + std::stringstream errss; + + // material + std::map material_map; + int material_id = -1; // -1 = invalid + + std::vector indices; + std::vector materials; + std::vector names; + names.reserve(2); + std::vector names_out; + + std::string linebuf; + while (inStream.peek() != -1) { + safeGetline(inStream, linebuf); + + // Trim newline '\r\n' or '\n' + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\n') + linebuf.erase(linebuf.size() - 1); + } + if (linebuf.size() > 0) { + if (linebuf[linebuf.size() - 1] == '\r') + linebuf.erase(linebuf.size() - 1); + } + + // Skip if empty line. + if (linebuf.empty()) { + continue; + } + + // Skip leading space. + const char* token = linebuf.c_str(); + token += strspn(token, " \t"); + + assert(token); + if (token[0] == '\0') continue; // empty line + + if (token[0] == '#') continue; // comment line + + // vertex + if (token[0] == 'v' && IS_SPACE((token[1]))) { + token += 2; + // TODO(syoyo): Support parsing vertex color extension. + real_t x, y, z, w; // w is optional. default = 1.0 + parseV(&x, &y, &z, &w, &token); + if (callback.vertex_cb) { + callback.vertex_cb(user_data, x, y, z, w); + } + continue; + } + + // normal + if (token[0] == 'v' && token[1] == 'n' && IS_SPACE((token[2]))) { + token += 3; + real_t x, y, z; + parseReal3(&x, &y, &z, &token); + if (callback.normal_cb) { + callback.normal_cb(user_data, x, y, z); + } + continue; + } + + // texcoord + if (token[0] == 'v' && token[1] == 't' && IS_SPACE((token[2]))) { + token += 3; + real_t x, y, z; // y and z are optional. default = 0.0 + parseReal3(&x, &y, &z, &token); + if (callback.texcoord_cb) { + callback.texcoord_cb(user_data, x, y, z); + } + continue; + } + + // face + if (token[0] == 'f' && IS_SPACE((token[1]))) { + token += 2; + token += strspn(token, " \t"); + + indices.clear(); + while (!IS_NEW_LINE(token[0])) { + vertex_index_t vi = parseRawTriple(&token); + + index_t idx; + idx.vertex_index = vi.v_idx; + idx.normal_index = vi.vn_idx; + idx.texcoord_index = vi.vt_idx; + + indices.push_back(idx); + size_t n = strspn(token, " \t\r"); + token += n; + } + + if (callback.index_cb && indices.size() > 0) { + callback.index_cb(user_data, &indices.at(0), + static_cast(indices.size())); + } + + continue; + } + + // use mtl + if ((0 == strncmp(token, "usemtl", 6)) && IS_SPACE((token[6]))) { + token += 7; + std::stringstream ss; + ss << token; + std::string namebuf = ss.str(); + + int newMaterialId = -1; + std::map::const_iterator it = + material_map.find(namebuf); + if (it != material_map.end()) { + newMaterialId = it->second; + } + else { + // { warn!! material not found } + if (warn && (!callback.usemtl_cb)) { + (*warn) += "material [ " + namebuf + " ] not found in .mtl\n"; + } + } + + if (newMaterialId != material_id) { + material_id = newMaterialId; + } + + if (callback.usemtl_cb) { + callback.usemtl_cb(user_data, namebuf.c_str(), material_id); + } + + continue; + } + + // load mtl + if ((0 == strncmp(token, "mtllib", 6)) && IS_SPACE((token[6]))) { + if (readMatFn) { + token += 7; + + std::vector filenames; + SplitString(std::string(token), ' ', '\\', filenames); + + if (filenames.empty()) { + if (warn) { + (*warn) += + "Looks like empty filename for mtllib. Use default " + "material. \n"; + } + } + else { + bool found = false; + for (size_t s = 0; s < filenames.size(); s++) { + std::string warn_mtl; + std::string err_mtl; + bool ok = (*readMatFn)(filenames[s].c_str(), &materials, + &material_map, &warn_mtl, &err_mtl); + + if (warn && (!warn_mtl.empty())) { + (*warn) += warn_mtl; // This should be warn message. + } + + if (err && (!err_mtl.empty())) { + (*err) += err_mtl; + } + + if (ok) { + found = true; + break; + } + } + + if (!found) { + if (warn) { + (*warn) += + "Failed to load material file(s). Use default " + "material.\n"; + } + } + else { + if (callback.mtllib_cb) { + callback.mtllib_cb(user_data, &materials.at(0), + static_cast(materials.size())); + } + } + } + } + + continue; + } + + // group name + if (token[0] == 'g' && IS_SPACE((token[1]))) { + names.clear(); + + while (!IS_NEW_LINE(token[0])) { + std::string str = parseString(&token); + names.push_back(str); + token += strspn(token, " \t\r"); // skip tag + } + + assert(names.size() > 0); + + if (callback.group_cb) { + if (names.size() > 1) { + // create const char* array. + names_out.resize(names.size() - 1); + for (size_t j = 0; j < names_out.size(); j++) { + names_out[j] = names[j + 1].c_str(); + } + callback.group_cb(user_data, &names_out.at(0), + static_cast(names_out.size())); + + } + else { + callback.group_cb(user_data, NULL, 0); + } + } + + continue; + } + + // object name + if (token[0] == 'o' && IS_SPACE((token[1]))) { + // @todo { multiple object name? } + token += 2; + + std::stringstream ss; + ss << token; + std::string object_name = ss.str(); + + if (callback.object_cb) { + callback.object_cb(user_data, object_name.c_str()); + } + + continue; + } + +#if 0 // @todo + if (token[0] == 't' && IS_SPACE(token[1])) { + tag_t tag; + + token += 2; + std::stringstream ss; + ss << token; + tag.name = ss.str(); + + token += tag.name.size() + 1; + + tag_sizes ts = parseTagTriple(&token); + + tag.intValues.resize(static_cast(ts.num_ints)); + + for (size_t i = 0; i < static_cast(ts.num_ints); ++i) { + tag.intValues[i] = atoi(token); + token += strcspn(token, "/ \t\r") + 1; + } + + tag.floatValues.resize(static_cast(ts.num_reals)); + for (size_t i = 0; i < static_cast(ts.num_reals); ++i) { + tag.floatValues[i] = parseReal(&token); + token += strcspn(token, "/ \t\r") + 1; + } + + tag.stringValues.resize(static_cast(ts.num_strings)); + for (size_t i = 0; i < static_cast(ts.num_strings); ++i) { + std::stringstream ss; + ss << token; + tag.stringValues[i] = ss.str(); + token += tag.stringValues[i].size() + 1; + } + + tags.push_back(tag); + } +#endif + + // Ignore unknown command. + } + + if (err) { + (*err) += errss.str(); + } + + return true; + } + + bool ObjReader::ParseFromFile(const std::string& filename, + const ObjReaderConfig& config) { + std::string mtl_search_path; + + if (config.mtl_search_path.empty()) { + // + // split at last '/'(for unixish system) or '\\'(for windows) to get + // the base directory of .obj file + // + size_t pos = filename.find_last_of("/\\"); + if (pos != std::string::npos) { + mtl_search_path = filename.substr(0, pos); + } + } + else { + mtl_search_path = config.mtl_search_path; + } + + valid_ = LoadObj(&attrib_, &shapes_, &materials_, &warning_, &error_, + filename.c_str(), mtl_search_path.c_str(), + config.triangulate, config.vertex_color); + + return valid_; + } + + bool ObjReader::ParseFromString(const std::string& obj_text, + const std::string& mtl_text, + const ObjReaderConfig& config) { + std::stringbuf obj_buf(obj_text); + std::stringbuf mtl_buf(mtl_text); + + std::istream obj_ifs(&obj_buf); + std::istream mtl_ifs(&mtl_buf); + + MaterialStreamReader mtl_ss(mtl_ifs); + + valid_ = LoadObj(&attrib_, &shapes_, &materials_, &warning_, &error_, + &obj_ifs, &mtl_ss, config.triangulate, config.vertex_color); + + return valid_; + } + +#ifdef __clang__ +#pragma clang diagnostic pop +#endif +} // namespace tinyobj + +#endif \ No newline at end of file diff --git a/img/100.PNG b/img/100.PNG new file mode 100644 index 0000000..72d57c2 Binary files /dev/null and b/img/100.PNG differ diff --git a/img/2000.PNG b/img/2000.PNG new file mode 100644 index 0000000..0cb2920 Binary files /dev/null and b/img/2000.PNG differ diff --git a/img/add.PNG b/img/add.PNG new file mode 100644 index 0000000..07d4c89 Binary files /dev/null and b/img/add.PNG differ diff --git a/img/bunny58.PNG b/img/bunny58.PNG new file mode 100644 index 0000000..2108ae9 Binary files /dev/null and b/img/bunny58.PNG differ diff --git a/img/bunnySmallF.PNG b/img/bunnySmallF.PNG new file mode 100644 index 0000000..691a4d9 Binary files /dev/null and b/img/bunnySmallF.PNG differ diff --git a/img/denoise100.PNG b/img/denoise100.PNG new file mode 100644 index 0000000..d77d124 Binary files /dev/null and b/img/denoise100.PNG differ diff --git a/img/denoiseTime1.png b/img/denoiseTime1.png new file mode 100644 index 0000000..f28b153 Binary files /dev/null and b/img/denoiseTime1.png differ diff --git a/img/denoiseTime2.png b/img/denoiseTime2.png new file mode 100644 index 0000000..b45b58f Binary files /dev/null and b/img/denoiseTime2.png differ diff --git a/img/filter.png b/img/filter.png new file mode 100644 index 0000000..129e1ea Binary files /dev/null and b/img/filter.png differ diff --git a/img/noisyBunny.PNG b/img/noisyBunny.PNG new file mode 100644 index 0000000..0808484 Binary files /dev/null and b/img/noisyBunny.PNG differ diff --git a/img/resolution.png b/img/resolution.png new file mode 100644 index 0000000..79148b0 Binary files /dev/null and b/img/resolution.png differ diff --git a/img/resolutionDenoise.png b/img/resolutionDenoise.png new file mode 100644 index 0000000..b822208 Binary files /dev/null and b/img/resolutionDenoise.png differ diff --git a/img/sameBuffer.PNG b/img/sameBuffer.PNG new file mode 100644 index 0000000..a550b84 Binary files /dev/null and b/img/sameBuffer.PNG differ diff --git a/img/white.PNG b/img/white.PNG new file mode 100644 index 0000000..d538b4a Binary files /dev/null and b/img/white.PNG differ diff --git a/scenes/bunny.txt b/scenes/bunny.txt new file mode 100644 index 0000000..daf8412 --- /dev/null +++ b/scenes/bunny.txt @@ -0,0 +1,126 @@ +// Emissive material (light) +MATERIAL 0 +RGB 1 1 1 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 5 + +// Diffuse white +MATERIAL 1 +RGB .98 .98 .98 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 0 + +// Diffuse red +MATERIAL 2 +RGB .85 .35 .35 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 0 + +// Diffuse green +MATERIAL 3 +RGB .35 .85 .35 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 0 + +// Specular white +MATERIAL 4 +RGB .98 .98 .98 +SPECEX 0 +SPECRGB .98 .98 .98 +REFL 1 +REFR 0.5 +REFRIOR 0 +EMITTANCE 0 + +// Refractive, specular red +MATERIAL 5 +RGB .85 .35 .35 +SPECEX 0 +SPECRGB .85 .35 .35 +REFL 0 +REFR 1 +REFRIOR 0 +EMITTANCE 0 + +// Camera +CAMERA +RES 800 800 +FOVY 45 +ITERATIONS 5 +DEPTH 8 +FILE bunny +EYE 0.0 5 10.5 +LOOKAT 0 5 0 +UP 0 1 0 + +// Ceiling light +OBJECT 0 +cube +material 0 +TRANS 0 10 0 +ROTAT 0 0 0 +SCALE 9 .3 9 + +// Floor +OBJECT 1 +cube +material 1 +TRANS 0 0 0 +ROTAT 0 0 0 +SCALE 10 .01 10 + +// Ceiling +OBJECT 2 +cube +material 1 +TRANS 0 10 0 +ROTAT 0 0 90 +SCALE .01 10 10 + +// Back wall +OBJECT 3 +cube +material 1 +TRANS 0 5 -5 +ROTAT 0 90 0 +SCALE .01 10 10 + +// Left wall +OBJECT 4 +cube +material 2 +TRANS -5 5 0 +ROTAT 0 0 0 +SCALE .01 10 10 + +// Right wall +OBJECT 5 +cube +material 3 +TRANS 5 5 0 +ROTAT 0 0 0 +SCALE .01 10 10 + +// bunny +OBJECT 6 +mesh +material 2 +TRANS -1 3 -1 +ROTAT 0 0 0 +SCALE 3 3 3 \ No newline at end of file diff --git a/scenes/cloud.txt b/scenes/cloud.txt new file mode 100644 index 0000000..65d6618 --- /dev/null +++ b/scenes/cloud.txt @@ -0,0 +1,96 @@ +// Emissive material (light) +MATERIAL 0 +RGB 1 1 1 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 5 + +// Diffuse white +MATERIAL 1 +RGB .95 .95 .95 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 0 + +// Diffuse grey +MATERIAL 2 +RGB .45 .45 .45 +SPECEX 0 +SPECRGB 0 0 0 +REFL 0 +REFR 0 +REFRIOR 0 +EMITTANCE 0 + +// Camera +CAMERA +RES 800 800 +FOVY 45 +ITERATIONS 5000 +DEPTH 8 +FILE cloud +EYE 0.0 5 10.5 +LOOKAT 0 5 0 +UP 0 1 0 + +// Ceiling light +OBJECT 0 +cube +material 0 +TRANS 0 10 0 +ROTAT 0 0 0 +SCALE 4 .3 4 + +// Floor +OBJECT 1 +cube +material 1 +TRANS 0 0 0 +ROTAT 0 0 0 +SCALE 10 .01 10 + +// Ceiling +OBJECT 2 +cube +material 1 +TRANS 0 10 0 +ROTAT 0 0 90 +SCALE .01 10 10 + +// Back wall +OBJECT 3 +cube +material 1 +TRANS 0 5 -5 +ROTAT 0 90 0 +SCALE .01 10 10 + +// Left wall +OBJECT 4 +cube +material 1 +TRANS -5 5 0 +ROTAT 0 0 0 +SCALE .01 10 10 + +// Right wall +OBJECT 5 +cube +material 1 +TRANS 5 5 0 +ROTAT 0 0 0 +SCALE .01 10 10 + +// cloud +OBJECT 6 +mesh +material 2 +TRANS -1 6 -1 +ROTAT 0 0 0 +SCALE .08 .08 .08 \ No newline at end of file diff --git a/scenes/cornell.txt b/scenes/cornell.txt index 83ff820..ca356d4 100644 --- a/scenes/cornell.txt +++ b/scenes/cornell.txt @@ -52,7 +52,7 @@ EMITTANCE 0 CAMERA RES 800 800 FOVY 45 -ITERATIONS 5000 +ITERATIONS 100 DEPTH 8 FILE cornell EYE 0.0 5 10.5 diff --git a/scenes/cornell_ceiling_light.txt b/scenes/cornell_ceiling_light.txt index 15af5f1..113033d 100644 --- a/scenes/cornell_ceiling_light.txt +++ b/scenes/cornell_ceiling_light.txt @@ -52,7 +52,7 @@ EMITTANCE 0 CAMERA RES 800 800 FOVY 45 -ITERATIONS 10 +ITERATIONS 100 DEPTH 8 FILE cornell EYE 0.0 5 10.5 diff --git a/src/interactions.h b/src/interactions.h index 144a9f5..498e072 100644 --- a/src/interactions.h +++ b/src/interactions.h @@ -45,18 +45,51 @@ glm::vec3 calculateRandomDirectionInHemisphere( */ __host__ __device__ void scatterRay( - PathSegment & pathSegment, - glm::vec3 intersect, - glm::vec3 normal, - const Material &m, - thrust::default_random_engine &rng) { - glm::vec3 newDirection; - if (m.hasReflective) { - newDirection = glm::reflect(pathSegment.ray.direction, normal); - } else { - newDirection = calculateRandomDirectionInHemisphere(normal, rng); + PathSegment& pathSegment, + glm::vec3 intersect, + glm::vec3 normal, + const Material& m, + thrust::default_random_engine& rng) +{ + thrust::uniform_real_distribution u01(0, 1); + float rand = u01(rng); + + if (!m.hasRefractive && !m.hasReflective) { // pure diffuse surface + glm::vec3 randDir = calculateRandomDirectionInHemisphere(normal, rng); + pathSegment.ray.origin = intersect + (0.001f * normal); + pathSegment.ray.direction = glm::normalize(randDir); + pathSegment.color *= m.color; } + else if (!m.hasRefractive && m.hasReflective) { // reflective surface + glm::vec3 reflected = glm::reflect(pathSegment.ray.direction, normal); + pathSegment.ray.origin = intersect + (0.001f * normal); + pathSegment.ray.direction = glm::normalize(reflected); + pathSegment.color *= m.specular.color; + } + else if (m.hasRefractive) { // refractive surface + // Schlick's approximation implementation for the process described in PBRT 8.2 + // https://pbr-book.org/3ed-2018/Reflection_Models/Specular_Reflection_and_Transmission + + // incident ray has component opposite normal so ray outside and need to swap + float eta = glm::dot(pathSegment.ray.direction, normal) < 0 ? 1.0f / m.indexOfRefraction : m.indexOfRefraction; - pathSegment.ray.direction = newDirection; - pathSegment.ray.origin = intersect + (newDirection * 0.0001f); + float cosTheta = glm::dot(-pathSegment.ray.direction, normal) / + (glm::length(pathSegment.ray.direction) * (glm::length(normal))); + + float r0 = glm::pow((1.f - eta) / (1.f + eta), 2); + float schlick = r0 + (1.f - r0) * glm::pow(1 - cosTheta, 5); + + if (schlick < rand) { // refract + pathSegment.ray.direction = glm::normalize(glm::refract(pathSegment.ray.direction, normal, eta)); + pathSegment.color *= m.color; + + pathSegment.ray.origin = intersect - 0.001f * normal; + } + else { //reflect + pathSegment.ray.direction = glm::normalize(glm::reflect(pathSegment.ray.direction, normal)); + pathSegment.color *= m.color; + + pathSegment.ray.origin = intersect + 0.001f * normal; + } + } } diff --git a/src/intersections.h b/src/intersections.h index c3e81f4..b868f17 100644 --- a/src/intersections.h +++ b/src/intersections.h @@ -19,6 +19,7 @@ __host__ __device__ inline unsigned int utilhash(unsigned int a) { return a; } +// CHECKITOUT /** * Compute a point at parameter value `t` on ray `r`. * Falls slightly short so that it doesn't intersect the object it's hitting. @@ -34,6 +35,7 @@ __host__ __device__ glm::vec3 multiplyMV(glm::mat4 m, glm::vec4 v) { return glm::vec3(m * v); } +// CHECKITOUT /** * Test intersection between a ray and a transformed cube. Untransformed, * the cube ranges from -0.5 to 0.5 in each axis and is centered at the origin. @@ -44,9 +46,9 @@ __host__ __device__ glm::vec3 multiplyMV(glm::mat4 m, glm::vec4 v) { * @return Ray parameter `t` value. -1 if no intersection. */ __host__ __device__ float boxIntersectionTest(Geom box, Ray r, - glm::vec3 &intersectionPoint, glm::vec3 &normal, bool &outside) { + glm::vec3& intersectionPoint, glm::vec3& normal, bool& outside) { Ray q; - q.origin = multiplyMV(box.inverseTransform, glm::vec4(r.origin , 1.0f)); + q.origin = multiplyMV(box.inverseTransform, glm::vec4(r.origin, 1.0f)); q.direction = glm::normalize(multiplyMV(box.inverseTransform, glm::vec4(r.direction, 0.0f))); float tmin = -1e38f; @@ -87,6 +89,7 @@ __host__ __device__ float boxIntersectionTest(Geom box, Ray r, return -1; } +// CHECKITOUT /** * Test intersection between a ray and a transformed sphere. Untransformed, * the sphere always has radius 0.5 and is centered at the origin. @@ -97,7 +100,7 @@ __host__ __device__ float boxIntersectionTest(Geom box, Ray r, * @return Ray parameter `t` value. -1 if no intersection. */ __host__ __device__ float sphereIntersectionTest(Geom sphere, Ray r, - glm::vec3 &intersectionPoint, glm::vec3 &normal, bool &outside) { + glm::vec3& intersectionPoint, glm::vec3& normal, bool& outside) { float radius = .5; glm::vec3 ro = multiplyMV(sphere.inverseTransform, glm::vec4(r.origin, 1.0f)); @@ -121,10 +124,12 @@ __host__ __device__ float sphereIntersectionTest(Geom sphere, Ray r, float t = 0; if (t1 < 0 && t2 < 0) { return -1; - } else if (t1 > 0 && t2 > 0) { + } + else if (t1 > 0 && t2 > 0) { t = min(t1, t2); outside = true; - } else { + } + else { t = max(t1, t2); outside = false; } @@ -139,3 +144,141 @@ __host__ __device__ float sphereIntersectionTest(Geom sphere, Ray r, return glm::length(r.origin - intersectionPoint); } + +/** + * Test intersection between a ray and a transformed mesh. + * + * @param intersectionPoint Output parameter for point of intersection. + * @param normal Output parameter for surface normal. + * @param outside Output param for whether the ray came from outside. + * @return Ray parameter `t` value. -1 if no intersection. + */ +__host__ __device__ float meshIntersectionTest(Geom mesh, Triangle* triangles, int numTriangles, Ray r, + glm::vec3& intersectionPoint, glm::vec3& normal, bool& outside) { + + glm::vec3 ro = multiplyMV(mesh.inverseTransform, glm::vec4(r.origin, 1.0f)); + glm::vec3 rd = glm::normalize(multiplyMV(mesh.inverseTransform, glm::vec4(r.direction, 0.0f))); + + Ray rt; + rt.origin = ro; + rt.direction = rd; + float tmin = FLT_MAX; + int currIndex = 0; + + for (int i = 0; i < numTriangles; i++) { + Triangle tri = triangles[i]; + glm::vec3 baryPos; // will hold the intersection point in barycentric coordinates + if (glm::intersectRayTriangle(rt.origin, rt.direction, tri.verts[0], tri.verts[1], tri.verts[2], baryPos)) { + float t = baryPos.z; + if (t > 0.0f && t < tmin) { + tmin = t; + currIndex = i; + } + } + } + + // the closest triangle + glm::vec3 objspaceIntersection = getPointOnRay(rt, tmin); + intersectionPoint = multiplyMV(mesh.transform, glm::vec4(objspaceIntersection, 1.0f)); + glm::vec3 surfaceNormal = glm::normalize(glm::cross( + triangles[currIndex].normals[0] - triangles[currIndex].normals[2], + triangles[currIndex].normals[0] - triangles[currIndex].normals[1])); + normal = glm::normalize(multiplyMV(mesh.transform, glm::vec4(surfaceNormal, 0.0f))); + + if (glm::dot(rt.origin, normal) < 0) { + outside = true; + } + else { + outside = false; + normal *= -1.0f; + } + return glm::length(r.origin - intersectionPoint); +} + +__host__ __device__ bool boundingBoxCheck(Geom box, Ray r, glm::vec3 bottomLeft, glm::vec3 topRight) { + Ray q; + q.origin = multiplyMV(box.inverseTransform, glm::vec4(r.origin, 1.0f)); + q.direction = glm::normalize(multiplyMV(box.inverseTransform, glm::vec4(r.direction, 0.0f))); + + float tmin = -1e38f; + float tmax = 1e38f; + glm::vec3 tmin_n; + glm::vec3 tmax_n; + for (int xyz = 0; xyz < 3; ++xyz) { + float qdxyz = q.direction[xyz]; + /*if (glm::abs(qdxyz) > 0.00001f)*/ { + float t1 = (bottomLeft[xyz] - q.origin[xyz]) / qdxyz; + float t2 = (topRight[xyz] - q.origin[xyz]) / qdxyz; + float ta = glm::min(t1, t2); + float tb = glm::max(t1, t2); + glm::vec3 n; + n[xyz] = t2 < t1 ? +1 : -1; + if (ta > 0 && ta > tmin) { + tmin = ta; + tmin_n = n; + } + if (tb < tmax) { + tmax = tb; + tmax_n = n; + } + } + } + + return tmax >= tmin && tmax > 0 ? true : false; +} + +__host__ __device__ float octreeIntersectionTest(OctreeNode& currNode, Geom mesh, Triangle* triangles, int numTriangles, Ray r, + glm::vec3& intersectionPoint, glm::vec3& normal, bool& outside, bool& isLeaf) { + + if (currNode.start == currNode.end) return -1; // this node doesn't enclose any triangles + isLeaf = currNode.children[0] == -1; + + Geom bound; + bound.type = CUBE; + bound.transform = mesh.transform; + bound.inverseTransform = mesh.inverseTransform; + bound.invTranspose = mesh.invTranspose; + + float t_min = FLT_MAX; + + if (!boundingBoxCheck(bound, r, currNode.bottomLeft, currNode.topRight)) { + return -1; + } + + glm::vec3 ro = multiplyMV(mesh.inverseTransform, glm::vec4(r.origin, 1.0f)); + glm::vec3 rd = glm::normalize(multiplyMV(mesh.inverseTransform, glm::vec4(r.direction, 0.0f))); + + Ray rt; + rt.origin = ro; + rt.direction = rd; + float tmin = FLT_MAX; + int currIndex = 0; + + for (int i = currNode.start; i < currNode.end; i++) { + Triangle tri = triangles[i]; + glm::vec3 baryPos; + if (glm::intersectRayTriangle(rt.origin, rt.direction, tri.verts[0], tri.verts[1], tri.verts[2], baryPos)) { + float t = baryPos.z; + if (t > 0.0f && t < tmin) { + tmin = t; + currIndex = i; + } + } + } + + glm::vec3 objspaceIntersection = getPointOnRay(rt, tmin); + intersectionPoint = multiplyMV(mesh.transform, glm::vec4(objspaceIntersection, 1.0f)); + glm::vec3 surfaceNormal = glm::normalize(glm::cross( + triangles[currIndex].normals[0] - triangles[currIndex].normals[2], + triangles[currIndex].normals[0] - triangles[currIndex].normals[1])); + normal = glm::normalize(multiplyMV(mesh.transform, glm::vec4(surfaceNormal, 0.0f))); + + if (glm::dot(rt.origin, normal) < 0) { + outside = true; + } + else { + outside = false; + normal *= -1.0f; + } + return glm::length(r.origin - intersectionPoint); +} diff --git a/src/main.cpp b/src/main.cpp index 4092ae4..94fcc7a 100644 --- a/src/main.cpp +++ b/src/main.cpp @@ -6,6 +6,10 @@ #include "../imgui/imgui_impl_glfw.h" #include "../imgui/imgui_impl_opengl3.h" +#define LOADOBJ 0 + + + static std::string startTimeString; // For camera controls @@ -60,7 +64,12 @@ int main(int argc, char** argv) { const char *sceneFile = argv[1]; // Load scene file +#if LOADOBJ + scene = new Scene(sceneFile, "../scenes/obj/bunny.obj"); + scene->makeOctree(); +#else scene = new Scene(sceneFile); +#endif // Set up camera stuff from loaded path tracer settings iteration = 0; @@ -162,13 +171,19 @@ void runCuda() { // execute the kernel int frame = 0; + pathtrace(frame, iteration); } if (ui_showGbuffer) { showGBuffer(pbo_dptr); } else { - showImage(pbo_dptr, iteration); + if (ui_denoise) { + showDenoised(pbo_dptr, iteration, ui_filterSize, ui_colorWeight, ui_normalWeight, ui_positionWeight); + } + else { + showImage(pbo_dptr, iteration); + } } // unmap buffer object diff --git a/src/pathtrace.cu b/src/pathtrace.cu index 23e5f90..5f8819d 100644 --- a/src/pathtrace.cu +++ b/src/pathtrace.cu @@ -4,6 +4,8 @@ #include #include #include +#include +#include #include "sceneStructs.h" #include "scene.h" @@ -13,8 +15,27 @@ #include "pathtrace.h" #include "intersections.h" #include "interactions.h" +#include "device_launch_parameters.h" #define ERRORCHECK 1 +#define SORT_MATERIALS 0 +#define CACHE_FIRST_BOUNCE 1 +#define STREAM_COMPACTION 1 + +#define DEPTH_OF_FIELD 0 +#define ANTI_ALIASING 1 + +// only one of BOUNDING_BOX and OCTREE should be 1 at any give time +#define BOUNDING_BOX 1 +#define OCTREE 0 + +// only one of these should be 1 at any give time +#define GBUFFER_T 0 +#define GBUFFER_NORM 0 +#define GBUFFER_POS 1 + +float totalTime = 0; + #define FILENAME (strrchr(__FILE__, '/') ? strrchr(__FILE__, '/') + 1 : __FILE__) #define checkCUDAError(msg) checkCUDAErrorFn(msg, FILENAME, __LINE__) @@ -38,6 +59,12 @@ void checkCUDAErrorFn(const char *msg, const char *file, int line) { #endif } +PerformanceTimer& timer() +{ + static PerformanceTimer timer; + return timer; +} + __host__ __device__ thrust::default_random_engine makeSeededRandomEngine(int iter, int index, int depth) { int h = utilhash((1 << 31) | (depth << 22) | iter) ^ utilhash(index); @@ -67,12 +94,34 @@ __global__ void sendImageToPBO(uchar4* pbo, glm::ivec2 resolution, } } -__global__ void gbufferToPBO(uchar4* pbo, glm::ivec2 resolution, GBufferPixel* gBuffer) { +__global__ void sendDenoiseToPBO(uchar4* pbo, glm::ivec2 resolution, glm::vec3* image) { int x = (blockIdx.x * blockDim.x) + threadIdx.x; int y = (blockIdx.y * blockDim.y) + threadIdx.y; if (x < resolution.x && y < resolution.y) { int index = x + (y * resolution.x); + glm::vec3 pix = image[index]; + + glm::ivec3 color; + color.x = glm::clamp((int)(pix.x * 255.0), 0, 255); + color.y = glm::clamp((int)(pix.y * 255.0), 0, 255); + color.z = glm::clamp((int)(pix.z * 255.0), 0, 255); + + // Each thread writes one pixel location in the texture (textel) + pbo[index].w = 0; + pbo[index].x = color.x; + pbo[index].y = color.y; + pbo[index].z = color.z; + } +} + +__global__ void gbufferToPBO(uchar4* pbo, glm::ivec2 resolution, GBufferPixel* gBuffer) { + int x = (blockIdx.x * blockDim.x) + threadIdx.x; + int y = (blockIdx.y * blockDim.y) + threadIdx.y; + int index = x + (y * resolution.x); +#if GBUFFER_T + if (x < resolution.x && y < resolution.y) { + float timeToIntersect = gBuffer[index].t * 256.0; pbo[index].w = 0; @@ -80,6 +129,24 @@ __global__ void gbufferToPBO(uchar4* pbo, glm::ivec2 resolution, GBufferPixel* g pbo[index].y = timeToIntersect; pbo[index].z = timeToIntersect; } +#elif GBUFFER_NORM + if (x < resolution.x && y < resolution.y) { + + pbo[index].w = 0; + pbo[index].x = glm::clamp(gBuffer[index].nor.x * 25.f, 0.f, 255.f); + pbo[index].y = glm::clamp(gBuffer[index].nor.y * 25.f, 0.f, 255.f); + pbo[index].z = glm::clamp(gBuffer[index].nor.z * 25.f, 0.f, 255.f); + } +#elif GBUFFER_POS + if (x < resolution.x && y < resolution.y) { + + pbo[index].w = 0; + pbo[index].x = glm::clamp(gBuffer[index].pos.x * 25.f, 0.f, 255.f); + pbo[index].y = glm::clamp(gBuffer[index].pos.y * 25.f, 0.f, 255.f); + pbo[index].z = glm::clamp(gBuffer[index].pos.z * 25.f, 0.f, 255.f); + } +#endif + } static Scene * hst_scene = NULL; @@ -90,7 +157,12 @@ static PathSegment * dev_paths = NULL; static ShadeableIntersection * dev_intersections = NULL; static GBufferPixel* dev_gBuffer = NULL; // TODO: static variables for device memory, any extra info you need, etc -// ... +static ShadeableIntersection* dev_intersections_cache = NULL; +static Triangle* dev_triangles = NULL; +static Triangle* dev_oct_triangles = NULL; +static OctreeNode* dev_octree = NULL; +static glm::vec3* dev_image1 = NULL; +static glm::vec3* dev_image2 = NULL; void pathtraceInit(Scene *scene) { hst_scene = scene; @@ -114,6 +186,26 @@ void pathtraceInit(Scene *scene) { cudaMalloc(&dev_gBuffer, pixelcount * sizeof(GBufferPixel)); // TODO: initialize any extra device memeory you need + cudaMalloc(&dev_intersections_cache, pixelcount * sizeof(ShadeableIntersection)); + cudaMemset(dev_intersections_cache, 0, pixelcount * sizeof(ShadeableIntersection)); + + cudaMalloc(&dev_triangles, scene->mesh.triangles.size() * sizeof(Triangle)); + cudaMemcpy(dev_triangles, scene->mesh.triangles.data(), scene->mesh.triangles.size() * sizeof(Triangle), cudaMemcpyHostToDevice); + + cudaMalloc(&dev_oct_triangles, scene->octTriangles.size() * sizeof(Triangle)); + cudaMemcpy(dev_oct_triangles, scene->octTriangles.data(), scene->octTriangles.size() * sizeof(Triangle), cudaMemcpyHostToDevice); + + cudaMalloc(&dev_octree, scene->octree.size() * sizeof(OctreeNode)); + cudaMemcpy(dev_octree, scene->octree.data(), scene->octree.size() * sizeof(OctreeNode), cudaMemcpyHostToDevice); + + cudaMalloc(&dev_gBuffer, pixelcount * sizeof(GBufferPixel)); + cudaMemset(dev_gBuffer, 0, pixelcount * sizeof(GBufferPixel)); + + cudaMalloc(&dev_image1, pixelcount * sizeof(glm::vec3)); + cudaMemset(dev_image1, 0, pixelcount * sizeof(glm::vec3)); + + cudaMalloc(&dev_image2, pixelcount * sizeof(glm::vec3)); + cudaMemset(dev_image2, 0, pixelcount * sizeof(glm::vec3)); checkCUDAError("pathtraceInit"); } @@ -126,10 +218,43 @@ void pathtraceFree() { cudaFree(dev_intersections); cudaFree(dev_gBuffer); // TODO: clean up any extra device memory you created + cudaFree(dev_intersections_cache); + cudaFree(dev_triangles); + cudaFree(dev_oct_triangles); + cudaFree(dev_octree); + cudaFree(dev_image1); + cudaFree(dev_image2); checkCUDAError("pathtraceFree"); } +/** +* Function to map a random point to a sample on a unit disk. Based off of PBRT 13.6.2 +* https://pbr-book.org/3ed-2018/Monte_Carlo_Integration/2D_Sampling_with_Multidimensional_Transformations#ConcentricSampleDisk +*/ +__host__ __device__ glm::vec2 concentricSampleDisk(glm::vec2 u) { + // Map uniform random numbers from input to -1 to 1 range + glm::vec2 uOffset = 2.f * u - glm::vec2(1.0f, 1.0f); + + // Handle degeneracy at origin + if (uOffset.x == 0.0f && uOffset.y == 0.0f) { + return glm::vec2(0.0f, 0.0f); + } + + // Apply concentric mapping to point + float theta, r; + if (glm::abs(uOffset.x) > glm::abs(uOffset.y)) { + r = uOffset.x; + theta = (PI / 4.0f) * (uOffset.y / uOffset.x); + } + else { + r = uOffset.y; + theta = (PI / 2.0f) - (PI / 4.0f) * (uOffset.x / uOffset.y); + } + + return r * glm::vec2(glm::cos(theta), glm::sin(theta)); +} + /** * Generate PathSegments with rays from the camera through the screen into the * scene, which is the first bounce of rays. @@ -140,137 +265,192 @@ void pathtraceFree() { */ __global__ void generateRayFromCamera(Camera cam, int iter, int traceDepth, PathSegment* pathSegments) { - int x = (blockIdx.x * blockDim.x) + threadIdx.x; - int y = (blockIdx.y * blockDim.y) + threadIdx.y; + int x = (blockIdx.x * blockDim.x) + threadIdx.x; + int y = (blockIdx.y * blockDim.y) + threadIdx.y; + + if (x < cam.resolution.x && y < cam.resolution.y) { + int index = x + (y * cam.resolution.x); + PathSegment& segment = pathSegments[index]; + + segment.ray.origin = cam.position; + segment.color = glm::vec3(1.0f, 1.0f, 1.0f); + + // TODO: implement antialiasing by jittering the ray + thrust::default_random_engine rng = makeSeededRandomEngine(iter, index, traceDepth); + thrust::uniform_real_distribution n11u(-1, 1); + +#if ANTI_ALIASING && !CACHE_FIRST_BOUNCE + float xOffset = n11u(rng); + float yOffset = n11u(rng); + segment.ray.direction = glm::normalize(cam.view + - cam.right * cam.pixelLength.x * ((float)x - (float)cam.resolution.x * 0.5f + xOffset) + - cam.up * cam.pixelLength.y * ((float)y - (float)cam.resolution.y * 0.5f + yOffset) + ); +#else + segment.ray.direction = glm::normalize(cam.view + - cam.right * cam.pixelLength.x * ((float)x - (float)cam.resolution.x * 0.5f) + - cam.up * cam.pixelLength.y * ((float)y - (float)cam.resolution.y * 0.5f) + ); +#endif + + segment.ray.direction = glm::normalize(cam.view + - cam.right * cam.pixelLength.x * ((float)x - (float)cam.resolution.x * 0.5f) + - cam.up * cam.pixelLength.y * ((float)y - (float)cam.resolution.y * 0.5f) + ); - if (x < cam.resolution.x && y < cam.resolution.y) { - int index = x + (y * cam.resolution.x); - PathSegment & segment = pathSegments[index]; +#if DEPTH_OF_FIELD + // adapted from PBRT 6.2.3 https://pbr-book.org/3ed-2018/Camera_Models/Projective_Camera_Models + float lensRadius = cam.lensRadius; + float focalDistance = cam.focalDistance; - segment.ray.origin = cam.position; - segment.color = glm::vec3(1.0f, 1.0f, 1.0f); + if (lensRadius > 0) { + // sample point on lens + glm::vec2 pLens = lensRadius * concentricSampleDisk(glm::vec2(n11u(rng), n11u(rng))); - segment.ray.direction = glm::normalize(cam.view - - cam.right * cam.pixelLength.x * ((float)x - (float)cam.resolution.x * 0.5f) - - cam.up * cam.pixelLength.y * ((float)y - (float)cam.resolution.y * 0.5f) - ); + //compute point on plane of focus + float ft = focalDistance / glm::dot(cam.view, segment.ray.direction); + glm::vec3 pFocus = cam.position + ft * segment.ray.direction; - segment.pixelIndex = index; - segment.remainingBounces = traceDepth; - } + segment.ray.origin = cam.position + (cam.right * pLens.x) + (cam.up * pLens.y); + segment.ray.direction = glm::normalize(pFocus - segment.ray.origin); + } +#endif + segment.pixelIndex = index; + segment.remainingBounces = traceDepth; + } } __global__ void computeIntersections( - int depth - , int num_paths - , PathSegment * pathSegments - , Geom * geoms - , int geoms_size - , ShadeableIntersection * intersections - ) + int depth + , int num_paths + , PathSegment* pathSegments + , Geom* geoms + , Triangle* triangles + , int numTriangles + , glm::vec3 bottomLeft + , glm::vec3 topRight + , OctreeNode* nodes + , int num_nodes + , int geoms_size + , ShadeableIntersection* intersections +) { - int path_index = blockIdx.x * blockDim.x + threadIdx.x; - - if (path_index < num_paths) - { - PathSegment pathSegment = pathSegments[path_index]; - - float t; - glm::vec3 intersect_point; - glm::vec3 normal; - float t_min = FLT_MAX; - int hit_geom_index = -1; - bool outside = true; - - glm::vec3 tmp_intersect; - glm::vec3 tmp_normal; - - // naive parse through global geoms - - for (int i = 0; i < geoms_size; i++) - { - Geom & geom = geoms[i]; - - if (geom.type == CUBE) - { - t = boxIntersectionTest(geom, pathSegment.ray, tmp_intersect, tmp_normal, outside); - } - else if (geom.type == SPHERE) - { - t = sphereIntersectionTest(geom, pathSegment.ray, tmp_intersect, tmp_normal, outside); - } - - // Compute the minimum t from the intersection tests to determine what - // scene geometry object was hit first. - if (t > 0.0f && t_min > t) - { - t_min = t; - hit_geom_index = i; - intersect_point = tmp_intersect; - normal = tmp_normal; - } - } - - if (hit_geom_index == -1) - { - intersections[path_index].t = -1.0f; - } - else - { - //The ray hits something - intersections[path_index].t = t_min; - intersections[path_index].materialId = geoms[hit_geom_index].materialid; - intersections[path_index].surfaceNormal = normal; - } - } + int path_index = blockIdx.x * blockDim.x + threadIdx.x; + + if (path_index >= num_paths) return; + + PathSegment pathSegment = pathSegments[path_index]; + + float t; + glm::vec3 intersect_point; + glm::vec3 normal; + float t_min = FLT_MAX; + int hit_geom_index = -1; + bool outside = true; + + glm::vec3 tmp_intersect; + glm::vec3 tmp_normal; + + // naive parse through global geoms + + for (int i = 0; i < geoms_size; i++) + { + Geom& geom = geoms[i]; + + if (geom.type == CUBE) + { + t = boxIntersectionTest(geom, pathSegment.ray, tmp_intersect, tmp_normal, outside); + } + else if (geom.type == SPHERE) + { + t = sphereIntersectionTest(geom, pathSegment.ray, tmp_intersect, tmp_normal, outside); + } + else if (geom.type == MESH) + { +#if BOUNDING_BOX + if (boundingBoxCheck(geom, pathSegment.ray, bottomLeft, topRight)) { + t = meshIntersectionTest(geom, triangles, numTriangles, pathSegment.ray, tmp_intersect, tmp_normal, outside); + } +#elif OCTREE + bool isLeaf; + for (int i = 0; i < num_nodes; i++) { + t = octreeIntersectionTest(nodes[i], geom, triangles, numTriangles, pathSegment.ray, tmp_intersect, tmp_normal, outside, isLeaf); + if (t > 0.0f && isLeaf) { + break; + } + } +#else + t = meshIntersectionTest(geom, triangles, numTriangles, pathSegment.ray, tmp_intersect, tmp_normal, outside); +#endif + } + + // Compute the minimum t from the intersection tests to determine what + // scene geometry object was hit first. + if (t > 0.0f && t_min > t) + { + t_min = t; + hit_geom_index = i; + intersect_point = tmp_intersect; + normal = tmp_normal; + } + } + + if (hit_geom_index == -1) + { + intersections[path_index].t = -1.0f; + } + else + { + //The ray hits something + intersections[path_index].t = t_min; + intersections[path_index].materialId = geoms[hit_geom_index].materialid; + intersections[path_index].surfaceNormal = normal; + } } -__global__ void shadeSimpleMaterials ( - int iter - , int num_paths - , ShadeableIntersection * shadeableIntersections - , PathSegment * pathSegments - , Material * materials - ) +__global__ void shadeMaterials( + int iter, + int numPaths, + int depth, + ShadeableIntersection* shadeableIntersections, + PathSegment* pathSegments, + Material* materials +) { - int idx = blockIdx.x * blockDim.x + threadIdx.x; - if (idx < num_paths) - { - ShadeableIntersection intersection = shadeableIntersections[idx]; - PathSegment segment = pathSegments[idx]; - if (segment.remainingBounces == 0) { - return; + int index = (blockIdx.x * blockDim.x) + threadIdx.x; + if (index >= numPaths) return; + + PathSegment& seg = pathSegments[index]; + ShadeableIntersection& inter = shadeableIntersections[index]; + + if (inter.t > 0.0f) { + thrust::default_random_engine rng = makeSeededRandomEngine(iter, index, depth); + thrust::uniform_real_distribution u01(0, 1); + + Material& mat = materials[inter.materialId]; + glm::vec3 matColor = mat.color; + + // mat is a light so terminate + if (mat.emittance > 0.0f) { + seg.remainingBounces = 0; + seg.color *= matColor * mat.emittance; + return; + } + + // determine new ray + if (seg.remainingBounces > 0) { + glm::vec3 intersect = getPointOnRay(seg.ray, inter.t); + scatterRay(seg, intersect, inter.surfaceNormal, mat, rng); + } + else { + seg.color = glm::vec3(0.0f, 0.0f, 0.0f); + } + seg.remainingBounces--; } - - if (intersection.t > 0.0f) { // if the intersection exists... - segment.remainingBounces--; - // Set up the RNG - thrust::default_random_engine rng = makeSeededRandomEngine(iter, idx, segment.remainingBounces); - - Material material = materials[intersection.materialId]; - glm::vec3 materialColor = material.color; - - // If the material indicates that the object was a light, "light" the ray - if (material.emittance > 0.0f) { - segment.color *= (materialColor * material.emittance); - segment.remainingBounces = 0; - } - else { - segment.color *= materialColor; - glm::vec3 intersectPos = intersection.t * segment.ray.direction + segment.ray.origin; - scatterRay(segment, intersectPos, intersection.surfaceNormal, material, rng); - } - // If there was no intersection, color the ray black. - // Lots of renderers use 4 channel color, RGBA, where A = alpha, often - // used for opacity, in which case they can indicate "no opacity". - // This can be useful for post-processing and image compositing. - } else { - segment.color = glm::vec3(0.0f); - segment.remainingBounces = 0; + else { + seg.remainingBounces = 0; + seg.color = glm::vec3(0.0f, 0.0f, 0.0f); } - - pathSegments[idx] = segment; - } } __global__ void generateGBuffer ( @@ -282,133 +462,305 @@ __global__ void generateGBuffer ( if (idx < num_paths) { gBuffer[idx].t = shadeableIntersections[idx].t; + gBuffer[idx].norm = shadeableIntersections[idx].surfaceNormal; + gBuffer[idx].pos = getPointOnRay(pathSegments[idx].ray, shadeableIntersections[idx].t); } } // Add the current iteration's output to the overall image -__global__ void finalGather(int nPaths, glm::vec3 * image, PathSegment * iterationPaths) +__global__ void finalGather(int nPaths, glm::vec3* image, PathSegment* iterationPaths) { - int index = (blockIdx.x * blockDim.x) + threadIdx.x; + int index = (blockIdx.x * blockDim.x) + threadIdx.x; - if (index < nPaths) - { - PathSegment iterationPath = iterationPaths[index]; - image[iterationPath.pixelIndex] += iterationPath.color; - } + if (index < nPaths) + { + PathSegment iterationPath = iterationPaths[index]; + image[iterationPath.pixelIndex] += iterationPath.color; + } } +// Comparison operator can be defined for thrust sort like here: https://stackoverflow.com/questions/5282039/sorting-objects-with-thrust-cuda +struct compMats { + __host__ __device__ bool operator()(const ShadeableIntersection& i1, const ShadeableIntersection& i2) { + return i1.materialId < i2.materialId; + } +}; + +struct continuePath { + __host__ __device__ bool operator()(const PathSegment path) { + return path.remainingBounces > 0; + } +}; + /** * Wrapper for the __global__ call that sets up the kernel calls and does a ton * of memory management */ void pathtrace(int frame, int iter) { const int traceDepth = hst_scene->state.traceDepth; - const Camera &cam = hst_scene->state.camera; + const Camera& cam = hst_scene->state.camera; const int pixelcount = cam.resolution.x * cam.resolution.y; - // 2D block for generating ray from camera + // 2D block for generating ray from camera const dim3 blockSize2d(8, 8); const dim3 blocksPerGrid2d( - (cam.resolution.x + blockSize2d.x - 1) / blockSize2d.x, - (cam.resolution.y + blockSize2d.y - 1) / blockSize2d.y); + (cam.resolution.x + blockSize2d.x - 1) / blockSize2d.x, + (cam.resolution.y + blockSize2d.y - 1) / blockSize2d.y); - // 1D block for path tracing - const int blockSize1d = 128; + // 1D block for path tracing + const int blockSize1d = 128; /////////////////////////////////////////////////////////////////////////// - // Pathtracing Recap: - // * Initialize array of path rays (using rays that come out of the camera) - // * You can pass the Camera object to that kernel. - // * Each path ray must carry at minimum a (ray, color) pair, - // * where color starts as the multiplicative identity, white = (1, 1, 1). - // * This has already been done for you. - // * NEW: For the first depth, generate geometry buffers (gbuffers) - // * For each depth: - // * Compute an intersection in the scene for each path ray. - // A very naive version of this has been implemented for you, but feel - // free to add more primitives and/or a better algorithm. - // Currently, intersection distance is recorded as a parametric distance, - // t, or a "distance along the ray." t = -1.0 indicates no intersection. - // * Color is attenuated (multiplied) by reflections off of any object - // * Stream compact away all of the terminated paths. - // You may use either your implementation or `thrust::remove_if` or its - // cousins. - // * Note that you can't really use a 2D kernel launch any more - switch - // to 1D. - // * Shade the rays that intersected something or didn't bottom out. - // That is, color the ray by performing a color computation according - // to the shader, then generate a new ray to continue the ray path. - // We recommend just updating the ray's PathSegment in place. - // Note that this step may come before or after stream compaction, - // since some shaders you write may also cause a path to terminate. - // * Finally: - // * if not denoising, add this iteration's results to the image - // * TODO: if denoising, run kernels that take both the raw pathtraced result and the gbuffer, and put the result in the "pbo" from opengl - - generateRayFromCamera <<>>(cam, iter, traceDepth, dev_paths); - checkCUDAError("generate camera ray"); - - int depth = 0; - PathSegment* dev_path_end = dev_paths + pixelcount; - int num_paths = dev_path_end - dev_paths; - - // --- PathSegment Tracing Stage --- - // Shoot ray into scene, bounce between objects, push shading chunks - - // Empty gbuffer - cudaMemset(dev_gBuffer, 0, pixelcount * sizeof(GBufferPixel)); - - // clean shading chunks - cudaMemset(dev_intersections, 0, pixelcount * sizeof(ShadeableIntersection)); - - bool iterationComplete = false; - while (!iterationComplete) { - - // tracing - dim3 numblocksPathSegmentTracing = (num_paths + blockSize1d - 1) / blockSize1d; - computeIntersections <<>> ( - depth - , num_paths - , dev_paths - , dev_geoms - , hst_scene->geoms.size() - , dev_intersections - ); - checkCUDAError("trace one bounce"); - cudaDeviceSynchronize(); - - if (depth == 0) { - generateGBuffer<<>>(num_paths, dev_intersections, dev_paths, dev_gBuffer); - } + timer().startGpuTimer(); + + generateRayFromCamera << > > (cam, iter, traceDepth, dev_paths); + checkCUDAError("generate camera ray"); + + int depth = 0; + PathSegment* dev_path_end = dev_paths + pixelcount; + int num_paths = dev_path_end - dev_paths; + int total_paths = num_paths; + + // --- PathSegment Tracing Stage --- + // Shoot ray into scene, bounce between objects, push shading chunks + + // clean gBuffer + cudaMemset(dev_gBuffer, 0, pixelcount * sizeof(GBufferPixel)); + + // clean shading chunks + cudaMemset(dev_intersections, 0, pixelcount * sizeof(ShadeableIntersection)); + + bool iterationComplete = false; + while (!iterationComplete) { + + dim3 numblocksPathSegmentTracing = (num_paths + blockSize1d - 1) / blockSize1d; + +#if CACHE_FIRST_BOUNCE +#if OCTREE + if (depth == 0 && iter == 1) { // first bounce of first iteration + computeIntersections << > > ( + depth + , num_paths + , dev_paths + , dev_geoms + , dev_oct_triangles + , hst_scene->octTriangles.size() + , hst_scene->mesh.bottomLeft + , hst_scene->mesh.topRight + , dev_octree + , hst_scene->octree.size() + , hst_scene->geoms.size() + , dev_intersections_cache); + checkCUDAError("First iter first bounce cache error"); + cudaDeviceSynchronize(); + cudaMemcpy(dev_intersections, dev_intersections_cache, pixelcount * sizeof(ShadeableIntersection), cudaMemcpyDeviceToDevice); + } + else if (depth == 0 && iter > 1) { // use the cached first bounce for all the following iterations + cudaMemcpy(dev_intersections, dev_intersections_cache, pixelcount * sizeof(ShadeableIntersection), cudaMemcpyDeviceToDevice); + } + else { // rest of the bounces can't be cached so compute + computeIntersections << > > ( + depth + , num_paths + , dev_paths + , dev_geoms + , dev_oct_triangles + , hst_scene->octTriangles.size() + , hst_scene->mesh.bottomLeft + , hst_scene->mesh.topRight + , dev_octree + , hst_scene->octree.size() + , hst_scene->geoms.size() + , dev_intersections); + checkCUDAError("trace one bounce"); + cudaDeviceSynchronize(); + } +#else + if (depth == 0 && iter == 1) { // first bounce of first iteration + computeIntersections << > > ( + depth + , num_paths + , dev_paths + , dev_geoms + , dev_triangles + , hst_scene->mesh.numTriangles + , hst_scene->mesh.bottomLeft + , hst_scene->mesh.topRight + , dev_octree + , hst_scene->octree.size() + , hst_scene->geoms.size() + , dev_intersections_cache); + checkCUDAError("First iter first bounce cache error"); + cudaDeviceSynchronize(); + cudaMemcpy(dev_intersections, dev_intersections_cache, pixelcount * sizeof(ShadeableIntersection), cudaMemcpyDeviceToDevice); + } + else if (depth == 0 && iter > 1) { // use the cached first bounce for all the following iterations + cudaMemcpy(dev_intersections, dev_intersections_cache, pixelcount * sizeof(ShadeableIntersection), cudaMemcpyDeviceToDevice); + } + else { // rest of the bounces can't be cached so compute + computeIntersections << > > ( + depth + , num_paths + , dev_paths + , dev_geoms + , dev_triangles + , hst_scene->mesh.numTriangles + , hst_scene->mesh.bottomLeft + , hst_scene->mesh.topRight + , dev_octree + , hst_scene->octree.size() + , hst_scene->geoms.size() + , dev_intersections); + checkCUDAError("trace one bounce"); + cudaDeviceSynchronize(); + } +#endif + +#else + computeIntersections << > > ( + depth + , num_paths + , dev_paths + , dev_geoms + , dev_triangles + , hst_scene->mesh.numTriangles + , hst_scene->mesh.bottomLeft + , hst_scene->mesh.topRight + , dev_octree + , hst_scene->octree.size() + , hst_scene->geoms.size() + , dev_intersections); + checkCUDAError("trace one bounce"); + cudaDeviceSynchronize(); +#endif + + if (depth == 0) { + generateGBuffer << > > + (num_paths, dev_intersections, dev_paths, dev_gBuffer); + } + depth++; + + shadeMaterials << > > ( + iter, + num_paths, + depth, + dev_intersections, + dev_paths, + dev_materials + ); + +#if SORT_MATERIALS + // Sort by material + thrust::stable_sort_by_key(thrust::device, dev_intersections, dev_intersections + num_paths, dev_paths, compMats()); +#endif + +#if STREAM_COMPACTION + // Stream compaction + dev_path_end = thrust::partition( + thrust::device, + dev_paths, + dev_path_end, + continuePath()); // moves all paths that can continue to front and returns new pointer to ending + num_paths = dev_path_end - dev_paths; +#endif + num_paths = dev_path_end - dev_paths; + iterationComplete = (depth >= hst_scene->state.traceDepth) || (num_paths <= 0); - depth++; + } - shadeSimpleMaterials<<>> ( - iter, - num_paths, - dev_intersections, - dev_paths, - dev_materials - ); - iterationComplete = depth == traceDepth; - } + timer().endGpuTimer(); + printElapsedTime(timer().getGpuElapsedTimeForPreviousOperation(), "(CUDA Measured)"); + totalTime += timer().getGpuElapsedTimeForPreviousOperation(); + if (iter == 100) std::cout << "Total Time: " << totalTime << "ms" << std::endl; - // Assemble this iteration and apply it to the image - dim3 numBlocksPixels = (pixelcount + blockSize1d - 1) / blockSize1d; - finalGather<<>>(num_paths, dev_image, dev_paths); + // Assemble this iteration and apply it to the image + dim3 numBlocksPixels = (pixelcount + blockSize1d - 1) / blockSize1d; + finalGather << > > (total_paths, dev_image, dev_paths); /////////////////////////////////////////////////////////////////////////// - // CHECKITOUT: use dev_image as reference if you want to implement saving denoised images. - // Otherwise, screenshots are also acceptable. // Retrieve image from GPU cudaMemcpy(hst_scene->state.image.data(), dev_image, - pixelcount * sizeof(glm::vec3), cudaMemcpyDeviceToHost); + pixelcount * sizeof(glm::vec3), cudaMemcpyDeviceToHost); checkCUDAError("pathtrace"); } +__global__ void denoiseATrous( + float c_phi, // color + float n_phi, // normal + float p_phi, //postiion + int stepWidth, + GBufferPixel* gBuffer, + glm::vec3* image_in, + glm::vec3* image_out, + glm::ivec2 camRes) { + + int x = (blockIdx.x * blockDim.x) + threadIdx.x; + int y = (blockIdx.y * blockDim.y) + threadIdx.y; + + if (x < camRes.x && y < camRes.y) { + float kern[5] = {0.0625f, 0.25f, 0.375f, 0.25f, 0.0625f}; + glm::vec3 sum = glm::vec3(0.f, 0.f, 0.f); + + int index = x + (y * camRes.x); + glm::vec3 cval = image_in[index]; + glm::vec3 nval = gBuffer[index].norm; + glm::vec3 pval = gBuffer[index].pos; + + float cum_w = 0.0f; + + for (int i = -2; i <= 2; i++) { + for (int j = -2; j <= 2; j++) { + float kernVal = kern[i + 2] * kern[j + 2]; + + // Find neighbors + glm::ivec2 uv = glm::clamp(glm::ivec2(x + (i * stepWidth), y + (j * stepWidth)), + glm::ivec2(0, 0), + camRes - glm::ivec2(1, 1)); + int uvIndex = uv.x + (uv.y * camRes.x); + + // Colors + glm::vec3 ctmp = image_in[uvIndex]; + glm::vec3 t = cval - ctmp; + float cdist = glm::dot(t, t); + float c_w = glm::min(glm::exp(-cdist / c_phi), 1.0f); + + // Normals + glm::vec3 ntmp = gBuffer[uvIndex].norm; + t = nval - ntmp; + float ndist = glm::max(glm::dot(t, t) / ((float)stepWidth * (float)stepWidth), 0.f); + float n_w = glm::min(glm::exp(-ndist / n_phi), 1.0f); + + // Positions + glm::vec3 ptmp = gBuffer[uvIndex].pos; + t = pval - ptmp; + float pdist = dot(t, t); + float p_w = glm::min(glm::exp(-pdist / p_phi), 1.0f); + + float weight = c_w * n_w * p_w; + sum += ctmp * weight * kernVal; + cum_w += weight * kernVal; + } + } + image_out[index] = sum / cum_w; + } +} + +__global__ void copyImageBuffer(glm::ivec2 camRes, int iter, glm::vec3* dest, const glm::vec3* src) { + int x = (blockIdx.x * blockDim.x) + threadIdx.x; + int y = (blockIdx.y * blockDim.y) + threadIdx.y; + + if (x < camRes.x && y < camRes.y) { + int index = x + (y * camRes.x); + + dest[index].x = src[index].x / iter; + dest[index].y = src[index].y / iter; + dest[index].z = src[index].z / iter; + } +} + // CHECKITOUT: this kernel "post-processes" the gbuffer/gbuffers into something that you can visualize for debugging. void showGBuffer(uchar4* pbo) { const Camera &cam = hst_scene->state.camera; @@ -422,7 +774,7 @@ void showGBuffer(uchar4* pbo) { } void showImage(uchar4* pbo, int iter) { -const Camera &cam = hst_scene->state.camera; + const Camera &cam = hst_scene->state.camera; const dim3 blockSize2d(8, 8); const dim3 blocksPerGrid2d( (cam.resolution.x + blockSize2d.x - 1) / blockSize2d.x, @@ -431,3 +783,33 @@ const Camera &cam = hst_scene->state.camera; // Send results to OpenGL buffer for rendering sendImageToPBO<<>>(pbo, cam.resolution, iter, dev_image); } + +void showDenoised(uchar4* pbo, int iter, int filterSize, float c_phi, float n_phi, float p_phi) { + timer().startGpuTimer(); + const Camera& cam = hst_scene->state.camera; + const int pixelcount = cam.resolution.x * cam.resolution.y; + const dim3 blockSize2d(8, 8); + const dim3 blocksPerGrid2d( + (cam.resolution.x + blockSize2d.x - 1) / blockSize2d.x, + (cam.resolution.y + blockSize2d.y - 1) / blockSize2d.y); + + + copyImageBuffer << > > (cam.resolution, iter, dev_image1, dev_image); + + int numIters = glm::ceil(glm::log2(filterSize)); + for (int i = 0; i < numIters; i++) { + int stepWidth = 1 << i; + denoiseATrous << > > + (c_phi, n_phi, p_phi, stepWidth, dev_gBuffer, dev_image1, dev_image2, cam.resolution); + + std::swap(dev_image1, dev_image2); // ping-pong buffers + } + + sendDenoiseToPBO << > > (pbo, cam.resolution, dev_image1); + + cudaMemcpy(hst_scene->state.image.data(), dev_image1, pixelcount * sizeof(glm::vec3), cudaMemcpyDeviceToHost); + + timer().endGpuTimer(); + std::cout << timer().getGpuElapsedTimeForPreviousOperation() << std::endl; + //printElapsedTime(timer().getGpuElapsedTimeForPreviousOperation(), "(Denoise CUDA Measured)"); +} diff --git a/src/pathtrace.h b/src/pathtrace.h index 9e12f44..3bc7689 100644 --- a/src/pathtrace.h +++ b/src/pathtrace.h @@ -8,3 +8,4 @@ void pathtraceFree(); void pathtrace(int frame, int iteration); void showGBuffer(uchar4 *pbo); void showImage(uchar4 *pbo, int iter); +void showDenoised(uchar4* pbo, int iter, int filterSize, float c_phi, float n_phi, float p_phi); diff --git a/src/scene.cpp b/src/scene.cpp index cbae043..554369f 100644 --- a/src/scene.cpp +++ b/src/scene.cpp @@ -1,3 +1,6 @@ +#define TINYOBJLOADER_IMPLEMENTATION +#include "tiny_obj_loader.h" + #include #include "scene.h" #include @@ -21,10 +24,42 @@ Scene::Scene(string filename) { if (strcmp(tokens[0].c_str(), "MATERIAL") == 0) { loadMaterial(tokens[1]); cout << " " << endl; - } else if (strcmp(tokens[0].c_str(), "OBJECT") == 0) { - loadGeom(tokens[1]); + } + else if (strcmp(tokens[0].c_str(), "OBJECT") == 0) { + loadGeom(tokens[1], ""); + cout << " " << endl; + } + else if (strcmp(tokens[0].c_str(), "CAMERA") == 0) { + loadCamera(); + cout << " " << endl; + } + } + } +} + +Scene::Scene(string filename, string objFilename) { + cout << "Reading scene from " << filename << " ..." << endl; + cout << " " << endl; + char* fname = (char*)filename.c_str(); + fp_in.open(fname); + if (!fp_in.is_open()) { + cout << "Error reading from file - aborting!" << endl; + throw; + } + while (fp_in.good()) { + string line; + utilityCore::safeGetline(fp_in, line); + if (!line.empty()) { + vector tokens = utilityCore::tokenizeString(line); + if (strcmp(tokens[0].c_str(), "MATERIAL") == 0) { + loadMaterial(tokens[1]); + cout << " " << endl; + } + else if (strcmp(tokens[0].c_str(), "OBJECT") == 0) { + loadGeom(tokens[1], objFilename); cout << " " << endl; - } else if (strcmp(tokens[0].c_str(), "CAMERA") == 0) { + } + else if (strcmp(tokens[0].c_str(), "CAMERA") == 0) { loadCamera(); cout << " " << endl; } @@ -32,12 +67,13 @@ Scene::Scene(string filename) { } } -int Scene::loadGeom(string objectid) { +int Scene::loadGeom(string objectid, string objFilename) { int id = atoi(objectid.c_str()); if (id != geoms.size()) { cout << "ERROR: OBJECT ID does not match expected number of geoms" << endl; return -1; - } else { + } + else { cout << "Loading Geom " << id << "..." << endl; Geom newGeom; string line; @@ -48,10 +84,19 @@ int Scene::loadGeom(string objectid) { if (strcmp(line.c_str(), "sphere") == 0) { cout << "Creating new sphere..." << endl; newGeom.type = SPHERE; - } else if (strcmp(line.c_str(), "cube") == 0) { + } + else if (strcmp(line.c_str(), "cube") == 0) { cout << "Creating new cube..." << endl; newGeom.type = CUBE; } + else if (strcmp(line.c_str(), "mesh") == 0) { + cout << "Creating new mesh..." << endl; + newGeom.type = MESH; + if (objFilename.empty()) { + std::cout << "Error in passing objFilename..." << std::endl; + } + loadMesh(objFilename); + } } //link material @@ -70,9 +115,11 @@ int Scene::loadGeom(string objectid) { //load tranformations if (strcmp(tokens[0].c_str(), "TRANS") == 0) { newGeom.translation = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); - } else if (strcmp(tokens[0].c_str(), "ROTAT") == 0) { + } + else if (strcmp(tokens[0].c_str(), "ROTAT") == 0) { newGeom.rotation = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); - } else if (strcmp(tokens[0].c_str(), "SCALE") == 0) { + } + else if (strcmp(tokens[0].c_str(), "SCALE") == 0) { newGeom.scale = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); } @@ -80,7 +127,7 @@ int Scene::loadGeom(string objectid) { } newGeom.transform = utilityCore::buildTransformationMatrix( - newGeom.translation, newGeom.rotation, newGeom.scale); + newGeom.translation, newGeom.rotation, newGeom.scale); newGeom.inverseTransform = glm::inverse(newGeom.transform); newGeom.invTranspose = glm::inverseTranspose(newGeom.transform); @@ -91,8 +138,8 @@ int Scene::loadGeom(string objectid) { int Scene::loadCamera() { cout << "Loading Camera ..." << endl; - RenderState &state = this->state; - Camera &camera = state.camera; + RenderState& state = this->state; + Camera& camera = state.camera; float fovy; //load static properties @@ -103,13 +150,17 @@ int Scene::loadCamera() { if (strcmp(tokens[0].c_str(), "RES") == 0) { camera.resolution.x = atoi(tokens[1].c_str()); camera.resolution.y = atoi(tokens[2].c_str()); - } else if (strcmp(tokens[0].c_str(), "FOVY") == 0) { + } + else if (strcmp(tokens[0].c_str(), "FOVY") == 0) { fovy = atof(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "ITERATIONS") == 0) { + } + else if (strcmp(tokens[0].c_str(), "ITERATIONS") == 0) { state.iterations = atoi(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "DEPTH") == 0) { + } + else if (strcmp(tokens[0].c_str(), "DEPTH") == 0) { state.traceDepth = atoi(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "FILE") == 0) { + } + else if (strcmp(tokens[0].c_str(), "FILE") == 0) { state.imageName = tokens[1]; } } @@ -120,11 +171,19 @@ int Scene::loadCamera() { vector tokens = utilityCore::tokenizeString(line); if (strcmp(tokens[0].c_str(), "EYE") == 0) { camera.position = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); - } else if (strcmp(tokens[0].c_str(), "LOOKAT") == 0) { + } + else if (strcmp(tokens[0].c_str(), "LOOKAT") == 0) { camera.lookAt = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); - } else if (strcmp(tokens[0].c_str(), "UP") == 0) { + } + else if (strcmp(tokens[0].c_str(), "UP") == 0) { camera.up = glm::vec3(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); } + else if (strcmp(tokens[0].c_str(), "FOCALDISTANCE") == 0) { + camera.focalDistance = atof(tokens[1].c_str()); + } + else if (strcmp(tokens[0].c_str(), "LENSRADIUS") == 0) { + camera.lensRadius = atof(tokens[1].c_str()); + } utilityCore::safeGetline(fp_in, line); } @@ -135,9 +194,9 @@ int Scene::loadCamera() { float fovx = (atan(xscaled) * 180) / PI; camera.fov = glm::vec2(fovx, fovy); - camera.right = glm::normalize(glm::cross(camera.view, camera.up)); - camera.pixelLength = glm::vec2(2 * xscaled / (float)camera.resolution.x - , 2 * yscaled / (float)camera.resolution.y); + camera.right = glm::normalize(glm::cross(camera.view, camera.up)); + camera.pixelLength = glm::vec2(2 * xscaled / (float)camera.resolution.x, + 2 * yscaled / (float)camera.resolution.y); camera.view = glm::normalize(camera.lookAt - camera.position); @@ -155,7 +214,8 @@ int Scene::loadMaterial(string materialid) { if (id != materials.size()) { cout << "ERROR: MATERIAL ID does not match expected number of materials" << endl; return -1; - } else { + } + else { cout << "Loading Material " << id << "..." << endl; Material newMaterial; @@ -165,20 +225,26 @@ int Scene::loadMaterial(string materialid) { utilityCore::safeGetline(fp_in, line); vector tokens = utilityCore::tokenizeString(line); if (strcmp(tokens[0].c_str(), "RGB") == 0) { - glm::vec3 color( atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str()) ); + glm::vec3 color(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); newMaterial.color = color; - } else if (strcmp(tokens[0].c_str(), "SPECEX") == 0) { + } + else if (strcmp(tokens[0].c_str(), "SPECEX") == 0) { newMaterial.specular.exponent = atof(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "SPECRGB") == 0) { + } + else if (strcmp(tokens[0].c_str(), "SPECRGB") == 0) { glm::vec3 specColor(atof(tokens[1].c_str()), atof(tokens[2].c_str()), atof(tokens[3].c_str())); newMaterial.specular.color = specColor; - } else if (strcmp(tokens[0].c_str(), "REFL") == 0) { + } + else if (strcmp(tokens[0].c_str(), "REFL") == 0) { newMaterial.hasReflective = atof(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "REFR") == 0) { + } + else if (strcmp(tokens[0].c_str(), "REFR") == 0) { newMaterial.hasRefractive = atof(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "REFRIOR") == 0) { + } + else if (strcmp(tokens[0].c_str(), "REFRIOR") == 0) { newMaterial.indexOfRefraction = atof(tokens[1].c_str()); - } else if (strcmp(tokens[0].c_str(), "EMITTANCE") == 0) { + } + else if (strcmp(tokens[0].c_str(), "EMITTANCE") == 0) { newMaterial.emittance = atof(tokens[1].c_str()); } } @@ -186,3 +252,147 @@ int Scene::loadMaterial(string materialid) { return 1; } } + +// Based off of reference example code from https://github.com/tinyobjloader/tinyobjloader/blob/master/README.md +int Scene::loadMesh(std::string inputFile) { + tinyobj::ObjReaderConfig reader_config; + reader_config.mtl_search_path = "../scenes/obj"; // Path to material files + + tinyobj::ObjReader reader; + + if (!reader.ParseFromFile(inputFile, reader_config)) { + if (!reader.Error().empty()) { + std::cerr << "TinyObjReader: " << reader.Error(); + } + exit(1); + } + + if (!reader.Warning().empty()) { + std::cout << "TinyObjReader: " << reader.Warning(); + } + + auto& attrib = reader.GetAttrib(); + auto& shapes = reader.GetShapes(); + auto& materials = reader.GetMaterials(); + + mesh.bottomLeft = glm::vec3(FLT_MAX, FLT_MAX, FLT_MAX); + mesh.topRight = glm::vec3(FLT_MIN, FLT_MIN, FLT_MIN); + + // Loop over shapes + for (size_t s = 0; s < shapes.size(); s++) { + // Loop over faces(polygon) + size_t index_offset = 0; + for (size_t f = 0; f < shapes[s].mesh.num_face_vertices.size(); f++) { + size_t fv = size_t(shapes[s].mesh.num_face_vertices[f]); + Triangle t; + + bool hasNormals = false; + // Loop over vertices in the face. + for (size_t v = 0; v < fv; v++) { + // access to vertex + tinyobj::index_t idx = shapes[s].mesh.indices[index_offset + v]; + tinyobj::real_t vx = attrib.vertices[3 * size_t(idx.vertex_index) + 0]; + tinyobj::real_t vy = attrib.vertices[3 * size_t(idx.vertex_index) + 1]; + tinyobj::real_t vz = attrib.vertices[3 * size_t(idx.vertex_index) + 2]; + + t.verts[v] = glm::vec3(vx, vy, vz); + + // Check if `normal_index` is zero or positive. negative = no normal data + if (idx.normal_index >= 0) { + tinyobj::real_t nx = attrib.normals[3 * size_t(idx.normal_index) + 0]; + tinyobj::real_t ny = attrib.normals[3 * size_t(idx.normal_index) + 1]; + tinyobj::real_t nz = attrib.normals[3 * size_t(idx.normal_index) + 2]; + + t.normals[v] = glm::vec3(nx, ny, nz); + hasNormals = true; + } + else { + hasNormals = false; + } + + mesh.bottomLeft = glm::min(t.verts[v], mesh.bottomLeft); + mesh.topRight = glm::max(t.verts[v], mesh.topRight); + } + if (!hasNormals) { + glm::vec3 normal = glm::cross(t.verts[0] - t.verts[1], t.verts[0] - t.verts[2]); + for (int i = 0; i < 3; i++) t.normals[i] = normal * (i / 1.0f); + } + + mesh.triangles.push_back(t); + index_offset += fv; + } + } + mesh.numTriangles = size(mesh.triangles); + std::cout << "There are " << size(mesh.triangles) << " triangles in the mesh for " << inputFile << std::endl; +} + +void Scene::makeOctree() { + OctreeNode root; + root.bottomLeft = mesh.bottomLeft; + root.topRight = mesh.topRight; + root.center = (root.bottomLeft + root.topRight) / 2.0f; + + root.start = -1; + root.end = -1; + + octree.push_back(root); + makeOctreeNode(root, 0); +} + +void Scene::makeOctreeNode(OctreeNode parent, int level) { + if (level >= MAX_LEVEL) { + for (int i = 0; i < 8; i++) parent.children[i] = -1; + return; + } + + // childrern take up the next 8 indices + int firstChild = octree.size(); + for (int i = 0; i < 8; i++) parent.children[i] = firstChild + i; + + glm::vec3 d = parent.center - parent.bottomLeft; + + // make children + for (int i = 0; i < 2; i++) { + for (int j = 0; j < 2; j++) { + for (int k = 0; k < 2; k++) { + OctreeNode child; + + glm::vec3 bottomLeft = parent.bottomLeft + glm::vec3(i * d.x, j * d.y, k * d.z); + glm::vec3 topRight = bottomLeft + glm::vec3(d.x, d.y, d.z); + + child.bottomLeft = bottomLeft; + child.topRight = topRight; + child.center = (child.bottomLeft + child.topRight) / 2.0f; + + child.start = -1; + child.end = -1; + + // check each triangle to see if it falls within this leaf + if (level == MAX_LEVEL - 1) { + child.start = octTriangles.size(); + for (int t = 0; t < mesh.numTriangles; t++) { + for (int v = 0; v < 3; v++) { + if (mesh.triangles[t].verts[v].x > bottomLeft.x && mesh.triangles[t].verts[v].x < topRight.x + && mesh.triangles[t].verts[v].y > bottomLeft.y && mesh.triangles[t].verts[v].y < topRight.y + && mesh.triangles[t].verts[v].z > bottomLeft.z && mesh.triangles[t].verts[v].z < topRight.z) { + + octTriangles.push_back(mesh.triangles[t]); + + // one of the vertices of this triangle falls in this leaf so move on to next triangle + break; + } + } + } + child.end = octTriangles.size(); + } + + octree.push_back(child); + } + } + } + + // make grandchildren + for (int i = 0; i < 8; i++) { + makeOctreeNode(octree[parent.children[i]], level + 1); + } +} diff --git a/src/scene.h b/src/scene.h index f29a917..ca8f139 100644 --- a/src/scene.h +++ b/src/scene.h @@ -8,19 +8,29 @@ #include "utilities.h" #include "sceneStructs.h" +#define MAX_LEVEL 2 + using namespace std; class Scene { private: ifstream fp_in; int loadMaterial(string materialid); - int loadGeom(string objectid); + int loadGeom(string objectid, string objFilename); int loadCamera(); + int loadMesh(std::string inputFile); public: Scene(string filename); + Scene(string filename, string objFilename); + + void makeOctree(); + void makeOctreeNode(OctreeNode parent, int level); ~Scene(); std::vector geoms; std::vector materials; + std::vector octree; + std::vector octTriangles; + Mesh mesh; RenderState state; }; diff --git a/src/sceneStructs.h b/src/sceneStructs.h index da7e558..f1dc105 100644 --- a/src/sceneStructs.h +++ b/src/sceneStructs.h @@ -10,6 +10,7 @@ enum GeomType { SPHERE, CUBE, + MESH }; struct Ray { @@ -17,6 +18,29 @@ struct Ray { glm::vec3 direction; }; +struct Triangle { + glm::vec3 verts[3]; + glm::vec3 normals[3]; +}; + +struct Mesh { + std::vector triangles; + int numTriangles; + + glm::vec3 bottomLeft; + glm::vec3 topRight; +}; + +struct OctreeNode { + glm::vec3 center; + glm::vec3 bottomLeft; + glm::vec3 topRight; + + int children[8]; + int start; + int end; +}; + struct Geom { enum GeomType type; int materialid; @@ -49,6 +73,8 @@ struct Camera { glm::vec3 right; glm::vec2 fov; glm::vec2 pixelLength; + float focalDistance; + float lensRadius; }; struct RenderState { @@ -79,4 +105,6 @@ struct ShadeableIntersection { // What information might be helpful for guiding a denoising filter? struct GBufferPixel { float t; + glm::vec3 norm; + glm::vec3 pos; }; diff --git a/src/utilities.h b/src/utilities.h index abb4f27..8437c9d 100644 --- a/src/utilities.h +++ b/src/utilities.h @@ -1,5 +1,9 @@ #pragma once +#include +#include +#include + #include "glm/glm.hpp" #include #include @@ -24,3 +28,113 @@ namespace utilityCore { extern std::string convertIntToString(int number); extern std::istream& safeGetline(std::istream& is, std::string& t); //Thanks to http://stackoverflow.com/a/6089413 } + +/** + * This class is used for timing the performance + * Uncopyable and unmovable + * + * Adapted from WindyDarian(https://github.com/WindyDarian) + * Reused from project 2 stream compaction + * (https://github.com/CIS565-Fall-2021/Project2-Stream-Compaction/blob/main/stream_compaction/common.h) + */ +class PerformanceTimer +{ +public: + PerformanceTimer() + { + cudaEventCreate(&event_start); + cudaEventCreate(&event_end); + } + + ~PerformanceTimer() + { + cudaEventDestroy(event_start); + cudaEventDestroy(event_end); + } + + void startCpuTimer() + { + if (cpu_timer_started) { throw std::runtime_error("CPU timer already started"); } + cpu_timer_started = true; + + time_start_cpu = std::chrono::high_resolution_clock::now(); + } + + void endCpuTimer() + { + time_end_cpu = std::chrono::high_resolution_clock::now(); + + if (!cpu_timer_started) { throw std::runtime_error("CPU timer not started"); } + + std::chrono::duration duro = time_end_cpu - time_start_cpu; + prev_elapsed_time_cpu_milliseconds = + static_cast(duro.count()); + + cpu_timer_started = false; + } + + void startGpuTimer() + { + if (gpu_timer_started) { throw std::runtime_error("GPU timer already started"); } + gpu_timer_started = true; + + cudaEventRecord(event_start); + } + + void endGpuTimer() + { + cudaEventRecord(event_end); + cudaEventSynchronize(event_end); + + if (!gpu_timer_started) { throw std::runtime_error("GPU timer not started"); } + + cudaEventElapsedTime(&prev_elapsed_time_gpu_milliseconds, event_start, event_end); + gpu_timer_started = false; + } + + float getCpuElapsedTimeForPreviousOperation() //noexcept //(damn I need VS 2015 + { + return prev_elapsed_time_cpu_milliseconds; + } + + float getGpuElapsedTimeForPreviousOperation() //noexcept + { + return prev_elapsed_time_gpu_milliseconds; + } + + bool getCpuTimerStarted() + { + return cpu_timer_started; + } + + bool getGpuTimerStarted() + { + return gpu_timer_started; + } + + // remove copy and move functions + PerformanceTimer(const PerformanceTimer&) = delete; + PerformanceTimer(PerformanceTimer&&) = delete; + PerformanceTimer& operator=(const PerformanceTimer&) = delete; + PerformanceTimer& operator=(PerformanceTimer&&) = delete; + +private: + cudaEvent_t event_start = nullptr; + cudaEvent_t event_end = nullptr; + + using time_point_t = std::chrono::high_resolution_clock::time_point; + time_point_t time_start_cpu; + time_point_t time_end_cpu; + + bool cpu_timer_started = false; + bool gpu_timer_started = false; + + float prev_elapsed_time_cpu_milliseconds = 0.f; + float prev_elapsed_time_gpu_milliseconds = 0.f; +}; + +template +void printElapsedTime(T time, std::string note = "") +{ + std::cout << " elapsed time: " << time << "ms " << note << std::endl; +}