Finish initial implementation of .obj parser in C89.

missing size() in radme

.clang-format

@@ -1,5 +1,5 @@
 ---
-BasedOnStyle: LLVM
+BasedOnStyle: Google
 IndentWidth: 2
 TabWidth: 2
 UseTab: Never

.travis.yml

@@ -1,5 +1,5 @@
 language: cpp
-sudo: false
+sudo: required
 matrix:
   include:
   - addons: &1
@@ -36,21 +36,17 @@ matrix:
     env: COMPILER_VERSION=4.9 BUILD_TYPE=Release
   - addons: *1
     compiler: clang
-    env: COMPILER_VERSION=3.7 BUILD_TYPE=Debug CFLAGS="-O0 --coverage" CXXFLAGS="-O0
-      --coverage" REPORT_COVERAGE=1
+    env: COMPILER_VERSION=3.7 BUILD_TYPE=Debug
 before_install:
 - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew upgrade; fi
-- if [ -n "$REPORT_COVERAGE" ]; then pip install --user cpp-coveralls; fi
+- if [ -n "$REPORT_COVERAGE" ]; then sudo apt-get update python; fi
+- if [ -n "$REPORT_COVERAGE" ]; then sudo apt-get install python-dev libffi-dev libssl-dev; fi
+- if [ -n "$REPORT_COVERAGE" ]; then sudo pip install --upgrade pip; fi
+- if [ -n "$REPORT_COVERAGE" ]; then CXX=g++ pip install --user requests[security]; fi
+- if [ -n "$REPORT_COVERAGE" ]; then CXX=g++ pip install --user cpp-coveralls; fi
 script:
-- mkdir build && cd build
-- export CC="${CC}-${COMPILER_VERSION}"
-- export CXX="${CXX}-${COMPILER_VERSION}"
-- ${CC} -v
-- cmake --version
-- cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DTINYOBJLOADER_BUILD_TEST_LOADER=On -G Ninja
-  ..
-- ninja
-- ./test_loader ../cornell_box.obj
+- cd tests
+- make check
 - if [ -n "$REPORT_COVERAGE" ]; then coveralls -b . -r .. -e examples -e tools -e
   jni -e python -e images -E ".*CompilerId.*" -E ".*feature_tests.*" ; fi
 - cd ..

CMakeLists.txt

@@ -12,8 +12,8 @@ set(tinyobjloader-Source
 	${CMAKE_CURRENT_SOURCE_DIR}/tiny_obj_loader.cc
 	)
 
-set(tinyobjloader-Test-Source
-	${CMAKE_CURRENT_SOURCE_DIR}/test.cc
+set(tinyobjloader-Example-Source
+	${CMAKE_CURRENT_SOURCE_DIR}/loader_example.cc
 	)
 
 set(tinyobjloader-examples-objsticher
@@ -22,14 +22,17 @@ set(tinyobjloader-examples-objsticher
 	${TINYOBJLOADEREXAMPLES_DIR}/obj_sticher/obj_sticher.cc
 	)
 
-add_library(tinyobjloader
-			${tinyobjloader-Source}
-	)
+option(TINYOBJLOADER_BUILD_TEST_LOADER "Build Example Loader Application" OFF)
+option(TINYOBJLOADER_COMPILATION_SHARED "Build as shared library" OFF)
 
-option(TINYOBJLOADER_BUILD_TEST_LOADER "Build Test Loader Application" OFF)
+if (TINYOBJLOADER_COMPILATION_SHARED)
+	add_library(tinyobjloader SHARED ${tinyobjloader-Source})
+else()
+	add_library(tinyobjloader STATIC ${tinyobjloader-Source})
+endif()
 
 if(TINYOBJLOADER_BUILD_TEST_LOADER)
-  add_executable(test_loader ${tinyobjloader-Test-Source})
+  add_executable(test_loader ${tinyobjloader-Example-Source})
   target_link_libraries(test_loader tinyobjloader)
 endif()
 

README.md

@@ -21,6 +21,7 @@ Tiny but powerful single file wavefront obj loader written in C++. No dependency
 What's new
 ----------
 
+* XX  YY, ZZZZ : New data strcutre and API!
 * Jan 29, 2016 : Support n-polygon(no triangulation) and OpenSubdiv crease tag! Thanks dboogert!
 * Nov 26, 2015 : Now single-header only!.
 * Nov 08, 2015 : Improved API.
@@ -39,11 +40,15 @@ What's new
 Example
 -------
 
-![Rungholt](https://github.com/syoyo/tinyobjloader/blob/master/images/rungholt.jpg?raw=true)
+![Rungholt](images/rungholt.jpg)
 
 tinyobjloader can successfully load 6M triangles Rungholt scene.
 http://graphics.cs.williams.edu/data/meshes.xml
 
+![](images/sanmugel.png) 
+
+* [examples/viewer/](examples/viewer) OpenGL .obj viewer 
+
 Use case
 --------
 
@@ -72,32 +77,39 @@ Features
 * Material
   * Unknown material attributes are returned as key-value(value is string) map.
 * Crease tag('t'). This is OpenSubdiv specific(not in wavefront .obj specification)
+* Callback API for custom loading.
 
 
 TODO
 ----
 
-* [ ] Support different indices for vertex/normal/texcoord
+* [ ] Fix Python binding.
+* [ ] Fix obj_sticker example.
+* [ ] More unit test codes.
 
 License
 -------
 
-Licensed under 2 clause BSD.
+Licensed under MIT license.
 
 Usage
 -----
 
-TinyObjLoader triangulate input .obj by default.
+`attrib_t` contains single and linear array of vertex data(position, normal and texcoord).
+Each `shape_t` does not contain vertex data but contains array index to `attrib_t`.
+See `loader_example.cc` for more details.
+
 ```c++
 #define TINYOBJLOADER_IMPLEMENTATION // define this in only *one* .cc
 #include "tiny_obj_loader.h"
 
 std::string inputfile = "cornell_box.obj";
+tinyobj::attrib_t attrib;
 std::vector<tinyobj::shape_t> shapes;
 std::vector<tinyobj::material_t> materials;
   
 std::string err;
-bool ret = tinyobj::LoadObj(shapes, materials, err, inputfile.c_str());
+bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, inputfile.c_str());
   
 if (!err.empty()) { // `err` may contain warning message.
   std::cerr << err << std::endl;
@@ -107,88 +119,51 @@ if (!ret) {
   exit(1);
 }
 
-std::cout << "# of shapes    : " << shapes.size() << std::endl;
-std::cout << "# of materials : " << materials.size() << std::endl;
-  
-for (size_t i = 0; i < shapes.size(); i++) {
-  printf("shape[%ld].name = %s\n", i, shapes[i].name.c_str());
-  printf("Size of shape[%ld].indices: %ld\n", i, shapes[i].mesh.indices.size());
-  printf("Size of shape[%ld].material_ids: %ld\n", i, shapes[i].mesh.material_ids.size());
-  assert((shapes[i].mesh.indices.size() % 3) == 0);
-  for (size_t f = 0; f < shapes[i].mesh.indices.size() / 3; f++) {
-    printf("  idx[%ld] = %d, %d, %d. mat_id = %d\n", f, shapes[i].mesh.indices[3*f+0], shapes[i].mesh.indices[3*f+1], shapes[i].mesh.indices[3*f+2], shapes[i].mesh.material_ids[f]);
-  }
-
-  printf("shape[%ld].vertices: %ld\n", i, shapes[i].mesh.positions.size());
-  assert((shapes[i].mesh.positions.size() % 3) == 0);
-  for (size_t v = 0; v < shapes[i].mesh.positions.size() / 3; v++) {
-    printf("  v[%ld] = (%f, %f, %f)\n", v,
-      shapes[i].mesh.positions[3*v+0],
-      shapes[i].mesh.positions[3*v+1],
-      shapes[i].mesh.positions[3*v+2]);
-  }
-}
-
-for (size_t i = 0; i < materials.size(); i++) {
-  printf("material[%ld].name = %s\n", i, materials[i].name.c_str());
-  printf("  material.Ka = (%f, %f ,%f)\n", materials[i].ambient[0], materials[i].ambient[1], materials[i].ambient[2]);
-  printf("  material.Kd = (%f, %f ,%f)\n", materials[i].diffuse[0], materials[i].diffuse[1], materials[i].diffuse[2]);
-  printf("  material.Ks = (%f, %f ,%f)\n", materials[i].specular[0], materials[i].specular[1], materials[i].specular[2]);
-  printf("  material.Tr = (%f, %f ,%f)\n", materials[i].transmittance[0], materials[i].transmittance[1], materials[i].transmittance[2]);
-  printf("  material.Ke = (%f, %f ,%f)\n", materials[i].emission[0], materials[i].emission[1], materials[i].emission[2]);
-  printf("  material.Ns = %f\n", materials[i].shininess);
-  printf("  material.Ni = %f\n", materials[i].ior);
-  printf("  material.dissolve = %f\n", materials[i].dissolve);
-  printf("  material.illum = %d\n", materials[i].illum);
-  printf("  material.map_Ka = %s\n", materials[i].ambient_texname.c_str());
-  printf("  material.map_Kd = %s\n", materials[i].diffuse_texname.c_str());
-  printf("  material.map_Ks = %s\n", materials[i].specular_texname.c_str());
-  printf("  material.map_Ns = %s\n", materials[i].specular_highlight_texname.c_str());
-  std::map<std::string, std::string>::const_iterator it(materials[i].unknown_parameter.begin());
-  std::map<std::string, std::string>::const_iterator itEnd(materials[i].unknown_parameter.end());
-  for (; it != itEnd; it++) {
-    printf("  material.%s = %s\n", it->first.c_str(), it->second.c_str());
-  }
-  printf("\n");
-}
-```
-
-Reading .obj without triangulation. Use `num_vertices[i]` to iterate over faces(indices). `num_vertices[i]` stores the number of vertices for ith face.
-```c++
-#define TINYOBJLOADER_IMPLEMENTATION // define this in only *one* .cc
-#include "tiny_obj_loader.h"
-
-std::string inputfile = "cornell_box.obj";
-std::vector<tinyobj::shape_t> shapes;
-std::vector<tinyobj::material_t> materials;
-  
-std::string err;
-bool triangulate = false;
-bool ret = tinyobj::LoadObj(shapes, materials, err, inputfile.c_str(), triangulate);
-  
-if (!err.empty()) { // `err` may contain warning message.
-  std::cerr << err << std::endl;
-}
-
-if (!ret) {
-  exit(1);
-}
-
-for (size_t i = 0; i < shapes.size(); i++) {
-
-  size_t indexOffset = 0;
-  for (size_t n = 0; n < shapes[i].mesh.num_vertices.size(); n++) {
-    int ngon = shapes[i].mesh.num_vertices[n];
-    for (size_t f = 0; f < ngon; f++) {
-      unsigend int v = shapes[i].mesh.indices[indexOffset + f];
-      printf("  face[%ld] v[%ld] = (%f, %f, %f)\n", n,
-        shapes[i].mesh.positions[3*v+0],
-        shapes[i].mesh.positions[3*v+1],
-        shapes[i].mesh.positions[3*v+2]);
+// 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++) {
+    int fv = shapes[s].mesh.num_face_vertices[f];
 
+    // 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];
+      float vx = attrib.vertices[3*idx.vertex_index+0];
+      float vy = attrib.vertices[3*idx.vertex_index+1];
+      float vz = attrib.vertices[3*idx.vertex_index+2];
+      float nx = attrib.normals[3*idx.normal_index+0];
+      float ny = attrib.normals[3*idx.normal_index+1];
+      float nz = attrib.normals[3*idx.normal_index+2];
+      float tx = attrib.texcoords[2*idx.texcoord_index+0];
+      float ty = attrib.texcoords[2*idx.texcoord_index+1];
     }
-    indexOffset += ngon;
-  }
+    index_offset += fv;
 
+    // per-face material
+    shapes[s].mesh.material_ids[f];
+  }
 }
+
 ```
+
+Optimized loader
+----------------
+
+Optimized multi-threaded .obj loader is available at `experimental/` directory.
+If you want absolute performance to load .obj data, this optimized loader will fit your purpose.
+Note that the optimized loader uses C++11 thread and it does less error checks but may work most .obj data.
+
+Here is some benchmark result. Time are measured on MacBook 12(Early 2016, Core m5 1.2GHz).
+
+* Rungholt scene(6M triangles)
+  * old version(v0.9.x): 15500 msecs.
+  * baseline(v1.0.x): 6800 msecs(2.3x faster than old version)
+  * optimised: 1500 msecs(10x faster than old version, 4.5x faster than basedline)
+
+
+Tests
+-----
+
+Unit tests are provided in `tests` directory. See `tests/README.md` for details.

appveyor.yml

@@ -1,12 +1,23 @@
 version: 0.9.{build}
 
-# scripts that runs after repo cloning.
-install:
-  - vcsetup.bat
-
 platform: x64
-configuration: Release
 
-build:
-  parallel: true
-  project: TinyObjLoaderSolution.sln
+install:
+  #######################################################################################
+  # All external dependencies are installed in C:\projects\deps
+  #######################################################################################
+  - mkdir C:\projects\deps
+  - cd C:\projects\deps
+
+  #######################################################################################
+  # Install Ninja
+  #######################################################################################
+  - set NINJA_URL="https://github.com/ninja-build/ninja/releases/download/v1.6.0/ninja-win.zip"
+  - appveyor DownloadFile %NINJA_URL% -FileName ninja.zip
+  - 7z x ninja.zip -oC:\projects\deps\ninja > nul
+  - set PATH=C:\projects\deps\ninja;%PATH%
+  - ninja --version
+
+build_script:
+  - cd tests
+  - vcbuild.bat

build.ninja

@@ -1,16 +1,53 @@
-# build.ninja
-cc     = clang
-cxx    = clang++
-cflags = -Werror -Weverything
-cxxflags = -Werror -Weverything
+ninja_required_version = 1.4
 
-rule compile
-  command = $cxx $cxxflags -c $in -o $out
+gnubuilddir = build
+gnudefines = 
+gnuincludes = -I.
+gnucflags = -O2 -g
+gnucxxflags = -O2 -g -pedantic -Wall -Wextra -Wcast-align -Wcast-qual $
+    -Wctor-dtor-privacy -Wdisabled-optimization -Wformat=2 -Winit-self $
+    -Wmissing-declarations -Wmissing-include-dirs -Wold-style-cast $
+    -Woverloaded-virtual -Wredundant-decls -Wshadow -Wsign-conversion $
+    -Wsign-promo -Wstrict-overflow=5 -Wswitch-default -Wundef -Werror $
+    -Wno-unused -fsanitize=address
+gnuldflags = -fsanitize=address
 
-rule link
-  command = $cxx $in -o $out
+pool link_pool
+  depth = 1
 
-build test.o: compile test.cc
-build test: link test.o
+rule gnucxx
+  command = $gnucxx -MMD -MF $out.d $gnudefines $gnuincludes $gnucxxflags $
+      -c $in -o $out
+  description = CXX $out
+  depfile = $out.d
+  deps = gcc
+rule gnucc
+  command = $gnucc -MMD -MF $out.d $gnudefines $gnuincludes $gnucflags -c $
+      $in -o $out
+  description = CC $out
+  depfile = $out.d
+  deps = gcc
+rule gnulink
+  command = $gnuld -o $out $in $libs $gnuldflags
+  description = LINK $out
+  pool = link_pool
+rule gnuar
+  command = $gnuar rsc $out $in
+  description = AR $out
+  pool = link_pool
+rule gnustamp
+  command = touch $out
+  description = STAMP $out
 
-default test
+gnucxx = g++
+gnucc = gcc
+gnuld = $gnucxx
+gnuar = ar
+
+build loader_example.o: gnucxx loader_example.cc
+
+
+build loader_example: gnulink loader_example.o
+build all: phony loader_example
+
+default all

examples/callback_api/Makefile

@@ -0,0 +1,2 @@
+all:
+	clang++ -I../../ -Wall -g -o example main.cc

examples/callback_api/main.cc

@@ -0,0 +1,155 @@
+#define TINYOBJLOADER_IMPLEMENTATION
+#include "tiny_obj_loader.h"
+
+#include <cstdio>
+#include <cstdlib>
+#include <cassert>
+#include <iostream>
+#include <sstream>
+#include <fstream>
+
+typedef struct
+{
+  std::vector<float> vertices;
+  std::vector<float> normals;
+  std::vector<float> texcoords;
+  std::vector<int>   v_indices;
+  std::vector<int>   vn_indices;
+  std::vector<int>   vt_indices;
+
+  std::vector<tinyobj::material_t> materials;
+
+} MyMesh;
+
+void vertex_cb(void *user_data, float x, float y, float z)
+{
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("v[%ld] = %f, %f, %f\n", mesh->vertices.size() / 3, x, y, z);
+
+  mesh->vertices.push_back(x);
+  mesh->vertices.push_back(y);
+  mesh->vertices.push_back(z);
+}
+
+void normal_cb(void *user_data, float x, float y, float z)
+{
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("vn[%ld] = %f, %f, %f\n", mesh->normals.size() / 3, x, y, z);
+
+  mesh->normals.push_back(x);
+  mesh->normals.push_back(y);
+  mesh->normals.push_back(z);
+}
+
+void texcoord_cb(void *user_data, float x, float y)
+{
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("vt[%ld] = %f, %f\n", mesh->texcoords.size() / 2, x, y);
+
+  mesh->texcoords.push_back(x);
+  mesh->texcoords.push_back(y);
+}
+
+void index_cb(void *user_data, int v_idx, int vn_idx, int vt_idx)
+{
+  // NOTE: the value of each index is raw value. 
+  // For example, the application must manually adjust the index with offset
+  // (e.g. v_indices.size()) when the value is negative(relative index).
+  // See fixIndex() function in tiny_obj_loader.h for details.
+  // Also, -2147483648(0x80000000) is set for the index value which does not exist in .obj 
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("idx[%ld] = %d, %d, %d\n", mesh->v_indices.size(), v_idx, vn_idx, vt_idx);
+
+  if (v_idx != 0x80000000) {
+    mesh->v_indices.push_back(v_idx);
+  }
+  if (vn_idx != 0x80000000) {
+    mesh->vn_indices.push_back(vn_idx);
+  }
+  if (vt_idx != 0x80000000) {
+    mesh->vt_indices.push_back(vt_idx);
+  }
+}
+
+void usemtl_cb(void *user_data, const char* name, int material_idx)
+{
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  if ((material_idx > -1) && (material_idx < mesh->materials.size())) {
+    printf("usemtl. material id = %d(name = %s)\n", material_idx, mesh->materials[material_idx].name.c_str());
+  } else {
+    printf("usemtl. name = %s\n", name);
+  }
+}
+
+void mtllib_cb(void *user_data, const tinyobj::material_t *materials, int num_materials)
+{
+  MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("mtllib. # of materials = %d\n", num_materials);
+
+  for (int i = 0; i < num_materials; i++) {
+    mesh->materials.push_back(materials[i]);
+  }
+}
+
+void group_cb(void *user_data, const char **names, int num_names)
+{
+  //MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("group : name = \n");
+
+  for (int i = 0; i < num_names; i++) {
+    printf("  %s\n", names[i]);
+  }
+}
+
+void object_cb(void *user_data, const char *name)
+{
+  //MyMesh *mesh = reinterpret_cast<MyMesh*>(user_data);
+  printf("object : name = %s\n", name);
+
+}
+
+int
+main(int argc, char** argv)
+{
+  tinyobj::callback_t cb;
+  cb.vertex_cb = vertex_cb;
+  cb.normal_cb = normal_cb;
+  cb.texcoord_cb = texcoord_cb;
+  cb.index_cb = index_cb;
+  cb.usemtl_cb = usemtl_cb;
+  cb.mtllib_cb = mtllib_cb;
+  cb.group_cb = group_cb;
+  cb.object_cb = object_cb;
+
+  MyMesh mesh;
+  std::string err;
+  std::ifstream ifs("../../models/cornell_box.obj");
+
+  if (ifs.fail()) {
+    std::cerr << "file not found." << std::endl;
+    return EXIT_FAILURE;
+  }
+
+  tinyobj::MaterialFileReader mtlReader("../../models/");
+  
+  bool ret = tinyobj::LoadObjWithCallback(&mesh, cb, &err, &ifs, &mtlReader);
+
+  if (!err.empty()) {
+    std::cerr << err << std::endl;
+  }
+
+  if (!ret) {
+    std::cerr << "Failed to parse .obj" << std::endl;
+    return EXIT_FAILURE;
+  }
+
+  printf("# of vertices         = %ld\n", mesh.vertices.size() / 3);
+  printf("# of normals          = %ld\n", mesh.normals.size() / 3);
+  printf("# of texcoords        = %ld\n", mesh.texcoords.size() / 2);
+  printf("# of vertex indices   = %ld\n", mesh.v_indices.size());
+  printf("# of normal indices   = %ld\n", mesh.vn_indices.size());
+  printf("# of texcoord indices   = %ld\n", mesh.vt_indices.size());
+  printf("# of materials = %ld\n", mesh.materials.size());
+
+  return EXIT_SUCCESS;
+}

examples/viewer/README.md

@@ -0,0 +1,37 @@
+# Simple .obj viewer with glew + glfw3 + OpenGL
+
+## Requirements
+
+* premake5
+* glfw3
+* glew
+
+
+## Build on MaCOSX
+
+Install glfw3 and glew using brew.
+Then,
+
+    $ premake5 gmake
+    $ make
+
+## Build on Linux
+
+Set `PKG_CONFIG_PATH` or Edit path to glfw3 and glew in `premake4.lua`
+
+Then,
+
+    $ premake5 gmake
+    $ make
+
+## Build on Windows.
+
+* Visual Studio 2013
+* Windows 64bit
+  * 32bit may work.
+
+Put glfw3 and glew library somewhere and replace include and lib path in `premake4.lua`
+
+Then,
+
+    > premake5.exe vs2013

examples/viewer/premake4.lua

@@ -0,0 +1,44 @@
+solution "objview"
+	-- location ( "build" )
+	configurations { "Release", "Debug" }
+	platforms {"native", "x64", "x32"}
+	
+	project "objview"
+
+		kind "ConsoleApp"
+		language "C++"
+		files { "viewer.cc", "trackball.cc" }
+		includedirs { "./" }
+		includedirs { "../../" }
+
+		configuration { "linux" }
+			linkoptions { "`pkg-config --libs glfw3`" }
+			links { "GL", "GLU", "m", "GLEW", "X11", "Xrandr", "Xinerama", "Xi", "Xxf86vm", "Xcursor", "dl" }
+			linkoptions { "-pthread" }
+
+		configuration { "windows" }
+			-- Path to GLFW3
+			includedirs { '../../../../local/glfw-3.1.2.bin.WIN64/include' }
+			libdirs { '../../../../local/glfw-3.1.2.bin.WIN64/lib-vc2013' }
+			-- Path to GLEW
+			includedirs { '../../../../local/glew-1.13.0/include' }
+			libdirs { '../../../../local/glew-1.13.0/lib/Release/x64' }
+
+			links { "glfw3", "glew32", "gdi32", "winmm", "user32", "glu32","opengl32", "kernel32" }
+			defines { "_CRT_SECURE_NO_WARNINGS" }
+
+		configuration { "macosx" }
+			includedirs { "/usr/local/include" }
+			buildoptions { "-Wno-deprecated-declarations" }
+			libdirs { "/usr/local/lib" }
+			links { "glfw3", "GLEW" }
+			linkoptions { "-framework OpenGL", "-framework Cocoa", "-framework IOKit", "-framework CoreVideo" }
+
+		configuration "Debug"
+			defines { "DEBUG" }
+			flags { "Symbols", "ExtraWarnings"}
+
+		configuration "Release"
+			defines { "NDEBUG" }
+			flags { "Optimize", "ExtraWarnings"}
+

examples/viewer/trackball.cc

@@ -0,0 +1,292 @@
+/*
+ * (c) Copyright 1993, 1994, Silicon Graphics, Inc.
+ * ALL RIGHTS RESERVED
+ * Permission to use, copy, modify, and distribute this software for
+ * any purpose and without fee is hereby granted, provided that the above
+ * copyright notice appear in all copies and that both the copyright notice
+ * and this permission notice appear in supporting documentation, and that
+ * the name of Silicon Graphics, Inc. not be used in advertising
+ * or publicity pertaining to distribution of the software without specific,
+ * written prior permission.
+ *
+ * THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU "AS-IS"
+ * AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR OTHERWISE,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY OR
+ * FITNESS FOR A PARTICULAR PURPOSE.  IN NO EVENT SHALL SILICON
+ * GRAPHICS, INC.  BE LIABLE TO YOU OR ANYONE ELSE FOR ANY DIRECT,
+ * SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY
+ * KIND, OR ANY DAMAGES WHATSOEVER, INCLUDING WITHOUT LIMITATION,
+ * LOSS OF PROFIT, LOSS OF USE, SAVINGS OR REVENUE, OR THE CLAIMS OF
+ * THIRD PARTIES, WHETHER OR NOT SILICON GRAPHICS, INC.  HAS BEEN
+ * ADVISED OF THE POSSIBILITY OF SUCH LOSS, HOWEVER CAUSED AND ON
+ * ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
+ * POSSESSION, USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * US Government Users Restricted Rights
+ * Use, duplication, or disclosure by the Government is subject to
+ * restrictions set forth in FAR 52.227.19(c)(2) or subparagraph
+ * (c)(1)(ii) of the Rights in Technical Data and Computer Software
+ * clause at DFARS 252.227-7013 and/or in similar or successor
+ * clauses in the FAR or the DOD or NASA FAR Supplement.
+ * Unpublished-- rights reserved under the copyright laws of the
+ * United States.  Contractor/manufacturer is Silicon Graphics,
+ * Inc., 2011 N.  Shoreline Blvd., Mountain View, CA 94039-7311.
+ *
+ * OpenGL(TM) is a trademark of Silicon Graphics, Inc.
+ */
+/*
+ * Trackball code:
+ *
+ * Implementation of a virtual trackball.
+ * Implemented by Gavin Bell, lots of ideas from Thant Tessman and
+ *   the August '88 issue of Siggraph's "Computer Graphics," pp. 121-129.
+ *
+ * Vector manip code:
+ *
+ * Original code from:
+ * David M. Ciemiewicz, Mark Grossman, Henry Moreton, and Paul Haeberli
+ *
+ * Much mucking with by:
+ * Gavin Bell
+ */
+#include <math.h>
+#include "trackball.h"
+
+/*
+ * This size should really be based on the distance from the center of
+ * rotation to the point on the object underneath the mouse.  That
+ * point would then track the mouse as closely as possible.  This is a
+ * simple example, though, so that is left as an Exercise for the
+ * Programmer.
+ */
+#define TRACKBALLSIZE (0.8)
+
+/*
+ * Local function prototypes (not defined in trackball.h)
+ */
+static float tb_project_to_sphere(float, float, float);
+static void normalize_quat(float[4]);
+
+static void vzero(float *v) {
+  v[0] = 0.0;
+  v[1] = 0.0;
+  v[2] = 0.0;
+}
+
+static void vset(float *v, float x, float y, float z) {
+  v[0] = x;
+  v[1] = y;
+  v[2] = z;
+}
+
+static void vsub(const float *src1, const float *src2, float *dst) {
+  dst[0] = src1[0] - src2[0];
+  dst[1] = src1[1] - src2[1];
+  dst[2] = src1[2] - src2[2];
+}
+
+static void vcopy(const float *v1, float *v2) {
+  register int i;
+  for (i = 0; i < 3; i++)
+    v2[i] = v1[i];
+}
+
+static void vcross(const float *v1, const float *v2, float *cross) {
+  float temp[3];
+
+  temp[0] = (v1[1] * v2[2]) - (v1[2] * v2[1]);
+  temp[1] = (v1[2] * v2[0]) - (v1[0] * v2[2]);
+  temp[2] = (v1[0] * v2[1]) - (v1[1] * v2[0]);
+  vcopy(temp, cross);
+}
+
+static float vlength(const float *v) {
+  return sqrt(v[0] * v[0] + v[1] * v[1] + v[2] * v[2]);
+}
+
+static void vscale(float *v, float div) {
+  v[0] *= div;
+  v[1] *= div;
+  v[2] *= div;
+}
+
+static void vnormal(float *v) { vscale(v, 1.0 / vlength(v)); }
+
+static float vdot(const float *v1, const float *v2) {
+  return v1[0] * v2[0] + v1[1] * v2[1] + v1[2] * v2[2];
+}
+
+static void vadd(const float *src1, const float *src2, float *dst) {
+  dst[0] = src1[0] + src2[0];
+  dst[1] = src1[1] + src2[1];
+  dst[2] = src1[2] + src2[2];
+}
+
+/*
+ * Ok, simulate a track-ball.  Project the points onto the virtual
+ * trackball, then figure out the axis of rotation, which is the cross
+ * product of P1 P2 and O P1 (O is the center of the ball, 0,0,0)
+ * Note:  This is a deformed trackball-- is a trackball in the center,
+ * but is deformed into a hyperbolic sheet of rotation away from the
+ * center.  This particular function was chosen after trying out
+ * several variations.
+ *
+ * It is assumed that the arguments to this routine are in the range
+ * (-1.0 ... 1.0)
+ */
+void trackball(float q[4], float p1x, float p1y, float p2x, float p2y) {
+  float a[3]; /* Axis of rotation */
+  float phi;  /* how much to rotate about axis */
+  float p1[3], p2[3], d[3];
+  float t;
+
+  if (p1x == p2x && p1y == p2y) {
+    /* Zero rotation */
+    vzero(q);
+    q[3] = 1.0;
+    return;
+  }
+
+  /*
+   * First, figure out z-coordinates for projection of P1 and P2 to
+   * deformed sphere
+   */
+  vset(p1, p1x, p1y, tb_project_to_sphere(TRACKBALLSIZE, p1x, p1y));
+  vset(p2, p2x, p2y, tb_project_to_sphere(TRACKBALLSIZE, p2x, p2y));
+
+  /*
+   *  Now, we want the cross product of P1 and P2
+   */
+  vcross(p2, p1, a);
+
+  /*
+   *  Figure out how much to rotate around that axis.
+   */
+  vsub(p1, p2, d);
+  t = vlength(d) / (2.0 * TRACKBALLSIZE);
+
+  /*
+   * Avoid problems with out-of-control values...
+   */
+  if (t > 1.0)
+    t = 1.0;
+  if (t < -1.0)
+    t = -1.0;
+  phi = 2.0 * asin(t);
+
+  axis_to_quat(a, phi, q);
+}
+
+/*
+ *  Given an axis and angle, compute quaternion.
+ */
+void axis_to_quat(float a[3], float phi, float q[4]) {
+  vnormal(a);
+  vcopy(a, q);
+  vscale(q, sin(phi / 2.0));
+  q[3] = cos(phi / 2.0);
+}
+
+/*
+ * Project an x,y pair onto a sphere of radius r OR a hyperbolic sheet
+ * if we are away from the center of the sphere.
+ */
+static float tb_project_to_sphere(float r, float x, float y) {
+  float d, t, z;
+
+  d = sqrt(x * x + y * y);
+  if (d < r * 0.70710678118654752440) { /* Inside sphere */
+    z = sqrt(r * r - d * d);
+  } else { /* On hyperbola */
+    t = r / 1.41421356237309504880;
+    z = t * t / d;
+  }
+  return z;
+}
+
+/*
+ * Given two rotations, e1 and e2, expressed as quaternion rotations,
+ * figure out the equivalent single rotation and stuff it into dest.
+ *
+ * This routine also normalizes the result every RENORMCOUNT times it is
+ * called, to keep error from creeping in.
+ *
+ * NOTE: This routine is written so that q1 or q2 may be the same
+ * as dest (or each other).
+ */
+
+#define RENORMCOUNT 97
+
+void add_quats(float q1[4], float q2[4], float dest[4]) {
+  static int count = 0;
+  float t1[4], t2[4], t3[4];
+  float tf[4];
+
+  vcopy(q1, t1);
+  vscale(t1, q2[3]);
+
+  vcopy(q2, t2);
+  vscale(t2, q1[3]);
+
+  vcross(q2, q1, t3);
+  vadd(t1, t2, tf);
+  vadd(t3, tf, tf);
+  tf[3] = q1[3] * q2[3] - vdot(q1, q2);
+
+  dest[0] = tf[0];
+  dest[1] = tf[1];
+  dest[2] = tf[2];
+  dest[3] = tf[3];
+
+  if (++count > RENORMCOUNT) {
+    count = 0;
+    normalize_quat(dest);
+  }
+}
+
+/*
+ * Quaternions always obey:  a^2 + b^2 + c^2 + d^2 = 1.0
+ * If they don't add up to 1.0, dividing by their magnitued will
+ * renormalize them.
+ *
+ * Note: See the following for more information on quaternions:
+ *
+ * - Shoemake, K., Animating rotation with quaternion curves, Computer
+ *   Graphics 19, No 3 (Proc. SIGGRAPH'85), 245-254, 1985.
+ * - Pletinckx, D., Quaternion calculus as a basic tool in computer
+ *   graphics, The Visual Computer 5, 2-13, 1989.
+ */
+static void normalize_quat(float q[4]) {
+  int i;
+  float mag;
+
+  mag = (q[0] * q[0] + q[1] * q[1] + q[2] * q[2] + q[3] * q[3]);
+  for (i = 0; i < 4; i++)
+    q[i] /= mag;
+}
+
+/*
+ * Build a rotation matrix, given a quaternion rotation.
+ *
+ */
+void build_rotmatrix(float m[4][4], const float q[4]) {
+  m[0][0] = 1.0 - 2.0 * (q[1] * q[1] + q[2] * q[2]);
+  m[0][1] = 2.0 * (q[0] * q[1] - q[2] * q[3]);
+  m[0][2] = 2.0 * (q[2] * q[0] + q[1] * q[3]);
+  m[0][3] = 0.0;
+
+  m[1][0] = 2.0 * (q[0] * q[1] + q[2] * q[3]);
+  m[1][1] = 1.0 - 2.0 * (q[2] * q[2] + q[0] * q[0]);
+  m[1][2] = 2.0 * (q[1] * q[2] - q[0] * q[3]);
+  m[1][3] = 0.0;
+
+  m[2][0] = 2.0 * (q[2] * q[0] - q[1] * q[3]);
+  m[2][1] = 2.0 * (q[1] * q[2] + q[0] * q[3]);
+  m[2][2] = 1.0 - 2.0 * (q[1] * q[1] + q[0] * q[0]);
+  m[2][3] = 0.0;
+
+  m[3][0] = 0.0;
+  m[3][1] = 0.0;
+  m[3][2] = 0.0;
+  m[3][3] = 1.0;
+}

examples/viewer/trackball.h

@@ -0,0 +1,75 @@
+/*
+ * (c) Copyright 1993, 1994, Silicon Graphics, Inc.
+ * ALL RIGHTS RESERVED
+ * Permission to use, copy, modify, and distribute this software for
+ * any purpose and without fee is hereby granted, provided that the above
+ * copyright notice appear in all copies and that both the copyright notice
+ * and this permission notice appear in supporting documentation, and that
+ * the name of Silicon Graphics, Inc. not be used in advertising
+ * or publicity pertaining to distribution of the software without specific,
+ * written prior permission.
+ *
+ * THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU "AS-IS"
+ * AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR OTHERWISE,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY OR
+ * FITNESS FOR A PARTICULAR PURPOSE.  IN NO EVENT SHALL SILICON
+ * GRAPHICS, INC.  BE LIABLE TO YOU OR ANYONE ELSE FOR ANY DIRECT,
+ * SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY
+ * KIND, OR ANY DAMAGES WHATSOEVER, INCLUDING WITHOUT LIMITATION,
+ * LOSS OF PROFIT, LOSS OF USE, SAVINGS OR REVENUE, OR THE CLAIMS OF
+ * THIRD PARTIES, WHETHER OR NOT SILICON GRAPHICS, INC.  HAS BEEN
+ * ADVISED OF THE POSSIBILITY OF SUCH LOSS, HOWEVER CAUSED AND ON
+ * ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
+ * POSSESSION, USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * US Government Users Restricted Rights
+ * Use, duplication, or disclosure by the Government is subject to
+ * restrictions set forth in FAR 52.227.19(c)(2) or subparagraph
+ * (c)(1)(ii) of the Rights in Technical Data and Computer Software
+ * clause at DFARS 252.227-7013 and/or in similar or successor
+ * clauses in the FAR or the DOD or NASA FAR Supplement.
+ * Unpublished-- rights reserved under the copyright laws of the
+ * United States.  Contractor/manufacturer is Silicon Graphics,
+ * Inc., 2011 N.  Shoreline Blvd., Mountain View, CA 94039-7311.
+ *
+ * OpenGL(TM) is a trademark of Silicon Graphics, Inc.
+ */
+/*
+ * trackball.h
+ * A virtual trackball implementation
+ * Written by Gavin Bell for Silicon Graphics, November 1988.
+ */
+
+/*
+ * Pass the x and y coordinates of the last and current positions of
+ * the mouse, scaled so they are from (-1.0 ... 1.0).
+ *
+ * The resulting rotation is returned as a quaternion rotation in the
+ * first paramater.
+ */
+void trackball(float q[4], float p1x, float p1y, float p2x, float p2y);
+
+void negate_quat(float *q, float *qn);
+
+/*
+ * Given two quaternions, add them together to get a third quaternion.
+ * Adding quaternions to get a compound rotation is analagous to adding
+ * translations to get a compound translation.  When incrementally
+ * adding rotations, the first argument here should be the new
+ * rotation, the second and third the total rotation (which will be
+ * over-written with the resulting new total rotation).
+ */
+void add_quats(float *q1, float *q2, float *dest);
+
+/*
+ * A useful function, builds a rotation matrix in Matrix based on
+ * given quaternion.
+ */
+void build_rotmatrix(float m[4][4], const float q[4]);
+
+/*
+ * This function computes a quaternion based on an axis (defined by
+ * the given vector) and an angle about which to rotate.  The angle is
+ * expressed in radians.  The result is put into the third argument.
+ */
+void axis_to_quat(float a[3], float phi, float q[4]);

examples/viewer/viewer.cc

@@ -0,0 +1,528 @@
+//
+// Simple .obj viewer(vertex only)
+//
+#include <vector>
+#include <string>
+#include <cstdio>
+#include <cstdlib>
+#include <iostream>
+#include <limits>
+#include <cmath>
+#include <cassert>
+#include <algorithm>	
+
+#include <GL/glew.h>
+
+#ifdef __APPLE__
+#include <OpenGL/glu.h>
+#else
+#include <GL/glu.h>
+#endif
+
+#include <GLFW/glfw3.h>
+
+#define TINYOBJLOADER_IMPLEMENTATION
+#include "../../tiny_obj_loader.h"
+
+#include "trackball.h"
+
+#ifdef _WIN32
+#ifdef __cplusplus
+extern "C" {
+#endif
+#include <windows.h>
+#include <mmsystem.h>
+#ifdef __cplusplus
+}
+#endif
+#pragma comment(lib, "winmm.lib")
+#else
+#if defined(__unix__) || defined(__APPLE__)
+#include <sys/time.h>
+#else
+#include <ctime>
+#endif
+#endif
+
+class timerutil {
+public:
+#ifdef _WIN32
+  typedef DWORD time_t;
+
+  timerutil() { ::timeBeginPeriod(1); }
+  ~timerutil() { ::timeEndPeriod(1); }
+
+  void start() { t_[0] = ::timeGetTime(); }
+  void end() { t_[1] = ::timeGetTime(); }
+
+  time_t sec() { return (time_t)((t_[1] - t_[0]) / 1000); }
+  time_t msec() { return (time_t)((t_[1] - t_[0])); }
+  time_t usec() { return (time_t)((t_[1] - t_[0]) * 1000); }
+  time_t current() { return ::timeGetTime(); }
+
+#else
+#if defined(__unix__) || defined(__APPLE__)
+  typedef unsigned long int time_t;
+
+  void start() { gettimeofday(tv + 0, &tz); }
+  void end() { gettimeofday(tv + 1, &tz); }
+
+  time_t sec() { return (time_t)(tv[1].tv_sec - tv[0].tv_sec); }
+  time_t msec() {
+    return this->sec() * 1000 +
+           (time_t)((tv[1].tv_usec - tv[0].tv_usec) / 1000);
+  }
+  time_t usec() {
+    return this->sec() * 1000000 + (time_t)(tv[1].tv_usec - tv[0].tv_usec);
+  }
+  time_t current() {
+    struct timeval t;
+    gettimeofday(&t, NULL);
+    return (time_t)(t.tv_sec * 1000 + t.tv_usec);
+  }
+
+#else // C timer
+  // using namespace std;
+  typedef clock_t time_t;
+
+  void start() { t_[0] = clock(); }
+  void end() { t_[1] = clock(); }
+
+  time_t sec() { return (time_t)((t_[1] - t_[0]) / CLOCKS_PER_SEC); }
+  time_t msec() { return (time_t)((t_[1] - t_[0]) * 1000 / CLOCKS_PER_SEC); }
+  time_t usec() { return (time_t)((t_[1] - t_[0]) * 1000000 / CLOCKS_PER_SEC); }
+  time_t current() { return (time_t)clock(); }
+
+#endif
+#endif
+
+private:
+#ifdef _WIN32
+  DWORD t_[2];
+#else
+#if defined(__unix__) || defined(__APPLE__)
+  struct timeval tv[2];
+  struct timezone tz;
+#else
+  time_t t_[2];
+#endif
+#endif
+};
+
+typedef struct {
+  GLuint vb;    // vertex buffer
+  int numTriangles;
+} DrawObject;
+
+std::vector<DrawObject> gDrawObjects;
+
+int width = 768;
+int height = 768;
+
+double prevMouseX, prevMouseY;
+bool mouseLeftPressed;
+bool mouseMiddlePressed;
+bool mouseRightPressed;
+float curr_quat[4];
+float prev_quat[4];
+float eye[3], lookat[3], up[3];
+
+GLFWwindow* window;
+
+void CheckErrors(std::string desc) {
+  GLenum e = glGetError();
+  if (e != GL_NO_ERROR) {
+    fprintf(stderr, "OpenGL error in \"%s\": %d (%d)\n", desc.c_str(), e, e);
+    exit(20);
+  }
+}
+
+void CalcNormal(float N[3], float v0[3], float v1[3], float v2[3]) {
+  float v10[3];
+  v10[0] = v1[0] - v0[0];
+  v10[1] = v1[1] - v0[1];
+  v10[2] = v1[2] - v0[2];
+
+  float v20[3];
+  v20[0] = v2[0] - v0[0];
+  v20[1] = v2[1] - v0[1];
+  v20[2] = v2[2] - v0[2];
+
+  N[0] = v20[1] * v10[2] - v20[2] * v10[1];
+  N[1] = v20[2] * v10[0] - v20[0] * v10[2];
+  N[2] = v20[0] * v10[1] - v20[1] * v10[0];
+
+  float len2 = N[0] * N[0] + N[1] * N[1] + N[2] * N[2];
+  if (len2 > 0.0f) {
+    float len = sqrtf(len2);
+    
+    N[0] /= len;
+    N[1] /= len;
+  }
+}
+
+bool LoadObjAndConvert(float bmin[3], float bmax[3], std::vector<DrawObject>& drawObjects, const char* filename)
+{
+  tinyobj::attrib_t attrib;
+  std::vector<tinyobj::shape_t> shapes;
+  std::vector<tinyobj::material_t> materials;
+
+  timerutil tm;
+
+  tm.start();
+  
+  std::string err;
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, filename, NULL);
+  if (!err.empty()) {
+    std::cerr << err << std::endl;
+  }
+
+  tm.end();
+  
+  if (!ret) {
+    std::cerr << "Failed to load " << filename << std::endl;
+    return false;
+  }
+
+  printf("Parsing time: %d [ms]\n", tm.msec());
+
+  printf("# of vertices  = %d\n", (int)(attrib.vertices.size()) / 3);
+  printf("# of normals   = %d\n", (int)(attrib.normals.size()) / 3);
+  printf("# of texcoords = %d\n", (int)(attrib.texcoords.size()) / 2);
+  printf("# of materials = %d\n", (int)materials.size());
+  printf("# of shapes    = %d\n", (int)shapes.size());
+
+  bmin[0] = bmin[1] = bmin[2] = std::numeric_limits<float>::max();
+  bmax[0] = bmax[1] = bmax[2] = -std::numeric_limits<float>::max();
+
+  {
+      for (size_t s = 0; s < shapes.size(); s++) {
+        DrawObject o;
+        std::vector<float> vb; // pos(3float), normal(3float), color(3float)
+        for (size_t f = 0; f < shapes[s].mesh.indices.size()/3; f++) {
+
+          tinyobj::index_t idx0 = shapes[s].mesh.indices[3*f+0];
+          tinyobj::index_t idx1 = shapes[s].mesh.indices[3*f+1];
+          tinyobj::index_t idx2 = shapes[s].mesh.indices[3*f+2];
+
+          float v[3][3];
+          for (int k = 0; k < 3; k++) {
+            int f0 = idx0.vertex_index;
+            int f1 = idx1.vertex_index;
+            int f2 = idx2.vertex_index;
+            assert(f0 >= 0);
+            assert(f1 >= 0);
+            assert(f2 >= 0);
+
+            v[0][k] = attrib.vertices[3*f0+k];
+            v[1][k] = attrib.vertices[3*f1+k];
+            v[2][k] = attrib.vertices[3*f2+k];
+            bmin[k] = std::min(v[0][k], bmin[k]);
+            bmin[k] = std::min(v[1][k], bmin[k]);
+            bmin[k] = std::min(v[2][k], bmin[k]);
+            bmax[k] = std::max(v[0][k], bmax[k]);
+            bmax[k] = std::max(v[1][k], bmax[k]);
+            bmax[k] = std::max(v[2][k], bmax[k]);
+          }
+
+          float n[3][3];
+
+          if (attrib.normals.size() > 0) { 
+            int f0 = idx0.normal_index;
+            int f1 = idx1.normal_index;
+            int f2 = idx2.normal_index;
+            assert(f0 >= 0);
+            assert(f1 >= 0);
+            assert(f2 >= 0);
+            for (int k = 0; k < 3; k++) {
+              n[0][k] = attrib.normals[3*f0+k];
+              n[1][k] = attrib.normals[3*f1+k];
+              n[2][k] = attrib.normals[3*f2+k];
+            }
+          } else {
+            // compute geometric normal
+            CalcNormal(n[0], v[0], v[1], v[2]);
+            n[1][0] = n[0][0]; n[1][1] = n[0][1]; n[1][2] = n[0][2];
+            n[2][0] = n[0][0]; n[2][1] = n[0][1]; n[2][2] = n[0][2];
+          }
+
+          for (int k = 0; k < 3; k++) {
+            vb.push_back(v[k][0]);
+            vb.push_back(v[k][1]);
+            vb.push_back(v[k][2]);
+            vb.push_back(n[k][0]);
+            vb.push_back(n[k][1]);
+            vb.push_back(n[k][2]);
+            // Use normal as color.
+            float c[3] = {n[k][0], n[k][1], n[k][2]};
+            float len2 = c[0] * c[0] + c[1] * c[1] + c[2] * c[2];
+            if (len2 > 0.0f) {
+              float len = sqrtf(len2);
+              
+              c[0] /= len;
+              c[1] /= len;
+              c[2] /= len;
+            }
+            vb.push_back(c[0] * 0.5 + 0.5);
+            vb.push_back(c[1] * 0.5 + 0.5);
+            vb.push_back(c[2] * 0.5 + 0.5);
+          }
+          
+        }
+
+        o.vb = 0;
+        o.numTriangles = 0;
+        if (vb.size() > 0) {
+          glGenBuffers(1, &o.vb);
+          glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+          glBufferData(GL_ARRAY_BUFFER, vb.size() * sizeof(float), &vb.at(0), GL_STATIC_DRAW);
+          o.numTriangles = vb.size() / 9 / 3;
+          printf("shape[%d] # of triangles = %d\n", static_cast<int>(s), o.numTriangles);
+        }
+
+        gDrawObjects.push_back(o);
+    }
+  }
+  
+  printf("bmin = %f, %f, %f\n", bmin[0], bmin[1], bmin[2]);
+  printf("bmax = %f, %f, %f\n", bmax[0], bmax[1], bmax[2]);
+
+  return true;
+}
+
+void reshapeFunc(GLFWwindow* window, int w, int h)
+{
+  int fb_w, fb_h;
+  // Get actual framebuffer size.
+  glfwGetFramebufferSize(window, &fb_w, &fb_h);
+
+  glViewport(0, 0, fb_w, fb_h);
+  glMatrixMode(GL_PROJECTION);
+  glLoadIdentity();
+  gluPerspective(45.0, (float)w / (float)h, 0.01f, 100.0f);
+  glMatrixMode(GL_MODELVIEW);
+  glLoadIdentity();
+
+  width = w;
+  height = h;
+}
+
+void keyboardFunc(GLFWwindow *window, int key, int scancode, int action, int mods) {
+    if(action == GLFW_PRESS || action == GLFW_REPEAT){
+        // Move camera
+        float mv_x = 0, mv_y = 0, mv_z = 0;
+        if(key == GLFW_KEY_K) mv_x += 1;
+        else if(key == GLFW_KEY_J) mv_x += -1;
+        else if(key == GLFW_KEY_L) mv_y += 1;
+        else if(key == GLFW_KEY_H) mv_y += -1;
+        else if(key == GLFW_KEY_P) mv_z += 1;
+        else if(key == GLFW_KEY_N) mv_z += -1;
+        //camera.move(mv_x * 0.05, mv_y * 0.05, mv_z * 0.05);
+        // Close window
+        if(key == GLFW_KEY_Q || key == GLFW_KEY_ESCAPE) glfwSetWindowShouldClose(window, GL_TRUE);
+
+        //init_frame = true;
+    }
+}
+
+void clickFunc(GLFWwindow* window, int button, int action, int mods){
+    if(button == GLFW_MOUSE_BUTTON_LEFT){
+        if(action == GLFW_PRESS){
+            mouseLeftPressed = true;
+            trackball(prev_quat, 0.0, 0.0, 0.0, 0.0);
+        } else if(action == GLFW_RELEASE){
+            mouseLeftPressed = false;
+        }
+    }
+    if(button == GLFW_MOUSE_BUTTON_RIGHT){
+        if(action == GLFW_PRESS){
+            mouseRightPressed = true;
+        } else if(action == GLFW_RELEASE){
+            mouseRightPressed = false;
+        }
+    }
+    if(button == GLFW_MOUSE_BUTTON_MIDDLE){
+        if(action == GLFW_PRESS){
+            mouseMiddlePressed = true;
+        } else if(action == GLFW_RELEASE){
+            mouseMiddlePressed = false;
+        }
+    }
+}
+
+void motionFunc(GLFWwindow* window, double mouse_x, double mouse_y){
+  float rotScale = 1.0f;
+  float transScale = 2.0f;
+
+    if(mouseLeftPressed){
+      trackball(prev_quat,
+          rotScale * (2.0f * prevMouseX - width) / (float)width,
+          rotScale * (height - 2.0f * prevMouseY) / (float)height,
+          rotScale * (2.0f * mouse_x - width) / (float)width,
+          rotScale * (height - 2.0f * mouse_y) / (float)height);
+
+      add_quats(prev_quat, curr_quat, curr_quat);
+    } else if (mouseMiddlePressed) {
+      eye[0] -= transScale * (mouse_x - prevMouseX) / (float)width;
+      lookat[0] -= transScale * (mouse_x - prevMouseX) / (float)width;
+      eye[1] += transScale * (mouse_y - prevMouseY) / (float)height;
+      lookat[1] += transScale * (mouse_y - prevMouseY) / (float)height;
+    } else if (mouseRightPressed) {
+      eye[2] += transScale * (mouse_y - prevMouseY) / (float)height;
+      lookat[2] += transScale * (mouse_y - prevMouseY) / (float)height;
+    }
+
+    // Update mouse point
+    prevMouseX = mouse_x;
+    prevMouseY = mouse_y;
+}
+
+void Draw(const std::vector<DrawObject>& drawObjects)
+{
+  glPolygonMode(GL_FRONT, GL_FILL);
+  glPolygonMode(GL_BACK, GL_FILL);
+
+  glEnable(GL_POLYGON_OFFSET_FILL);
+  glPolygonOffset(1.0, 1.0);
+  glColor3f(1.0f, 1.0f, 1.0f);
+  for (size_t i = 0; i < drawObjects.size(); i++) {
+    DrawObject o = drawObjects[i];
+    if (o.vb < 1) {
+      continue;
+    }
+ 
+    glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glEnableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float)*3));
+    glColorPointer(3, GL_FLOAT, 36, (const void*)(sizeof(float)*6));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * o.numTriangles);
+    CheckErrors("drawarrays");
+  }
+
+  // draw wireframe
+  glDisable(GL_POLYGON_OFFSET_FILL);
+  glPolygonMode(GL_FRONT, GL_LINE);
+  glPolygonMode(GL_BACK, GL_LINE);
+
+  glColor3f(0.0f, 0.0f, 0.4f);
+  for (size_t i = 0; i < drawObjects.size(); i++) {
+    DrawObject o = drawObjects[i];
+    if (o.vb < 1) {
+      continue;
+    }
+ 
+    glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glDisableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float)*3));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * o.numTriangles);
+    CheckErrors("drawarrays");
+  }
+}
+
+static void Init() {
+  trackball(curr_quat, 0, 0, 0, 0);
+
+  eye[0] = 0.0f;
+  eye[1] = 0.0f;
+  eye[2] = 3.0f;
+
+  lookat[0] = 0.0f;
+  lookat[1] = 0.0f;
+  lookat[2] = 0.0f;
+
+  up[0] = 0.0f;
+  up[1] = 1.0f;
+  up[2] = 0.0f;
+}
+
+
+int main(int argc, char **argv)
+{
+  if (argc < 2) {
+    std::cout << "Needs input.obj\n" << std::endl;
+    return 0;
+  }
+
+  Init();
+ 
+
+  if(!glfwInit()){
+    std::cerr << "Failed to initialize GLFW." << std::endl;
+    return -1;
+  }
+
+  
+
+  window = glfwCreateWindow(width, height, "Obj viewer", NULL, NULL);
+  if(window == NULL){
+    std::cerr << "Failed to open GLFW window. " << std::endl;
+    glfwTerminate();
+    return 1;
+  }
+
+  glfwMakeContextCurrent(window);
+  glfwSwapInterval(1);
+
+  // Callback
+  glfwSetWindowSizeCallback(window, reshapeFunc);
+  glfwSetKeyCallback(window, keyboardFunc);
+  glfwSetMouseButtonCallback(window, clickFunc);
+  glfwSetCursorPosCallback(window, motionFunc);
+
+  glewExperimental = true;
+  if (glewInit() != GLEW_OK) {
+    std::cerr << "Failed to initialize GLEW." << std::endl;
+    return -1;
+  }
+
+  reshapeFunc(window, width, height);
+
+  float bmin[3], bmax[3];
+  if (false == LoadObjAndConvert(bmin, bmax, gDrawObjects, argv[1])) {
+    return -1;
+  }
+
+  float maxExtent = 0.5f * (bmax[0] - bmin[0]);
+  if (maxExtent < 0.5f * (bmax[1] - bmin[1])) {
+    maxExtent = 0.5f * (bmax[1] - bmin[1]);
+  }
+  if (maxExtent < 0.5f * (bmax[2] - bmin[2])) {
+    maxExtent = 0.5f * (bmax[2] - bmin[2]);
+  }
+
+  while(glfwWindowShouldClose(window) == GL_FALSE) {
+    glfwPollEvents();
+    glClearColor(0.1f, 0.2f, 0.3f, 1.0f);
+    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+    glEnable(GL_DEPTH_TEST);
+
+    // camera & rotate
+    glMatrixMode(GL_MODELVIEW);
+    glLoadIdentity();
+    GLfloat mat[4][4];
+    gluLookAt(eye[0], eye[1], eye[2], lookat[0], lookat[1], lookat[2], up[0], up[1], up[2]);
+    build_rotmatrix(mat, curr_quat);
+    glMultMatrixf(&mat[0][0]);
+
+    // Fit to -1, 1
+    glScalef(1.0f / maxExtent, 1.0f / maxExtent, 1.0f / maxExtent);
+
+    // Centerize object.
+    glTranslatef(-0.5*(bmax[0] + bmin[0]), -0.5*(bmax[1] + bmin[1]), -0.5*(bmax[2] + bmin[2]));
+  
+    Draw(gDrawObjects);
+
+    glfwSwapBuffers(window);
+  }
+
+  glfwTerminate();
+}

experimental/LICENSE.ltalloc

@@ -0,0 +1,26 @@
+Copyright (c) 2013, Alexander Tretyak
+Copyright (c) 2015, r-lyeh
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+    * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+    * Neither the name of the author nor the names of its
+contributors may be used to endorse or promote products derived
+from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

experimental/Makefile

@@ -0,0 +1,12 @@
+UNAME_S := $(shell uname -s)
+ifeq ($(UNAME_S),Darwin)
+LD_FLAGS=-framework OpenGL -lglfw3 -lglew
+endif
+ifeq ($(UNAME_S),Linux)
+LD_FLAGS=-lGL -lGLU -lglfw3 -lGLEW -lX11 -lXrandr -lXinerama -lXxf86vm -lXcursor -lm -pthread -ldl
+endif
+
+all:
+	clang -c trackball.c
+	clang -o viewer -g -O0 -pedantic -ansi -Wno-deprecated-declarations viewer-c.c trackball.o $(LD_FLAGS)
+#	clang -pedantic -ansi -Weverything -Wno-deprecated-declarations viewer-c.c trackball.o -framework OpenGL -lglfw3 -lglew

experimental/ltalloc.cc

@@ -0,0 +1,973 @@
+/*
+Copyright (c) 2013, Alexander Tretyak
+Copyright (c) 2015, r-lyeh
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+
+    * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+    * Neither the name of the author nor the names of its
+contributors may be used to endorse or promote products derived
+from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+  Note: The latest version of this memory allocator obtainable at
+        http://ltalloc.googlecode.com/hg/ltalloc.cc
+
+  Project URL: http://code.google.com/p/ltalloc
+*/
+
+#include "ltalloc.h"
+
+#define LTALLOC_VERSION "2.0.0" /* (2015/06/16) - ltcalloc(), ltmsize(), ltrealloc(), ltmemalign(), LTALLOC_AUTO_GC_INTERVAL 
+#define LTALLOC_VERSION "1.0.0" /* (2015/06/16) - standard STL allocator provided [see ltalloc.hpp file](ltalloc.hpp)
+#define LTALLOC_VERSION "0.0.0" /* (2013/xx/xx) - fork from public repository */
+
+//Customizable constants
+//#define LTALLOC_DISABLE_OPERATOR_NEW_OVERRIDE
+//#define LTALLOC_AUTO_GC_INTERVAL 3.0
+#ifndef LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO
+#define LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO 2//determines how accurately size classes are spaced (i.e. when = 0, allocation requests are rounded up to the nearest power of two (2^n), when = 1, rounded to 2^n, or (2^n)*1.5, when = 2, rounded to 2^n, (2^n)*1.25, (2^n)*1.5, or (2^n)*1.75, and so on); this parameter have direct influence on memory fragmentation - bigger values lead to reducing internal fragmentation (which can be approximately estimated as pow(0.5, VALUE)*100%), but at the same time increasing external fragmentation
+#endif
+#define CHUNK_SIZE     (64*1024)//size of chunk (basic allocation unit for all allocations of size <= MAX_BLOCK_SIZE); must be a power of two (as well as all following parameters), also should not be less than allocation granularity on Windows (which is always 64K by design of NT kernel)
+#define CACHE_LINE_SIZE 64
+#define MAX_NUM_OF_BLOCKS_IN_BATCH 256//maximum number of blocks to move between a thread cache and a central cache in one shot
+static const unsigned int MAX_BATCH_SIZE = 64*1024;//maximum total size of blocks to move between a thread cache and a central cache in one shot (corresponds to half size of thread cache of each size class)
+static const unsigned int MAX_BLOCK_SIZE = CHUNK_SIZE;//requesting memory of any size greater than this value will lead to direct call of system virtual memory allocation routine
+
+//Platform-specific stuff
+
+#ifdef __cplusplus
+#define CPPCODE(code) code
+#include <new>
+#else
+#define CPPCODE(code)
+#endif
+
+#ifdef LTALLOC_AUTO_GC_INTERVAL
+#include <time.h>
+#	if			LTALLOC_AUTO_GC_INTERVAL <= 0
+#		undef	LTALLOC_AUTO_GC_INTERVAL 
+#		define	LTALLOC_AUTO_GC_INTERVAL 3.00
+#	endif
+#endif
+
+#ifdef __GNUC__
+
+#define __STDC_LIMIT_MACROS
+#include <stdint.h> //for SIZE_MAX
+#include <limits.h> //for UINT_MAX
+#define alignas(a) __attribute__((aligned(a)))
+#define thread_local __thread
+#define NOINLINE __attribute__((noinline))
+#define CAS_LOCK(lock) __sync_lock_test_and_set(lock, 1)
+#define SPINLOCK_RELEASE(lock) __sync_lock_release(lock)
+#define PAUSE __asm__ __volatile__("pause" ::: "memory")
+#define BSR(r, v) r = CODE3264(__builtin_clz(v) ^ 31, __builtin_clzll(v) ^ 63)//x ^ 31 = 31 - x, but gcc does not optimize 31 - __builtin_clz(x) to bsr(x), but generates 31 - (bsr(x) ^ 31)
+
+#elif _MSC_VER
+
+#define _ALLOW_KEYWORD_MACROS
+#include <limits.h> //for SIZE_MAX and UINT_MAX
+#define alignas(a) __declspec(align(a))
+#define thread_local __declspec(thread)
+#define NOINLINE __declspec(noinline)
+#define CAS_LOCK(lock) _InterlockedExchange((long*)lock, 1)
+#define SPINLOCK_RELEASE(lock) _InterlockedExchange((long*)lock, 0)
+#define PAUSE _mm_pause()
+#define BSR(r, v) CODE3264(_BitScanReverse, _BitScanReverse64)((unsigned long*)&r, v)
+CPPCODE(extern "C") long _InterlockedExchange(long volatile *, long);
+CPPCODE(extern "C") void _mm_pause();
+#pragma warning(disable: 4127 4201 4324 4290)//"conditional expression is constant", "nonstandard extension used : nameless struct/union", and "structure was padded due to __declspec(align())"
+
+#else
+#error Unsupported compiler
+#endif
+
+#if __GNUC__ || __INTEL_COMPILER
+#define likely(x) __builtin_expect(!!(x), 1)
+#define unlikely(x) __builtin_expect(!!(x), 0)
+#else
+#define likely(x) (x)
+#define unlikely(x) (x)
+#endif
+
+static void SPINLOCK_ACQUIRE(volatile int *lock) {if (CAS_LOCK(lock)) while (*lock || CAS_LOCK(lock)) PAUSE;}
+
+#include <assert.h>
+#include <string.h> //for memset
+
+#if SIZE_MAX == UINT_MAX
+#define CODE3264(c32, c64) c32
+#else
+#define CODE3264(c32, c64) c64
+#endif
+typedef char CODE3264_check[sizeof(void*) == CODE3264(4, 8) ? 1 : -1];
+
+#ifdef _WIN32
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+
+#define VMALLOC(size) VirtualAlloc(NULL, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE)
+#define VMFREE(p, size) VirtualFree(p, 0, MEM_RELEASE)
+
+#else
+
+#include <sys/mman.h>
+#include <unistd.h>
+
+#define VMALLOC(size) (void*)(((uintptr_t)mmap(NULL, size, PROT_READ|PROT_WRITE, MAP_ANON|MAP_PRIVATE, -1, 0)+1)&~1)//with the conversion of MAP_FAILED to 0
+#define VMFREE(p, size) munmap(p, size)
+
+static size_t page_size()
+{
+	assert((uintptr_t)MAP_FAILED+1 == 0);//have to use dynamic check somewhere, because some gcc versions (e.g. 4.4.5) won't compile typedef char MAP_FAILED_value_static_check[(uintptr_t)MAP_FAILED+1 == 0 ? 1 : -1];
+	static size_t pagesize = 0;
+	if (!pagesize) pagesize = sysconf(_SC_PAGE_SIZE);//assuming that writing of size_t value is atomic, so this can be done safely in different simultaneously running threads
+	return pagesize;
+}
+
+typedef struct PTrieNode            // Compressed radix tree (patricia trie) for storing sizes of system allocations
+{                                   // This implementation have some specific properties:
+    uintptr_t keys[2];              // 1. There are no separate leaf nodes (with both null children), as leaf's value is stored directly in place of corresponding child node pointer. Least significant bit of that pointer used to determine its meaning (i.e., is it a value, or child node pointer).
+    struct PTrieNode *childNodes[2];// 2. Inserting a new element (key/value) into this tree require to create always an exactly one new node (and similarly for remove key/node operation).
+} PTrieNode;                        // 3. Tree always contains just one node with null child (i.e. all but one nodes in any possible tree are always have two children).
+#define PTRIE_NULL_NODE (PTrieNode*)(uintptr_t)1
+static PTrieNode *ptrieRoot = PTRIE_NULL_NODE, *ptrieFreeNodesList = NULL, *ptrieNewAllocatedPage = NULL;
+static volatile int ptrieLock = 0;
+
+static uintptr_t ptrie_lookup(uintptr_t key)
+{
+	PTrieNode *node = ptrieRoot;
+	uintptr_t *lastKey = NULL;
+	while (!((uintptr_t)node & 1))
+	{
+		int branch = (key >> (node->keys[0] & 0xFF)) & 1;
+		lastKey = &node->keys[branch];
+		node = node->childNodes[branch];
+	}
+	assert(lastKey && (*lastKey & ~0xFF) == key);
+	return (uintptr_t)node & ~1;
+}
+
+static void ptrie_insert(uintptr_t key, uintptr_t value, PTrieNode *newNode/* = (PTrieNode*)malloc(sizeof(PTrieNode))*/)
+{
+	PTrieNode **node = &ptrieRoot, *n;
+	uintptr_t *prevKey = NULL, x, pkey;
+	unsigned int index, b;
+	assert(!((value & 1) | (key & 0xFF)));//check constraints for key/value
+	for (;;)
+	{
+		n = *node;
+		if (!((uintptr_t)n & 1))//not a leaf
+		{
+			int prefixEnd = n->keys[0] & 0xFF;
+			x = key ^ n->keys[0];// & ~0xFF;
+			if (!(x & (~(uintptr_t)1 << prefixEnd))) {//prefix matches, so go on
+				int branch = (key >> prefixEnd) & 1;
+				node = &n->childNodes[branch];
+				prevKey = &n->keys[branch];
+			} else {//insert a new node before current
+				pkey = n->keys[0] & ~0xFF;
+				break;
+			}
+		} else {//leaf
+			if (*node == PTRIE_NULL_NODE) {
+				*node = newNode;
+				newNode->keys[0] = key;//left prefixEnd = 0, so all following insertions will be before this node
+				newNode->childNodes[0] = (PTrieNode*)(value | 1);
+				newNode->childNodes[1] = PTRIE_NULL_NODE;
+				return;
+			} else {
+				pkey = *prevKey & ~0xFF;
+				x = key ^ pkey;
+				assert(x/*key != pkey*/ && "key already inserted");
+				break;
+			}
+		}
+	}
+	BSR(index, x);
+	b = (key >> index) & 1;
+	newNode->keys[b] = key;
+	newNode->keys[b^1] = pkey;
+	newNode->keys[0] |= index;
+	newNode->childNodes[b] = (PTrieNode*)(value | 1);
+	newNode->childNodes[b^1] = n;
+	*node = newNode;
+}
+
+static uintptr_t ptrie_remove(uintptr_t key)
+{
+	PTrieNode **node = &ptrieRoot;
+	uintptr_t *pkey = NULL;
+	assert(ptrieRoot != PTRIE_NULL_NODE && "trie is empty!");
+	for (;;)
+	{
+		PTrieNode *n = *node;
+		int branch = (key >> (n->keys[0] & 0xFF)) & 1;
+		PTrieNode *cn = n->childNodes[branch];//current child node
+		if ((uintptr_t)cn & 1)//leaf
+		{
+			PTrieNode *other = n->childNodes[branch^1];
+			assert((n->keys[branch] & ~0xFF) == key);
+			assert(cn != PTRIE_NULL_NODE && "node's key is probably broken");
+		//	if (other == PTRIE_NULL_NODE) *node = PTRIE_NULL_NODE; else//special handling for null child nodes is not necessary
+			if (((uintptr_t)other & 1) && other != PTRIE_NULL_NODE)//if other node is not a pointer
+				*pkey = (n->keys[branch^1] & ~0xFF) | ((*pkey) & 0xFF);
+			*node = other;
+			*(PTrieNode**)n = ptrieFreeNodesList; ptrieFreeNodesList = n;//free(n);
+			return (uintptr_t)cn & ~1;
+		}
+		pkey = &n->keys[branch];
+		node = &n->childNodes[branch];
+	}
+}
+#endif
+
+static void *sys_aligned_alloc(size_t alignment, size_t size)
+{
+	void *p = VMALLOC(size);//optimistically try mapping precisely the right amount before falling back to the slow method
+	assert(!(alignment & (alignment-1)) && "alignment must be a power of two");
+	if ((uintptr_t)p & (alignment-1)/* && p != MAP_FAILED*/)
+	{
+		VMFREE(p, size);
+#ifdef _WIN32
+		{static DWORD allocationGranularity = 0;
+		if (!allocationGranularity) {
+			SYSTEM_INFO si;
+			GetSystemInfo(&si);
+			allocationGranularity = si.dwAllocationGranularity;
+		}
+		if ((uintptr_t)p < 16*1024*1024)//fill "bubbles" (reserve unaligned regions) at the beginning of virtual address space, otherwise there will be always falling back to the slow method
+			VirtualAlloc(p, alignment - ((uintptr_t)p & (alignment-1)), MEM_RESERVE, PAGE_NOACCESS);
+		do
+		{
+			p = VirtualAlloc(NULL, size + alignment - allocationGranularity, MEM_RESERVE, PAGE_NOACCESS);
+			if (p == NULL) return NULL;
+			VirtualFree(p, 0, MEM_RELEASE);//unfortunately, WinAPI doesn't support release a part of allocated region, so release a whole region
+			p = VirtualAlloc((void*)(((uintptr_t)p + (alignment-1)) & ~(alignment-1)), size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE);
+		} while (p == NULL);}
+#else
+		p = VMALLOC(size + alignment - page_size());
+		if (p/* != MAP_FAILED*/)
+		{
+			uintptr_t ap = ((uintptr_t)p + (alignment-1)) & ~(alignment-1);
+			uintptr_t diff = ap - (uintptr_t)p;
+			if (diff) VMFREE(p, diff);
+			diff = alignment - page_size() - diff;
+			assert((intptr_t)diff >= 0);
+			if (diff) VMFREE((void*)(ap + size), diff);
+			return (void*)ap;
+		}
+#endif
+	}
+	//if (p == 0) p = sys_aligned_alloc(alignment, size);//just in case (because 0 pointer is handled specially elsewhere)
+	//if (p == MAP_FAILED) p = NULL;
+	return p;
+}
+
+static NOINLINE void sys_free(void *p)
+{
+	if (p == NULL) return;
+#ifdef _WIN32
+	VirtualFree(p, 0, MEM_RELEASE);
+#else
+	SPINLOCK_ACQUIRE(&ptrieLock);
+	size_t size = ptrie_remove((uintptr_t)p);
+	SPINLOCK_RELEASE(&ptrieLock);
+	munmap(p, size);
+#endif
+}
+
+static void release_thread_cache(void*);
+#ifdef __GNUC__
+#include <pthread.h>
+#pragma weak pthread_once
+#pragma weak pthread_key_create
+#pragma weak pthread_setspecific
+static pthread_key_t pthread_key;
+static pthread_once_t init_once = PTHREAD_ONCE_INIT;
+static void init_pthread_key() { pthread_key_create(&pthread_key, release_thread_cache); }
+static thread_local int thread_initialized = 0;
+static void init_pthread_destructor()//must be called only when some block placed into a thread cache's free list
+{
+	if (unlikely(!thread_initialized))
+	{
+		thread_initialized = 1;
+		if (pthread_once)
+		{
+			pthread_once(&init_once, init_pthread_key);
+			pthread_setspecific(pthread_key, (void*)1);//set nonzero value to force calling of release_thread_cache() on thread terminate
+		}
+	}
+}
+#else
+static void NTAPI on_tls_callback(PVOID h, DWORD reason, PVOID pv) { h; pv; if (reason == DLL_THREAD_DETACH) release_thread_cache(0); }
+#pragma comment(linker, "/INCLUDE:" CODE3264("_","") "p_thread_callback_ltalloc")
+#pragma const_seg(".CRT$XLL")
+extern CPPCODE("C") const PIMAGE_TLS_CALLBACK p_thread_callback_ltalloc = on_tls_callback;
+#pragma const_seg()
+#define init_pthread_destructor()
+#endif
+
+//End of platform-specific stuff
+
+#define MAX_BLOCK_SIZE (MAX_BLOCK_SIZE < CHUNK_SIZE - (CHUNK_SIZE >> (1 + LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)) ? \
+                        MAX_BLOCK_SIZE : CHUNK_SIZE - (CHUNK_SIZE >> (1 + LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)))
+#define NUMBER_OF_SIZE_CLASSES ((sizeof(void*)*8 + 1) << LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)
+
+typedef struct FreeBlock
+{
+	struct FreeBlock *next,
+					 *nextBatch;//in the central cache blocks are organized into batches to allow fast moving blocks from thread cache and back
+} FreeBlock;
+
+typedef struct alignas(CACHE_LINE_SIZE) ChunkBase//force sizeof(Chunk) = cache line size to avoid false sharing
+{
+	unsigned int sizeClass;
+} Chunk;
+
+typedef struct alignas(CACHE_LINE_SIZE) ChunkSm//chunk of smallest blocks of size = sizeof(void*)
+{
+	unsigned int sizeClass;//struct ChunkBase chunk;
+	struct ChunkSm *prev, *next;
+	int numBatches;
+#define NUM_OF_BATCHES_IN_CHUNK_SM CHUNK_SIZE/(sizeof(void*)*MAX_NUM_OF_BLOCKS_IN_BATCH)
+	FreeBlock *batches[NUM_OF_BATCHES_IN_CHUNK_SM];//batches of blocks inside ChunkSm have to be stored separately (as smallest blocks of size = sizeof(void*) do not have enough space to store second pointer for the batch)
+} ChunkSm;
+
+typedef struct alignas(CACHE_LINE_SIZE)//align needed to prevent cache line sharing between adjacent classes accessed from different threads
+{
+	volatile int lock;
+	unsigned int freeBlocksInLastChunk;
+	char *lastChunk;//Chunk or ChunkSm
+	union {
+		FreeBlock *firstBatch;
+		ChunkSm *chunkWithFreeBatches;
+	};
+	FreeBlock *freeList;//short list of free blocks that for some reason are not organized into batches
+	unsigned int freeListSize;//should be less than batch size
+	uintptr_t minChunkAddr, maxChunkAddr;
+} CentralCache;
+static CentralCache centralCache[NUMBER_OF_SIZE_CLASSES];// = {{0}};
+
+typedef struct
+{
+	FreeBlock *freeList;
+	FreeBlock *tempList;//intermediate list providing a hysteresis in order to avoid a corner case of too frequent moving free blocks to the central cache and back from
+	int counter;//number of blocks in freeList (used to determine when to move free blocks list to the central cache)
+} ThreadCache;
+static thread_local ThreadCache threadCache[NUMBER_OF_SIZE_CLASSES];// = {{0}};
+
+static struct
+{
+	volatile int lock;
+	void *freeChunk;
+	size_t size;
+} pad = {0, NULL, 0};
+
+static CPPCODE(inline) unsigned int get_size_class(size_t size)
+{
+	unsigned int index;
+#if _MSC_VER && LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO == 2
+	static const unsigned char small_size_classes[256] = {//have to use a const array here, because MS compiler unfortunately does not evaluate _BitScanReverse with a constant argument at compile time (as gcc does for __builtin_clz)
+#if CODE3264(1, 0)
+		131, 4, 15, 17, 19, 20, 21, 22, 23, 24, 24, 25, 25, 26, 26, 27, 27, 28, 28, 28, 28, 29, 29, 29, 29, 30, 30, 30, 30, 31, 31, 31, 31, 32, 32, 32, 32, 32, 32, 32, 32, 33, 33, 33, 33, 33, 33, 33, 33, 34, 34, 34, 34, 34, 34, 34, 34, 35, 35, 35, 35, 35, 35, 35, 35, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 36, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 37, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43
+#else
+		131, 15, 19, 21, 23, 24, 25, 26, 27, 28, 28, 29, 29, 30, 30, 31, 31, 32, 32, 32, 32, 33, 33, 33, 33, 34, 34, 34, 34, 35, 35, 35, 35, 36, 36, 36, 36, 36, 36, 36, 36, 37, 37, 37, 37, 37, 37, 37, 37, 38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47
+#endif
+	};
+	if (size < 256 * sizeof(void*) - (sizeof(void*)-1))
+		return small_size_classes[(size + (sizeof(void*)-1)) / sizeof(void*)];
+#endif
+
+	size = (size + (sizeof(void*)-1)) & ~(sizeof(void*)-1);//minimum block size is sizeof(void*), doing this is better than just "size = max(size, sizeof(void*))"
+
+	BSR(index, (size-1)|1);//"|1" needed because result of BSR is undefined for zero input
+#if LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO == 0
+	return index;
+#else
+	return (index<<LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO) + (unsigned int)((size-1) >> (index-LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO));
+#endif
+}
+
+static unsigned int class_to_size(unsigned int c)
+{
+#if LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO == 0
+	return 2 << c;
+#else
+#if LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO >= CODE3264(2, 3)
+	if (unlikely(c < (LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO<<LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)))//for block sizes less than or equal to pow(2, LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)
+		return 2 << (c>>LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO);
+	else
+#endif
+	{
+		c -= (1<<LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)-1;
+		return ((c & ((1<<LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)-1)) | (1<<LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)) << ((c>>LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)-LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO);
+	}
+#endif
+}
+
+static unsigned int batch_size(unsigned int sizeClass)//calculates a number of blocks to move between a thread cache and a central cache in one shot
+{
+	return ((MAX_BATCH_SIZE-1) >> (sizeClass >> LTALLOC_SIZE_CLASSES_SUBPOWER_OF_TWO)) & (MAX_NUM_OF_BLOCKS_IN_BATCH-1);
+}
+
+CPPCODE(template <bool> static) void *ltmalloc(size_t size);
+
+CPPCODE(template <bool throw_>) static void *fetch_from_central_cache(size_t size, ThreadCache *tc, unsigned int sizeClass)
+{
+	void *p;
+	if (likely(size-1u <= MAX_BLOCK_SIZE-1u))//<=> if (size <= MAX_BLOCK_SIZE && size != 0)
+	{
+		FreeBlock *fb = tc->tempList;
+		if (fb)
+		{
+			assert(tc->counter == (int)batch_size(sizeClass)+1);
+			tc->counter = 1;
+			tc->freeList = fb->next;
+			tc->tempList = NULL;
+			return fb;
+		}
+
+		assert(tc->counter == 0 || tc->counter == (int)batch_size(sizeClass)+1);
+		tc->counter = 1;
+
+		{CentralCache *cc = &centralCache[sizeClass];
+		SPINLOCK_ACQUIRE(&cc->lock);
+		if (unlikely(!cc->firstBatch))//no free batch
+		{no_free_batch:{
+			unsigned int batchSize = batch_size(sizeClass)+1;
+
+			if (cc->freeList)
+			{
+				assert(cc->freeListSize);
+				if (likely(cc->freeListSize <= batchSize + 1))
+				{
+					tc->counter = batchSize - cc->freeListSize + 1;
+				//	batchSize = cc->freeListSize;
+					cc->freeListSize = 0;
+					fb = cc->freeList;
+					cc->freeList = NULL;
+				}
+				else
+				{
+					cc->freeListSize -= batchSize;
+					fb = cc->freeList;
+					{FreeBlock *b = cc->freeList;
+					while (--batchSize) b = b->next;
+					cc->freeList = b->next;
+					b->next = NULL;}
+				}
+				SPINLOCK_RELEASE(&cc->lock);
+				tc->freeList = fb->next;
+				init_pthread_destructor();//this call must be placed carefully to allow recursive memory allocation from pthread_key_create (in case when ltalloc replaces the system malloc)
+				return fb;
+			}
+
+			{unsigned int blockSize = class_to_size(sizeClass);
+			if (cc->freeBlocksInLastChunk)
+			{
+				char *firstFree = cc->lastChunk;
+				assert(cc->lastChunk && cc->freeBlocksInLastChunk == (CHUNK_SIZE - ((uintptr_t)cc->lastChunk & (CHUNK_SIZE-1)))/blockSize);
+				if (cc->freeBlocksInLastChunk < batchSize) {
+					tc->counter = batchSize - cc->freeBlocksInLastChunk + 1;
+					batchSize = cc->freeBlocksInLastChunk;
+				}
+				cc->freeBlocksInLastChunk -= batchSize;
+				cc->lastChunk += blockSize * batchSize;
+				if (cc->freeBlocksInLastChunk == 0) {
+					assert(((uintptr_t)cc->lastChunk & (CHUNK_SIZE-1)) == 0);
+					cc->lastChunk = ((char**)cc->lastChunk)[-1];
+					if (cc->lastChunk)
+						cc->freeBlocksInLastChunk = (CHUNK_SIZE - ((uintptr_t)cc->lastChunk & (CHUNK_SIZE-1)))/blockSize;
+				}
+				SPINLOCK_RELEASE(&cc->lock);
+				fb = (FreeBlock*)firstFree;
+				while (--batchSize)
+					firstFree = (char*)(((FreeBlock*)firstFree)->next = (FreeBlock*)(firstFree + blockSize));
+				((FreeBlock*)firstFree)->next = NULL;
+				tc->freeList = fb->next;
+				init_pthread_destructor();
+				return fb;
+			}
+
+			//Allocate new chunk
+			SPINLOCK_RELEASE(&cc->lock);//release lock for a while
+
+			SPINLOCK_ACQUIRE(&pad.lock);
+			if (pad.freeChunk)
+			{
+				p = pad.freeChunk;
+				pad.freeChunk = *(void**)p;
+				pad.size -= CHUNK_SIZE;
+				SPINLOCK_RELEASE(&pad.lock);
+				((char**)((char*)p + CHUNK_SIZE))[-1] = 0;
+			} else {
+				SPINLOCK_RELEASE(&pad.lock);
+				p = sys_aligned_alloc(CHUNK_SIZE, CHUNK_SIZE);
+				if (unlikely(!p)) { CPPCODE(if (throw_) throw std::bad_alloc(); else) return NULL; }
+			}
+
+#define CHUNK_IS_SMALL unlikely(sizeClass < get_size_class(2*sizeof(void*)))
+			{unsigned int numBlocksInChunk = (CHUNK_SIZE - (CHUNK_IS_SMALL ? sizeof(ChunkSm) : sizeof(Chunk)))/blockSize;
+#ifndef _WIN32
+			//intptr_t sz = ((CHUNK_SIZE - numBlocksInChunk*blockSize) & ~(page_size()-1)) - page_size();
+			//if (sz > 0) mprotect((char*)p + page_size(), sz, PROT_NONE);//munmap((char*)p + page_size(), sz);//to make possible unmapping, we need to be more careful when returning memory to the system, not simply VMFREE(firstFreeChunk, CHUNK_SIZE), so let there be just mprotect
+#endif
+			assert(((char**)((char*)p + CHUNK_SIZE))[-1] == 0);//assume that allocated memory is always zero filled (on first access); it is better not to zero it explicitly because it will lead to allocation of physical page which may never needed otherwise
+			if (numBlocksInChunk < batchSize) {
+				tc->counter = batchSize - numBlocksInChunk + 1;
+				batchSize = numBlocksInChunk;
+			}
+
+			//Prepare chunk
+			((Chunk*)p)->sizeClass = sizeClass;
+			{char *firstFree = (char*)p + CHUNK_SIZE - numBlocksInChunk*blockSize;//blocks in chunk are located in such way to achieve a maximum possible alignment
+			fb = (FreeBlock*)firstFree;
+			{int n = batchSize; while (--n)
+				firstFree = (char*)(((FreeBlock*)firstFree)->next = (FreeBlock*)(firstFree + blockSize));}
+			((FreeBlock*)firstFree)->next = NULL;
+			firstFree += blockSize;
+
+			SPINLOCK_ACQUIRE(&cc->lock);
+			if ((uintptr_t)p < cc->minChunkAddr || !cc->minChunkAddr) cc->minChunkAddr = (uintptr_t)p;
+			if ((uintptr_t)p > cc->maxChunkAddr                     ) cc->maxChunkAddr = (uintptr_t)p;
+
+			if (CHUNK_IS_SMALL)//special handling for smallest blocks of size = sizeof(void*)
+			{
+				ChunkSm *cs = (ChunkSm*)p;
+				cs->numBatches = 0;
+				//Insert new chunk right after chunkWithFreeBatches
+				cs->prev = cc->chunkWithFreeBatches;
+				if (cc->chunkWithFreeBatches) {
+					cs->next = cc->chunkWithFreeBatches->next;
+					if (cc->chunkWithFreeBatches->next) cc->chunkWithFreeBatches->next->prev = cs;
+					cc->chunkWithFreeBatches->next = cs;
+				} else {
+					cs->next = NULL;
+					cc->chunkWithFreeBatches = cs;
+				}
+			}
+
+			if (unlikely(cc->freeBlocksInLastChunk))//so happened that other thread have already allocated chunk for the same size class while the lock was released
+			{
+				//Hook pointer to the current lastChunk at the end of new chunk (another way is just put all blocks to cc->freeList which is much less effecient)
+				((char**)(((uintptr_t)firstFree & ~(CHUNK_SIZE-1)) + CHUNK_SIZE))[-1] = cc->lastChunk;
+			}
+			cc->freeBlocksInLastChunk = numBlocksInChunk - batchSize;
+			cc->lastChunk = firstFree;
+		}}}}}
+		else {
+			if (!CHUNK_IS_SMALL)//smallest blocks of size = sizeof(void*) are handled specially
+			{
+				fb = cc->firstBatch;
+				cc->firstBatch = fb->nextBatch;
+			}
+			else//size of block = sizeof(void*)
+			{
+				ChunkSm *cs = cc->chunkWithFreeBatches;
+				if (unlikely(cs->numBatches == 0))
+				{
+					if (unlikely(cs->prev == NULL)) goto no_free_batch;
+					cs = cc->chunkWithFreeBatches = cs->prev;
+					assert(cs->numBatches == NUM_OF_BATCHES_IN_CHUNK_SM);
+				}
+				fb = cs->batches[--cs->numBatches];
+			}
+		}
+		SPINLOCK_RELEASE(&cc->lock);}
+		tc->freeList = fb->next;
+		init_pthread_destructor();
+		return fb;
+	}
+	else//allocate block directly from the system
+	{
+		if (unlikely(size == 0)) return ltmalloc CPPCODE(<throw_>)(1);//return NULL;//doing this check here is better than on the top level
+
+		size = (size + CHUNK_SIZE-1) & ~(CHUNK_SIZE-1);
+		p = sys_aligned_alloc(CHUNK_SIZE, size);
+#ifndef _WIN32
+		if (p) {
+			SPINLOCK_ACQUIRE(&ptrieLock);
+			PTrieNode *newNode;
+			if (ptrieFreeNodesList)
+				ptrieFreeNodesList = *(PTrieNode**)(newNode = ptrieFreeNodesList);
+			else if (ptrieNewAllocatedPage) {
+				newNode = ptrieNewAllocatedPage;
+				if (!((uintptr_t)++ptrieNewAllocatedPage & (page_size()-1)))
+					ptrieNewAllocatedPage = ((PTrieNode**)ptrieNewAllocatedPage)[-1];
+			} else {
+				SPINLOCK_RELEASE(&ptrieLock);
+				newNode = (PTrieNode*)VMALLOC(page_size());
+				if (unlikely(!newNode)) { CPPCODE(if (throw_) throw std::bad_alloc(); else) return NULL; }
+				assert(((char**)((char*)newNode + page_size()))[-1] == 0);
+				SPINLOCK_ACQUIRE(&ptrieLock);
+				((PTrieNode**)((char*)newNode + page_size()))[-1] = ptrieNewAllocatedPage;//in case if other thread also have just allocated a new page
+				ptrieNewAllocatedPage = newNode + 1;
+			}
+			ptrie_insert((uintptr_t)p, size, newNode);
+			SPINLOCK_RELEASE(&ptrieLock);
+		}
+#endif
+		CPPCODE(if (throw_) if (unlikely(!p)) throw std::bad_alloc();)
+		return p;
+	}
+}
+
+CPPCODE(template <bool throw_> static) void *ltmalloc(size_t size)
+{
+	unsigned int sizeClass = get_size_class(size);
+	ThreadCache *tc = &threadCache[sizeClass];
+	FreeBlock *fb = tc->freeList;
+	if (likely(fb))
+	{
+		tc->freeList = fb->next;
+		tc->counter++;
+		return fb;
+	}
+	else
+		return fetch_from_central_cache CPPCODE(<throw_>)(size, tc, sizeClass);
+}
+CPPCODE(void *ltmalloc(size_t size) {return ltmalloc<false>(size);})//for possible external usage
+
+static void add_batch_to_central_cache(CentralCache *cc, unsigned int sizeClass, FreeBlock *batch)
+{
+	if (!CHUNK_IS_SMALL)
+	{
+		batch->nextBatch = cc->firstBatch;
+		cc->firstBatch = batch;
+	}
+	else
+	{
+		ChunkSm *cs = cc->chunkWithFreeBatches;
+		if (unlikely(cs->numBatches == NUM_OF_BATCHES_IN_CHUNK_SM))
+		{
+			cs = cc->chunkWithFreeBatches = cc->chunkWithFreeBatches->next;
+			assert(cs && cs->numBatches == 0);
+		}
+		cs->batches[cs->numBatches++] = batch;
+	}
+}
+
+static NOINLINE void move_to_central_cache(ThreadCache *tc, unsigned int sizeClass)
+{
+	init_pthread_destructor();//needed for cases when freed memory was allocated in the other thread and no alloc was called in this thread till its termination
+
+	tc->counter = batch_size(sizeClass);
+	if (tc->tempList)//move temp list to the central cache
+	{
+		CentralCache *cc = &centralCache[sizeClass];
+		SPINLOCK_ACQUIRE(&cc->lock);
+		add_batch_to_central_cache(cc, sizeClass, tc->tempList);
+		SPINLOCK_RELEASE(&cc->lock);
+	}
+// 	else if (unlikely(!tc->freeList))//this is a first call (i.e. when counter = 0) - just initialization of counter needed
+// 	{
+// 		tc->counter--;
+// 		return;
+// 	}
+
+	tc->tempList = tc->freeList;
+	tc->freeList = NULL;
+}
+
+void ltfree(void *p)
+{
+	if (likely((uintptr_t)p & (CHUNK_SIZE-1)))
+	{
+		unsigned int sizeClass = ((Chunk*)((uintptr_t)p & ~(CHUNK_SIZE-1)))->sizeClass;
+		ThreadCache *tc = &threadCache[sizeClass];
+
+		if (unlikely(--tc->counter < 0))
+			move_to_central_cache(tc, sizeClass);
+
+		((FreeBlock*)p)->next = tc->freeList;
+		tc->freeList = (FreeBlock*)p;
+	}
+	else
+		sys_free(p);
+}
+
+size_t ltmsize(void *p)
+{
+	if (likely((uintptr_t)p & (CHUNK_SIZE-1)))
+	{
+		return class_to_size(((Chunk*)((uintptr_t)p & ~(CHUNK_SIZE-1)))->sizeClass);
+	}
+	else
+	{
+		if (p == NULL) return 0;
+#ifdef _WIN32
+		{MEMORY_BASIC_INFORMATION mi;
+		VirtualQuery(p, &mi, sizeof(mi));
+		return mi.RegionSize;}
+#else
+		SPINLOCK_ACQUIRE(&ptrieLock);
+		size_t size = ptrie_lookup((uintptr_t)p);
+		SPINLOCK_RELEASE(&ptrieLock);
+		return size;
+#endif
+	}
+}
+
+static void release_thread_cache(void *p)
+{
+	unsigned int sizeClass = 0; (void)p;
+	for (;sizeClass < NUMBER_OF_SIZE_CLASSES; sizeClass++)
+	{
+		ThreadCache *tc = &threadCache[sizeClass];
+		if (tc->freeList || tc->tempList)
+		{
+			FreeBlock *tail = tc->freeList;
+			unsigned int freeListSize = 1;
+			CentralCache *cc = &centralCache[sizeClass];
+
+			if (tail)
+				while (tail->next)//search for end of list
+					tail = tail->next, freeListSize++;
+
+			SPINLOCK_ACQUIRE(&cc->lock);
+			if (tc->tempList)
+				add_batch_to_central_cache(cc, sizeClass, tc->tempList);
+			if (tc->freeList) {//append tc->freeList to cc->freeList
+				tail->next = cc->freeList;
+				cc->freeList = tc->freeList;
+				assert(freeListSize == batch_size(sizeClass)+1 - tc->counter);
+				cc->freeListSize += freeListSize;
+			}
+			SPINLOCK_RELEASE(&cc->lock);
+		}
+	}
+}
+
+void ltsqueeze(size_t padsz)
+{
+	unsigned int sizeClass = get_size_class(2*sizeof(void*));//skip small chunks because corresponding batches can not be efficiently detached from the central cache (if that becomes relevant, may be it worths to reimplement batches for small chunks from array to linked lists)
+	for (;sizeClass < NUMBER_OF_SIZE_CLASSES; sizeClass++)
+	{
+		CentralCache *cc = &centralCache[sizeClass];
+		if (cc->maxChunkAddr - cc->minChunkAddr <= CHUNK_SIZE)//preliminary check without lock (assume that writing to minChunkAddr/maxChunkAddr is atomic)
+			continue;
+
+		SPINLOCK_ACQUIRE(&cc->lock);
+		if (cc->maxChunkAddr - cc->minChunkAddr <= CHUNK_SIZE) {//quick check for theoretical possibility that at least one chunk is totally free
+			SPINLOCK_RELEASE(&cc->lock);
+			continue;
+		}
+		{uintptr_t minChunkAddr = cc->minChunkAddr;
+		size_t bufferSize = ((cc->maxChunkAddr - minChunkAddr) / CHUNK_SIZE + 1) * sizeof(short);
+		//Quickly detach all batches of the current size class from the central cache
+		unsigned int freeListSize = cc->freeListSize;
+		FreeBlock *firstBatch = cc->firstBatch, *freeList = cc->freeList;
+		cc->firstBatch = NULL;
+		cc->freeList = NULL;
+		cc->freeListSize = 0;
+		SPINLOCK_RELEASE(&cc->lock);
+
+		//1. Find out chunks with only free blocks via a simple counting the number of free blocks in each chunk
+		{char buffer[32*1024];//enough for 1GB address space
+		unsigned short *inChunkFreeBlocks = (unsigned short*)(bufferSize <= sizeof(buffer) ? memset(buffer, 0, bufferSize) : VMALLOC(bufferSize));
+		unsigned int numBlocksInChunk = (CHUNK_SIZE - (/*CHUNK_IS_SMALL ? sizeof(ChunkSm) : */sizeof(Chunk)))/class_to_size(sizeClass);
+		FreeBlock **pbatch, *block, **pblock;
+		Chunk *firstFreeChunk = NULL;
+		assert(numBlocksInChunk < (1U<<(sizeof(short)*8)));//in case if CHUNK_SIZE is too big that total count of blocks in it doesn't fit at short type (...may be use static_assert instead?)
+		if (inChunkFreeBlocks)//consider VMALLOC can fail
+		{
+			for (pbatch = &firstBatch; *pbatch; pbatch = &(*pbatch)->nextBatch)
+				for (block = *pbatch; block; block = block->next)
+#define FREE_BLOCK(block) \
+					if (++inChunkFreeBlocks[((uintptr_t)block - minChunkAddr) / CHUNK_SIZE] == numBlocksInChunk)/*chunk is totally free*/\
+					{\
+						Chunk *chunk = (Chunk*)((uintptr_t)block & ~(CHUNK_SIZE-1));\
+						assert(chunk->sizeClass == sizeClass);/*just in case check before overwriting this info*/\
+						*(Chunk**)chunk = firstFreeChunk;/*put nextFreeChunk pointer right at the beginning of Chunk as there are always must be a space for one pointer before first memory block*/\
+						firstFreeChunk = chunk;\
+					}
+				FREE_BLOCK(block)
+			for (pblock = &freeList; *pblock; pblock = &(*pblock)->next)
+				FREE_BLOCK(*pblock)
+#undef FREE_BLOCK
+		}
+		else {
+			for (pbatch = &firstBatch; *pbatch; pbatch = &(*pbatch)->nextBatch);
+			for (pblock = &freeList; *pblock; pblock = &(*pblock)->next);
+		}
+
+		if (firstFreeChunk)//is anything to release
+		{
+			//2. Unlink all matching blocks from the corresponding free lists
+			FreeBlock *additionalBatchesList = NULL, *additionalBlocksList = NULL, **abatch = &additionalBatchesList, **ablock = &additionalBlocksList;
+			unsigned int additionalBlocksListSize = 0, batchSize = batch_size(sizeClass)+1;
+			for (pbatch = &firstBatch; *pbatch;)
+			{
+				for (block = *pbatch; block; block = block->next)
+					if (inChunkFreeBlocks[((uintptr_t)block - minChunkAddr) / CHUNK_SIZE] == numBlocksInChunk)//if at least one block belongs to a releasable chunk, then this batch should be handled specially
+					{
+						FreeBlock *nextBatch = (*pbatch)->nextBatch;
+						for (block = *pbatch; block;)//re-add blocks of not-for-release chunks and organize them into another batches' list (to join it with the main later)
+							if (inChunkFreeBlocks[((uintptr_t)block - minChunkAddr) / CHUNK_SIZE] != numBlocksInChunk)//skip matching-for-release blocks
+							{
+								*ablock = block;
+								do//this loop needed only to minimize memory write operations, otherwise a simpler approach could be used (like in the next loop below)
+								{
+									ablock = &block->next;
+									block = block->next;
+									if (++additionalBlocksListSize == batchSize)
+									{
+										abatch = &(*abatch = additionalBlocksList)->nextBatch;
+										*abatch = NULL;
+										*ablock = NULL;
+										ablock = &additionalBlocksList;
+										additionalBlocksList = NULL;
+										additionalBlocksListSize = 0;
+										break;//to force *ablock = block; for starting a new batch
+									}
+								} while (block && inChunkFreeBlocks[((uintptr_t)block - minChunkAddr) / CHUNK_SIZE] != numBlocksInChunk);
+							}
+							else
+								block = block->next;
+						*ablock = NULL;
+						*pbatch = nextBatch;//unlink batch
+						goto continue_;
+					}
+				pbatch = &(*pbatch)->nextBatch;
+continue_:;
+			}
+			for (block = freeList; block;)
+				if (inChunkFreeBlocks[((uintptr_t)block - minChunkAddr) / CHUNK_SIZE] != numBlocksInChunk)
+				{
+					//*pblock = (*pblock)->next, freeListSize--;//unlink block
+					ablock = &(*ablock = block)->next;
+					block = block->next;
+					*ablock = NULL;
+					if (++additionalBlocksListSize == batchSize)
+					{
+						abatch = &(*abatch = additionalBlocksList)->nextBatch;
+						*abatch = NULL;
+						ablock = &additionalBlocksList;
+						additionalBlocksList = NULL;
+						additionalBlocksListSize = 0;
+					}
+				}
+				else
+					block = block->next;
+			//Add additional lists
+			*abatch = *pbatch;
+			*pbatch = additionalBatchesList;
+			pblock = ablock;
+			freeList = additionalBlocksList;
+			freeListSize = additionalBlocksListSize;
+
+			//Return back all left not-for-release blocks to the central cache as quickly as possible (as other threads may want to allocate a new memory)
+#define GIVE_LISTS_BACK_TO_CC \
+			SPINLOCK_ACQUIRE(&cc->lock);\
+			*pbatch = cc->firstBatch;\
+			cc->firstBatch = firstBatch;\
+			*pblock = cc->freeList;\
+			cc->freeList = freeList;\
+			cc->freeListSize += freeListSize;\
+			SPINLOCK_RELEASE(&cc->lock);\
+			if (bufferSize > sizeof(buffer)) VMFREE(inChunkFreeBlocks, bufferSize);//this better to do before 3. as kernel is likely optimized for release of just allocated range
+			GIVE_LISTS_BACK_TO_CC
+
+			if (padsz)
+			{
+				SPINLOCK_ACQUIRE(&pad.lock);
+				if (pad.size < padsz)
+				{
+					Chunk *first = firstFreeChunk, **c;
+					do//put off free chunks up to a specified pad size
+					{
+						c = (Chunk**)firstFreeChunk;
+						firstFreeChunk = *c;
+						pad.size += CHUNK_SIZE;
+					} while (pad.size < padsz && firstFreeChunk);
+					*c = (Chunk*)pad.freeChunk;
+					pad.freeChunk = first;
+				}
+				SPINLOCK_RELEASE(&pad.lock);
+			}
+
+			//3. Return memory to the system
+			while (firstFreeChunk)
+			{
+				Chunk *nextFreeChunk = *(Chunk**)firstFreeChunk;
+				VMFREE(firstFreeChunk, CHUNK_SIZE);
+				firstFreeChunk = nextFreeChunk;
+			}
+		}
+		else//nothing to release - just return batches back to the central cache
+		{
+			GIVE_LISTS_BACK_TO_CC
+#undef GIVE_LISTS_BACK_TO_CC
+		}}}
+	}
+}
+
+#if defined(__cplusplus) && !defined(LTALLOC_DISABLE_OPERATOR_NEW_OVERRIDE)
+void *operator new  (size_t size) throw(std::bad_alloc)          {return ltmalloc<true> (size);}
+void *operator new  (size_t size, const std::nothrow_t&) throw() {return ltmalloc<false>(size);}
+void *operator new[](size_t size) throw(std::bad_alloc)          {return ltmalloc<true> (size);}
+void *operator new[](size_t size, const std::nothrow_t&) throw() {return ltmalloc<false>(size);}
+
+void operator delete  (void* p)                        throw() {ltfree(p);}
+void operator delete  (void* p, const std::nothrow_t&) throw() {ltfree(p);}
+void operator delete[](void* p)                        throw() {ltfree(p);}
+void operator delete[](void* p, const std::nothrow_t&) throw() {ltfree(p);}
+#endif
+
+/* @r-lyeh's { */
+#include <string.h>
+void *ltcalloc(size_t elems, size_t size) {
+	size *= elems;
+	return memset( ltmalloc( size ), 0, size );
+}
+void *ltmemalign( size_t align, size_t size ) {
+	return --align, ltmalloc( (size+align)&~align );
+}
+void *ltrealloc( void *ptr, size_t sz ) {
+	if( !ptr ) return ltmalloc( sz );
+	if( !sz  ) return ltfree( ptr ), (void *)0;
+	size_t osz = ltmsize( ptr );
+	if( sz <= osz ) {
+		return ptr;
+	}
+	void *nptr = memcpy( ltmalloc(sz), ptr, osz );
+	ltfree( ptr );
+
+#ifdef LTALLOC_AUTO_GC_INTERVAL
+	/* this is kind of compromise; the following timer is to guarantee
+	that memory gets wiped out at least every given seconds between consecutive 
+	ltrealloc() calls (I am assuming frequency usage for ltrealloc() is smaller
+	than ltmalloc() or ltfree() too) - @r-lyeh */
+			clock_t now = clock();
+	static	clock_t then = now;
+	if( ( double(now - then) / CLOCKS_PER_SEC ) > LTALLOC_AUTO_GC_INTERVAL ) {
+		ltsqueeze(0);
+	}
+	then = now;
+#endif
+
+	return nptr;
+}
+/* } */

experimental/ltalloc.h

@@ -0,0 +1,14 @@
+#include <stdlib.h>  /*a more portable size_t definition than stddef.h itself*/
+#ifdef __cplusplus
+extern "C" {
+#endif
+void*  ltmalloc(size_t);
+void   ltfree(void*);
+void*  ltrealloc( void *, size_t );
+void*  ltcalloc( size_t, size_t );
+void*  ltmemalign( size_t, size_t );
+void   ltsqueeze(size_t pad); /*return memory to system (see README.md)*/
+size_t ltmsize(void*);
+#ifdef __cplusplus
+}
+#endif

experimental/ltalloc.hpp

@@ -0,0 +1,59 @@
+// based on code by Jerry Coffin (most likely Public Domain)
+// - rlyeh
+
+#pragma once
+
+#include <stdlib.h>
+#include <new>
+#include <limits>
+
+#include "ltalloc.h"
+
+namespace lt {
+template <class T>
+struct allocator {
+    typedef size_t size_type;
+    typedef ptrdiff_t difference_type;
+    typedef T* pointer;
+    typedef const T* const_pointer;
+    typedef T& reference;
+    typedef const T& const_reference;
+    typedef T value_type;
+
+    template <class U> struct rebind { typedef allocator<U> other; };
+    allocator() throw() {}
+    allocator(const allocator&) throw() {}
+
+    template <class U> allocator(const allocator<U>&) throw(){}
+
+    ~allocator() throw() {}
+
+    pointer address(reference x) const { return &x; }
+    const_pointer address(const_reference x) const { return &x; }
+
+    pointer allocate(size_type s, void const * = 0) {
+        if (0 == s)
+            return NULL;
+        pointer temp = (pointer)ltmalloc(s * sizeof(T));
+        if (temp == NULL)
+            throw std::bad_alloc();
+        return temp;
+    }
+
+    void deallocate(pointer p, size_type) {
+        ltfree(p);
+    }
+
+    size_type max_size() const throw() {
+        return std::numeric_limits<size_t>::max() / sizeof(T);
+    }
+
+    void construct(pointer p, const T& val) {
+        new((void *)p) T(val);
+    }
+
+    void destroy(pointer p) {
+        p->~T();
+    }
+};
+}

experimental/premake4.lua

@@ -0,0 +1,60 @@
+newoption {
+   trigger     = "with-zlib",
+   description = "Build with zlib."
+}
+
+solution "objview"
+	-- location ( "build" )
+	configurations { "Release", "Debug" }
+	platforms {"native", "x64", "x32"}
+	
+	project "objview"
+
+	kind "ConsoleApp"
+	language "C++"
+	files { "viewer.cc", "trackball.cc", "ltalloc.cc" }
+	includedirs { "./" }
+	includedirs { "../../" }
+
+	buildoptions { "-std=c++11" }
+
+	if _OPTIONS['with-zlib'] then
+		defines { 'ENABLE_ZLIB' }
+		links { 'z' }
+	end	
+
+	-- Uncomment if you want address sanitizer(gcc/clang only)
+	--buildoptions { "-fsanitize=address" }
+	--linkoptions { "-fsanitize=address" }
+
+	configuration { "linux" }
+		linkoptions { "`pkg-config --libs glfw3`" }
+		links { "GL", "GLU", "m", "GLEW", "X11", "Xrandr", "Xinerama", "Xi", "Xxf86vm", "Xcursor", "dl" }
+		linkoptions { "-pthread" }
+
+	configuration { "windows" }
+		-- Path to GLFW3
+		includedirs { '../../../../local/glfw-3.1.2.bin.WIN64/include' }
+		libdirs { '../../../../local/glfw-3.1.2.bin.WIN64/lib-vc2013' }
+		-- Path to GLEW
+		includedirs { '../../../../local/glew-1.13.0/include' }
+		libdirs { '../../../../local/glew-1.13.0/lib/Release/x64' }
+
+		links { "glfw3", "glew32", "gdi32", "winmm", "user32", "glu32","opengl32", "kernel32" }
+		defines { "_CRT_SECURE_NO_WARNINGS" }
+
+	configuration { "macosx" }
+		includedirs { "/usr/local/include" }
+		buildoptions { "-Wno-deprecated-declarations" }
+		libdirs { "/usr/local/lib" }
+		links { "glfw3", "GLEW" }
+		linkoptions { "-framework OpenGL", "-framework Cocoa", "-framework IOKit", "-framework CoreVideo" }
+
+	configuration "Debug"
+		defines { "DEBUG" }
+		flags { "Symbols"}
+
+	configuration "Release"
+		defines { "NDEBUG" }
+		flags { "Optimize"}
+

experimental/trackball.c

@@ -0,0 +1,292 @@
+/*
+ * (c) Copyright 1993, 1994, Silicon Graphics, Inc.
+ * ALL RIGHTS RESERVED
+ * Permission to use, copy, modify, and distribute this software for
+ * any purpose and without fee is hereby granted, provided that the above
+ * copyright notice appear in all copies and that both the copyright notice
+ * and this permission notice appear in supporting documentation, and that
+ * the name of Silicon Graphics, Inc. not be used in advertising
+ * or publicity pertaining to distribution of the software without specific,
+ * written prior permission.
+ *
+ * THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU "AS-IS"
+ * AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR OTHERWISE,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY OR
+ * FITNESS FOR A PARTICULAR PURPOSE.  IN NO EVENT SHALL SILICON
+ * GRAPHICS, INC.  BE LIABLE TO YOU OR ANYONE ELSE FOR ANY DIRECT,
+ * SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY
+ * KIND, OR ANY DAMAGES WHATSOEVER, INCLUDING WITHOUT LIMITATION,
+ * LOSS OF PROFIT, LOSS OF USE, SAVINGS OR REVENUE, OR THE CLAIMS OF
+ * THIRD PARTIES, WHETHER OR NOT SILICON GRAPHICS, INC.  HAS BEEN
+ * ADVISED OF THE POSSIBILITY OF SUCH LOSS, HOWEVER CAUSED AND ON
+ * ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
+ * POSSESSION, USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * US Government Users Restricted Rights
+ * Use, duplication, or disclosure by the Government is subject to
+ * restrictions set forth in FAR 52.227.19(c)(2) or subparagraph
+ * (c)(1)(ii) of the Rights in Technical Data and Computer Software
+ * clause at DFARS 252.227-7013 and/or in similar or successor
+ * clauses in the FAR or the DOD or NASA FAR Supplement.
+ * Unpublished-- rights reserved under the copyright laws of the
+ * United States.  Contractor/manufacturer is Silicon Graphics,
+ * Inc., 2011 N.  Shoreline Blvd., Mountain View, CA 94039-7311.
+ *
+ * OpenGL(TM) is a trademark of Silicon Graphics, Inc.
+ */
+/*
+ * Trackball code:
+ *
+ * Implementation of a virtual trackball.
+ * Implemented by Gavin Bell, lots of ideas from Thant Tessman and
+ *   the August '88 issue of Siggraph's "Computer Graphics," pp. 121-129.
+ *
+ * Vector manip code:
+ *
+ * Original code from:
+ * David M. Ciemiewicz, Mark Grossman, Henry Moreton, and Paul Haeberli
+ *
+ * Much mucking with by:
+ * Gavin Bell
+ */
+#include <math.h>
+#include "trackball.h"
+
+/*
+ * This size should really be based on the distance from the center of
+ * rotation to the point on the object underneath the mouse.  That
+ * point would then track the mouse as closely as possible.  This is a
+ * simple example, though, so that is left as an Exercise for the
+ * Programmer.
+ */
+#define TRACKBALLSIZE (0.8)
+
+/*
+ * Local function prototypes (not defined in trackball.h)
+ */
+static float tb_project_to_sphere(float, float, float);
+static void normalize_quat(float[4]);
+
+static void vzero(float *v) {
+  v[0] = 0.0;
+  v[1] = 0.0;
+  v[2] = 0.0;
+}
+
+static void vset(float *v, float x, float y, float z) {
+  v[0] = x;
+  v[1] = y;
+  v[2] = z;
+}
+
+static void vsub(const float *src1, const float *src2, float *dst) {
+  dst[0] = src1[0] - src2[0];
+  dst[1] = src1[1] - src2[1];
+  dst[2] = src1[2] - src2[2];
+}
+
+static void vcopy(const float *v1, float *v2) {
+  register int i;
+  for (i = 0; i < 3; i++)
+    v2[i] = v1[i];
+}
+
+static void vcross(const float *v1, const float *v2, float *cross) {
+  float temp[3];
+
+  temp[0] = (v1[1] * v2[2]) - (v1[2] * v2[1]);
+  temp[1] = (v1[2] * v2[0]) - (v1[0] * v2[2]);
+  temp[2] = (v1[0] * v2[1]) - (v1[1] * v2[0]);
+  vcopy(temp, cross);
+}
+
+static float vlength(const float *v) {
+  return sqrt(v[0] * v[0] + v[1] * v[1] + v[2] * v[2]);
+}
+
+static void vscale(float *v, float div) {
+  v[0] *= div;
+  v[1] *= div;
+  v[2] *= div;
+}
+
+static void vnormal(float *v) { vscale(v, 1.0 / vlength(v)); }
+
+static float vdot(const float *v1, const float *v2) {
+  return v1[0] * v2[0] + v1[1] * v2[1] + v1[2] * v2[2];
+}
+
+static void vadd(const float *src1, const float *src2, float *dst) {
+  dst[0] = src1[0] + src2[0];
+  dst[1] = src1[1] + src2[1];
+  dst[2] = src1[2] + src2[2];
+}
+
+/*
+ * Ok, simulate a track-ball.  Project the points onto the virtual
+ * trackball, then figure out the axis of rotation, which is the cross
+ * product of P1 P2 and O P1 (O is the center of the ball, 0,0,0)
+ * Note:  This is a deformed trackball-- is a trackball in the center,
+ * but is deformed into a hyperbolic sheet of rotation away from the
+ * center.  This particular function was chosen after trying out
+ * several variations.
+ *
+ * It is assumed that the arguments to this routine are in the range
+ * (-1.0 ... 1.0)
+ */
+void trackball(float q[4], float p1x, float p1y, float p2x, float p2y) {
+  float a[3]; /* Axis of rotation */
+  float phi;  /* how much to rotate about axis */
+  float p1[3], p2[3], d[3];
+  float t;
+
+  if (p1x == p2x && p1y == p2y) {
+    /* Zero rotation */
+    vzero(q);
+    q[3] = 1.0;
+    return;
+  }
+
+  /*
+   * First, figure out z-coordinates for projection of P1 and P2 to
+   * deformed sphere
+   */
+  vset(p1, p1x, p1y, tb_project_to_sphere(TRACKBALLSIZE, p1x, p1y));
+  vset(p2, p2x, p2y, tb_project_to_sphere(TRACKBALLSIZE, p2x, p2y));
+
+  /*
+   *  Now, we want the cross product of P1 and P2
+   */
+  vcross(p2, p1, a);
+
+  /*
+   *  Figure out how much to rotate around that axis.
+   */
+  vsub(p1, p2, d);
+  t = vlength(d) / (2.0 * TRACKBALLSIZE);
+
+  /*
+   * Avoid problems with out-of-control values...
+   */
+  if (t > 1.0)
+    t = 1.0;
+  if (t < -1.0)
+    t = -1.0;
+  phi = 2.0 * asin(t);
+
+  axis_to_quat(a, phi, q);
+}
+
+/*
+ *  Given an axis and angle, compute quaternion.
+ */
+void axis_to_quat(float a[3], float phi, float q[4]) {
+  vnormal(a);
+  vcopy(a, q);
+  vscale(q, sin(phi / 2.0));
+  q[3] = cos(phi / 2.0);
+}
+
+/*
+ * Project an x,y pair onto a sphere of radius r OR a hyperbolic sheet
+ * if we are away from the center of the sphere.
+ */
+static float tb_project_to_sphere(float r, float x, float y) {
+  float d, t, z;
+
+  d = sqrt(x * x + y * y);
+  if (d < r * 0.70710678118654752440) { /* Inside sphere */
+    z = sqrt(r * r - d * d);
+  } else { /* On hyperbola */
+    t = r / 1.41421356237309504880;
+    z = t * t / d;
+  }
+  return z;
+}
+
+/*
+ * Given two rotations, e1 and e2, expressed as quaternion rotations,
+ * figure out the equivalent single rotation and stuff it into dest.
+ *
+ * This routine also normalizes the result every RENORMCOUNT times it is
+ * called, to keep error from creeping in.
+ *
+ * NOTE: This routine is written so that q1 or q2 may be the same
+ * as dest (or each other).
+ */
+
+#define RENORMCOUNT 97
+
+void add_quats(float q1[4], float q2[4], float dest[4]) {
+  static int count = 0;
+  float t1[4], t2[4], t3[4];
+  float tf[4];
+
+  vcopy(q1, t1);
+  vscale(t1, q2[3]);
+
+  vcopy(q2, t2);
+  vscale(t2, q1[3]);
+
+  vcross(q2, q1, t3);
+  vadd(t1, t2, tf);
+  vadd(t3, tf, tf);
+  tf[3] = q1[3] * q2[3] - vdot(q1, q2);
+
+  dest[0] = tf[0];
+  dest[1] = tf[1];
+  dest[2] = tf[2];
+  dest[3] = tf[3];
+
+  if (++count > RENORMCOUNT) {
+    count = 0;
+    normalize_quat(dest);
+  }
+}
+
+/*
+ * Quaternions always obey:  a^2 + b^2 + c^2 + d^2 = 1.0
+ * If they don't add up to 1.0, dividing by their magnitued will
+ * renormalize them.
+ *
+ * Note: See the following for more information on quaternions:
+ *
+ * - Shoemake, K., Animating rotation with quaternion curves, Computer
+ *   Graphics 19, No 3 (Proc. SIGGRAPH'85), 245-254, 1985.
+ * - Pletinckx, D., Quaternion calculus as a basic tool in computer
+ *   graphics, The Visual Computer 5, 2-13, 1989.
+ */
+static void normalize_quat(float q[4]) {
+  int i;
+  float mag;
+
+  mag = (q[0] * q[0] + q[1] * q[1] + q[2] * q[2] + q[3] * q[3]);
+  for (i = 0; i < 4; i++)
+    q[i] /= mag;
+}
+
+/*
+ * Build a rotation matrix, given a quaternion rotation.
+ *
+ */
+void build_rotmatrix(float m[4][4], const float q[4]) {
+  m[0][0] = 1.0 - 2.0 * (q[1] * q[1] + q[2] * q[2]);
+  m[0][1] = 2.0 * (q[0] * q[1] - q[2] * q[3]);
+  m[0][2] = 2.0 * (q[2] * q[0] + q[1] * q[3]);
+  m[0][3] = 0.0;
+
+  m[1][0] = 2.0 * (q[0] * q[1] + q[2] * q[3]);
+  m[1][1] = 1.0 - 2.0 * (q[2] * q[2] + q[0] * q[0]);
+  m[1][2] = 2.0 * (q[1] * q[2] - q[0] * q[3]);
+  m[1][3] = 0.0;
+
+  m[2][0] = 2.0 * (q[2] * q[0] - q[1] * q[3]);
+  m[2][1] = 2.0 * (q[1] * q[2] + q[0] * q[3]);
+  m[2][2] = 1.0 - 2.0 * (q[1] * q[1] + q[0] * q[0]);
+  m[2][3] = 0.0;
+
+  m[3][0] = 0.0;
+  m[3][1] = 0.0;
+  m[3][2] = 0.0;
+  m[3][3] = 1.0;
+}

experimental/trackball.h

@@ -0,0 +1,75 @@
+/*
+ * (c) Copyright 1993, 1994, Silicon Graphics, Inc.
+ * ALL RIGHTS RESERVED
+ * Permission to use, copy, modify, and distribute this software for
+ * any purpose and without fee is hereby granted, provided that the above
+ * copyright notice appear in all copies and that both the copyright notice
+ * and this permission notice appear in supporting documentation, and that
+ * the name of Silicon Graphics, Inc. not be used in advertising
+ * or publicity pertaining to distribution of the software without specific,
+ * written prior permission.
+ *
+ * THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU "AS-IS"
+ * AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR OTHERWISE,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY OR
+ * FITNESS FOR A PARTICULAR PURPOSE.  IN NO EVENT SHALL SILICON
+ * GRAPHICS, INC.  BE LIABLE TO YOU OR ANYONE ELSE FOR ANY DIRECT,
+ * SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY
+ * KIND, OR ANY DAMAGES WHATSOEVER, INCLUDING WITHOUT LIMITATION,
+ * LOSS OF PROFIT, LOSS OF USE, SAVINGS OR REVENUE, OR THE CLAIMS OF
+ * THIRD PARTIES, WHETHER OR NOT SILICON GRAPHICS, INC.  HAS BEEN
+ * ADVISED OF THE POSSIBILITY OF SUCH LOSS, HOWEVER CAUSED AND ON
+ * ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
+ * POSSESSION, USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * US Government Users Restricted Rights
+ * Use, duplication, or disclosure by the Government is subject to
+ * restrictions set forth in FAR 52.227.19(c)(2) or subparagraph
+ * (c)(1)(ii) of the Rights in Technical Data and Computer Software
+ * clause at DFARS 252.227-7013 and/or in similar or successor
+ * clauses in the FAR or the DOD or NASA FAR Supplement.
+ * Unpublished-- rights reserved under the copyright laws of the
+ * United States.  Contractor/manufacturer is Silicon Graphics,
+ * Inc., 2011 N.  Shoreline Blvd., Mountain View, CA 94039-7311.
+ *
+ * OpenGL(TM) is a trademark of Silicon Graphics, Inc.
+ */
+/*
+ * trackball.h
+ * A virtual trackball implementation
+ * Written by Gavin Bell for Silicon Graphics, November 1988.
+ */
+
+/*
+ * Pass the x and y coordinates of the last and current positions of
+ * the mouse, scaled so they are from (-1.0 ... 1.0).
+ *
+ * The resulting rotation is returned as a quaternion rotation in the
+ * first paramater.
+ */
+void trackball(float q[4], float p1x, float p1y, float p2x, float p2y);
+
+void negate_quat(float *q, float *qn);
+
+/*
+ * Given two quaternions, add them together to get a third quaternion.
+ * Adding quaternions to get a compound rotation is analagous to adding
+ * translations to get a compound translation.  When incrementally
+ * adding rotations, the first argument here should be the new
+ * rotation, the second and third the total rotation (which will be
+ * over-written with the resulting new total rotation).
+ */
+void add_quats(float *q1, float *q2, float *dest);
+
+/*
+ * A useful function, builds a rotation matrix in Matrix based on
+ * given quaternion.
+ */
+void build_rotmatrix(float m[4][4], const float q[4]);
+
+/*
+ * This function computes a quaternion based on an axis (defined by
+ * the given vector) and an angle about which to rotate.  The angle is
+ * expressed in radians.  The result is put into the third argument.
+ */
+void axis_to_quat(float a[3], float phi, float q[4]);

experimental/viewer-c.c

@@ -0,0 +1,622 @@
+#include <GL/glew.h>
+#include <float.h>
+#include <limits.h>
+#include <math.h>
+
+#ifdef __APPLE__
+#include <OpenGL/glu.h>
+#else
+#include <GL/glu.h>
+#endif
+
+#include <GLFW/glfw3.h>
+
+#include "trackball.h"
+
+#define TINYOBJ_LOADER_C_IMPLEMENTATION
+#include "tinyobj_loader_c.h"
+
+#define MAX_OBJECTS (128)
+
+typedef struct {
+  GLuint vb;
+  int numTriangles;
+} DrawObject;
+
+static DrawObject gDrawObject;
+
+static int width = 768;
+static int height = 768;
+
+static float prevMouseX, prevMouseY;
+static int mouseLeftPressed;
+static int mouseMiddlePressed;
+static int mouseRightPressed;
+static float curr_quat[4];
+static float prev_quat[4];
+static float eye[3], lookat[3], up[3];
+
+static GLFWwindow* gWindow;
+
+static void CheckErrors(const char* desc) {
+  GLenum e = glGetError();
+  if (e != GL_NO_ERROR) {
+    fprintf(stderr, "OpenGL error in \"%s\": %d (%d)\n", desc, e, e);
+    exit(20);
+  }
+}
+
+static void CalcNormal(float N[3], float v0[3], float v1[3], float v2[3]) {
+  float v10[3];
+  float v20[3];
+  float len2;
+
+  v10[0] = v1[0] - v0[0];
+  v10[1] = v1[1] - v0[1];
+  v10[2] = v1[2] - v0[2];
+
+  v20[0] = v2[0] - v0[0];
+  v20[1] = v2[1] - v0[1];
+  v20[2] = v2[2] - v0[2];
+
+  N[0] = v20[1] * v10[2] - v20[2] * v10[1];
+  N[1] = v20[2] * v10[0] - v20[0] * v10[2];
+  N[2] = v20[0] * v10[1] - v20[1] * v10[0];
+
+  len2 = N[0] * N[0] + N[1] * N[1] + N[2] * N[2];
+  if (len2 > 0.0f) {
+    float len = sqrt(len2);
+
+    N[0] /= len;
+    N[1] /= len;
+  }
+}
+
+static const char* mmap_file(size_t* len, const char* filename) {
+#ifdef _WIN64
+  HANDLE file =
+      CreateFileA(filename, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING,
+                  FILE_ATTRIBUTE_NORMAL | FILE_FLAG_SEQUENTIAL_SCAN, NULL);
+  assert(file != INVALID_HANDLE_VALUE);
+
+  HANDLE fileMapping = CreateFileMapping(file, NULL, PAGE_READONLY, 0, 0, NULL);
+  assert(fileMapping != INVALID_HANDLE_VALUE);
+
+  LPVOID fileMapView = MapViewOfFile(fileMapping, FILE_MAP_READ, 0, 0, 0);
+  auto fileMapViewChar = (const char*)fileMapView;
+  assert(fileMapView != NULL);
+#else
+
+  FILE* f;
+  long file_size;
+  struct stat sb;
+  char* p;
+  int fd;
+
+  (*len) = 0;
+
+  f = fopen(filename, "r");
+  fseek(f, 0, SEEK_END);
+  file_size = ftell(f);
+  fclose(f);
+
+  fd = open(filename, O_RDONLY);
+  if (fd == -1) {
+    perror("open");
+    return NULL;
+  }
+
+  if (fstat(fd, &sb) == -1) {
+    perror("fstat");
+    return NULL;
+  }
+
+  if (!S_ISREG(sb.st_mode)) {
+    fprintf(stderr, "%s is not a file\n", "lineitem.tbl");
+    return NULL;
+  }
+
+  p = (char*)mmap(0, (size_t)file_size, PROT_READ, MAP_SHARED, fd, 0);
+
+  if (p == MAP_FAILED) {
+    perror("mmap");
+    return NULL;
+  }
+
+  if (close(fd) == -1) {
+    perror("close");
+    return NULL;
+  }
+
+  (*len) = (size_t)file_size;
+
+  return p;
+
+#endif
+}
+
+#if 0
+static int gz_load(std::vector<char>* buf, const char* filename)
+{
+#ifdef ENABLE_ZLIB
+    gzFile file;
+    file = gzopen (filename, "r");
+    if (! file) {
+        fprintf (stderr, "gzopen of '%s' failed: %s.\n", filename,
+                 strerror (errno));
+        exit (EXIT_FAILURE);
+        return false;
+    }
+    while (1) {
+        int err;                    
+        int bytes_read;
+        unsigned char buffer[1024];
+        bytes_read = gzread (file, buffer, 1024);
+        buf->insert(buf->end(), buffer, buffer + 1024); 
+        //printf ("%s", buffer);
+        if (bytes_read < 1024) {
+            if (gzeof (file)) {
+                break;
+            }
+            else {
+                const char * error_string;
+                error_string = gzerror (file, & err);
+                if (err) {
+                    fprintf (stderr, "Error: %s.\n", error_string);
+                    exit (EXIT_FAILURE);
+                    return false;
+                }
+            }
+        }
+    }
+    gzclose (file);
+    return true;
+#else
+  return false;
+#endif
+}
+#endif
+
+static const char* get_file_data(size_t* len, const char* filename) {
+  const char* ext = strrchr(filename, '.');
+
+  size_t data_len = 0;
+  const char* data = NULL;
+
+  if (strcmp(ext, ".gz") == 0) {
+    assert(0); /* todo */
+
+#if 0
+    std::vector<char> buf;
+    bool ret = gz_load(&buf, filename);
+
+    if (ret) {
+      char *p = static_cast<char*>(malloc(buf.size() + 1));  // @fixme { implement deleter }
+      memcpy(p, &buf.at(0), buf.size());
+      p[buf.size()] = '\0';
+      data = p;
+      data_len = buf.size();
+    }
+#endif
+
+  } else {
+    data = mmap_file(&data_len, filename);
+  }
+
+  (*len) = data_len;
+  return data;
+}
+
+static int LoadObjAndConvert(float bmin[3], float bmax[3],
+                             const char* filename) {
+  tinyobj_attrib_t attrib;
+  tinyobj_shape_t* shapes = NULL;
+  size_t num_shapes;
+  tinyobj_material_t* materials = NULL;
+  size_t num_materials;
+
+  size_t data_len = 0;
+  const char* data = get_file_data(&data_len, filename);
+  if (data == NULL) {
+    exit(-1);
+    return 0;
+  }
+  printf("filesize: %d\n", (int)data_len);
+
+  {
+    unsigned int flags = TINYOBJ_FLAG_TRIANGULATE;
+    int ret = tinyobj_parse_obj(&attrib, &shapes, &num_shapes, &materials,
+                                &num_materials, data, data_len, flags);
+    if (ret != TINYOBJ_SUCCESS) {
+      return 0;
+    }
+
+    printf("# of shapes    = %d\n", (int)num_shapes);
+    printf("# of materiasl = %d\n", (int)num_materials);
+
+    {
+      int i;
+      for (i = 0; i < num_shapes; i++) {
+        printf("shape[%d] name = %s\n", i, shapes[i].name);
+      }
+    }
+  }
+
+  bmin[0] = bmin[1] = bmin[2] = FLT_MAX;
+  bmax[0] = bmax[1] = bmax[2] = -FLT_MAX;
+
+  {
+    DrawObject o;
+    float* vb;
+    /* std::vector<float> vb; //  */
+    size_t face_offset = 0;
+    size_t i;
+
+    /* Assume triangulated face. */
+    size_t num_triangles = attrib.num_face_num_verts;
+    size_t stride = 9; /* 9 = pos(3float), normal(3float), color(3float) */
+
+    vb = (float*)malloc(sizeof(float) * stride * num_triangles * 3);
+
+    for (i = 0; i < attrib.num_face_num_verts; i++) {
+      size_t f;
+      assert(attrib.face_num_verts[i] % 3 ==
+             0); /* assume all triangle faces. */
+      for (f = 0; f < attrib.face_num_verts[i] / 3; f++) {
+        int k;
+        float v[3][3];
+        float n[3][3];
+        float c[3];
+        float len2;
+
+        tinyobj_vertex_index_t idx0 = attrib.faces[face_offset + 3 * f + 0];
+        tinyobj_vertex_index_t idx1 = attrib.faces[face_offset + 3 * f + 1];
+        tinyobj_vertex_index_t idx2 = attrib.faces[face_offset + 3 * f + 2];
+
+        for (k = 0; k < 3; k++) {
+          int f0 = idx0.v_idx;
+          int f1 = idx1.v_idx;
+          int f2 = idx2.v_idx;
+          assert(f0 >= 0);
+          assert(f1 >= 0);
+          assert(f2 >= 0);
+
+          v[0][k] = attrib.vertices[3 * f0 + k];
+          v[1][k] = attrib.vertices[3 * f1 + k];
+          v[2][k] = attrib.vertices[3 * f2 + k];
+          bmin[k] = (v[0][k] < bmin[k]) ? v[0][k] : bmin[k];
+          bmin[k] = (v[1][k] < bmin[k]) ? v[1][k] : bmin[k];
+          bmin[k] = (v[2][k] < bmin[k]) ? v[2][k] : bmin[k];
+          bmax[k] = (v[0][k] > bmax[k]) ? v[0][k] : bmax[k];
+          bmax[k] = (v[1][k] > bmax[k]) ? v[1][k] : bmax[k];
+          bmax[k] = (v[2][k] > bmax[k]) ? v[2][k] : bmax[k];
+        }
+
+        if (attrib.num_normals > 0) {
+          int f0 = idx0.vn_idx;
+          int f1 = idx1.vn_idx;
+          int f2 = idx2.vn_idx;
+          if (f0 >= 0 && f1 >= 0 && f2 >= 0) {
+            assert(3 * f0 + 2 < attrib.num_normals);
+            assert(3 * f1 + 2 < attrib.num_normals);
+            assert(3 * f2 + 2 < attrib.num_normals);
+            for (k = 0; k < 3; k++) {
+              n[0][k] = attrib.normals[3 * f0 + k];
+              n[1][k] = attrib.normals[3 * f1 + k];
+              n[2][k] = attrib.normals[3 * f2 + k];
+            }
+          } else { /* normal index is not defined for this face */
+            /* compute geometric normal */
+            CalcNormal(n[0], v[0], v[1], v[2]);
+            n[1][0] = n[0][0];
+            n[1][1] = n[0][1];
+            n[1][2] = n[0][2];
+            n[2][0] = n[0][0];
+            n[2][1] = n[0][1];
+            n[2][2] = n[0][2];
+          }
+        } else {
+          /* compute geometric normal */
+          CalcNormal(n[0], v[0], v[1], v[2]);
+          n[1][0] = n[0][0];
+          n[1][1] = n[0][1];
+          n[1][2] = n[0][2];
+          n[2][0] = n[0][0];
+          n[2][1] = n[0][1];
+          n[2][2] = n[0][2];
+        }
+
+        for (k = 0; k < 3; k++) {
+          vb[(3 * i + k) * stride + 0] = v[k][0];
+          vb[(3 * i + k) * stride + 1] = v[k][1];
+          vb[(3 * i + k) * stride + 2] = v[k][2];
+          vb[(3 * i + k) * stride + 3] = n[k][0];
+          vb[(3 * i + k) * stride + 4] = n[k][1];
+          vb[(3 * i + k) * stride + 5] = n[k][2];
+
+          /* Use normal as color. */
+          c[0] = n[k][0];
+          c[1] = n[k][1];
+          c[2] = n[k][2];
+          len2 = c[0] * c[0] + c[1] * c[1] + c[2] * c[2];
+          if (len2 > 0.0f) {
+            float len = (float)sqrt(len2);
+
+            c[0] /= len;
+            c[1] /= len;
+            c[2] /= len;
+          }
+
+          vb[(3 * i + k) * stride + 6] = (c[0] * 0.5 + 0.5);
+          vb[(3 * i + k) * stride + 7] = (c[1] * 0.5 + 0.5);
+          vb[(3 * i + k) * stride + 8] = (c[2] * 0.5 + 0.5);
+        }
+      }
+      face_offset += attrib.face_num_verts[i];
+    }
+
+    o.vb = 0;
+    o.numTriangles = 0;
+    if (num_triangles > 0) {
+      glGenBuffers(1, &o.vb);
+      glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+      glBufferData(GL_ARRAY_BUFFER, num_triangles * 3 * stride * sizeof(float),
+                   vb, GL_STATIC_DRAW);
+      o.numTriangles = num_triangles;
+    }
+
+    free(vb);
+
+    gDrawObject = o;
+  }
+
+  printf("bmin = %f, %f, %f\n", bmin[0], bmin[1], bmin[2]);
+  printf("bmax = %f, %f, %f\n", bmax[0], bmax[1], bmax[2]);
+
+  tinyobj_attrib_free(&attrib);
+  tinyobj_shapes_free(shapes, num_shapes);
+  tinyobj_materials_free(materials, num_materials);
+
+  return 1;
+}
+
+static void reshapeFunc(GLFWwindow* window, int w, int h) {
+  int fb_w, fb_h;
+  glfwGetFramebufferSize(window, &fb_w, &fb_h);
+
+  glViewport(0, 0, fb_w, fb_h);
+  glMatrixMode(GL_PROJECTION);
+  glLoadIdentity();
+  gluPerspective(45.0, (GLdouble)w / (GLdouble)h, (GLdouble)0.01f,
+                 (GLdouble)100.0f);
+  glMatrixMode(GL_MODELVIEW);
+  glLoadIdentity();
+
+  width = w;
+  height = h;
+}
+
+static void keyboardFunc(GLFWwindow* window, int key, int scancode, int action,
+                         int mods) {
+  (void)window;
+  (void)scancode;
+  (void)mods;
+  if (action == GLFW_PRESS || action == GLFW_REPEAT) {
+    /* Move camera */
+    float mv_x = 0, mv_y = 0, mv_z = 0;
+    if (key == GLFW_KEY_K)
+      mv_x += 1;
+    else if (key == GLFW_KEY_J)
+      mv_x += -1;
+    else if (key == GLFW_KEY_L)
+      mv_y += 1;
+    else if (key == GLFW_KEY_H)
+      mv_y += -1;
+    else if (key == GLFW_KEY_P)
+      mv_z += 1;
+    else if (key == GLFW_KEY_N)
+      mv_z += -1;
+
+    if (key == GLFW_KEY_Q || key == GLFW_KEY_ESCAPE)
+      glfwSetWindowShouldClose(window, GL_TRUE);
+  }
+}
+
+static void clickFunc(GLFWwindow* window, int button, int action, int mods) {
+  (void)window;
+  (void)mods;
+  if (button == GLFW_MOUSE_BUTTON_LEFT) {
+    if (action == GLFW_PRESS) {
+      mouseLeftPressed = 1;
+      trackball(prev_quat, 0.0, 0.0, 0.0, 0.0);
+    } else if (action == GLFW_RELEASE) {
+      mouseLeftPressed = 0;
+    }
+  }
+  if (button == GLFW_MOUSE_BUTTON_RIGHT) {
+    if (action == GLFW_PRESS) {
+      mouseRightPressed = 1;
+    } else if (action == GLFW_RELEASE) {
+      mouseRightPressed = 0;
+    }
+  }
+  if (button == GLFW_MOUSE_BUTTON_MIDDLE) {
+    if (action == GLFW_PRESS) {
+      mouseMiddlePressed = 1;
+    } else if (action == GLFW_RELEASE) {
+      mouseMiddlePressed = 0;
+    }
+  }
+}
+
+static void motionFunc(GLFWwindow* window, double mouse_x, double mouse_y) {
+  float rotScale = 1.0f;
+  float transScale = 2.0f;
+
+  (void)window;
+
+  if (mouseLeftPressed) {
+    trackball(prev_quat, rotScale * (2.0f * prevMouseX - width) / (float)width,
+              rotScale * (height - 2.0f * prevMouseY) / (float)height,
+              rotScale * (2.0f * (float)mouse_x - width) / (float)width,
+              rotScale * (height - 2.0f * (float)mouse_y) / (float)height);
+
+    add_quats(prev_quat, curr_quat, curr_quat);
+  } else if (mouseMiddlePressed) {
+    eye[0] -= transScale * ((float)mouse_x - prevMouseX) / (float)width;
+    lookat[0] -= transScale * ((float)mouse_x - prevMouseX) / (float)width;
+    eye[1] += transScale * ((float)mouse_y - prevMouseY) / (float)height;
+    lookat[1] += transScale * ((float)mouse_y - prevMouseY) / (float)height;
+  } else if (mouseRightPressed) {
+    eye[2] += transScale * ((float)mouse_y - prevMouseY) / (float)height;
+    lookat[2] += transScale * ((float)mouse_y - prevMouseY) / (float)height;
+  }
+
+  prevMouseX = (float)mouse_x;
+  prevMouseY = (float)mouse_y;
+}
+
+static void Draw(const DrawObject* draw_object) {
+  int i;
+
+  glPolygonMode(GL_FRONT, GL_FILL);
+  glPolygonMode(GL_BACK, GL_FILL);
+
+  glEnable(GL_POLYGON_OFFSET_FILL);
+  glPolygonOffset(1.0, 1.0);
+  glColor3f(1.0f, 1.0f, 1.0f);
+  if (draw_object->vb >= 1) {
+    glBindBuffer(GL_ARRAY_BUFFER, draw_object->vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glEnableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float) * 3));
+    glColorPointer(3, GL_FLOAT, 36, (const void*)(sizeof(float) * 6));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * draw_object->numTriangles);
+    CheckErrors("drawarrays");
+  }
+
+  /* draw wireframe */
+  glDisable(GL_POLYGON_OFFSET_FILL);
+  glPolygonMode(GL_FRONT, GL_LINE);
+  glPolygonMode(GL_BACK, GL_LINE);
+
+  glColor3f(0.0f, 0.0f, 0.4f);
+
+  if (draw_object->vb >= 1) {
+    glBindBuffer(GL_ARRAY_BUFFER, draw_object->vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glDisableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float) * 3));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * draw_object->numTriangles);
+    CheckErrors("drawarrays");
+  }
+}
+
+static void Init() {
+  trackball(curr_quat, 0, 0, 0, 0);
+
+  eye[0] = 0.0f;
+  eye[1] = 0.0f;
+  eye[2] = 3.0f;
+
+  lookat[0] = 0.0f;
+  lookat[1] = 0.0f;
+  lookat[2] = 0.0f;
+
+  up[0] = 0.0f;
+  up[1] = 1.0f;
+  up[2] = 0.0f;
+}
+
+int main(int argc, char** argv) {
+  if (argc < 2) {
+    fprintf(stderr, "Needs input.obj\n");
+    return 0;
+  }
+
+  Init();
+
+  printf("Initialize GLFW...\n");
+
+  if (!glfwInit()) {
+    fprintf(stderr, "Failed to initialize GLFW.\n");
+    return -1;
+  }
+
+  gWindow = glfwCreateWindow(width, height, "Obj viewer", NULL, NULL);
+  if (gWindow == NULL) {
+    fprintf(stderr, "Failed to open GLFW window.\n");
+    glfwTerminate();
+    return 1;
+  }
+
+  glfwMakeContextCurrent(gWindow);
+  glfwSwapInterval(1);
+
+  glfwSetWindowSizeCallback(gWindow, reshapeFunc);
+  glfwSetKeyCallback(gWindow, keyboardFunc);
+  glfwSetMouseButtonCallback(gWindow, clickFunc);
+  glfwSetCursorPosCallback(gWindow, motionFunc);
+
+  /* glewExperimental = 1; */
+  if (glewInit() != GLEW_OK) {
+    fprintf(stderr, "Failed to initialize GLEW.\n");
+    return -1;
+  }
+
+  reshapeFunc(gWindow, width, height);
+
+  {
+    float bmin[3], bmax[3];
+    float maxExtent;
+    if (0 == LoadObjAndConvert(bmin, bmax, argv[1])) {
+      printf("failed to load & conv\n");
+      return -1;
+    }
+
+    maxExtent = 0.5f * (bmax[0] - bmin[0]);
+    if (maxExtent < 0.5f * (bmax[1] - bmin[1])) {
+      maxExtent = 0.5f * (bmax[1] - bmin[1]);
+    }
+    if (maxExtent < 0.5f * (bmax[2] - bmin[2])) {
+      maxExtent = 0.5f * (bmax[2] - bmin[2]);
+    }
+
+    while (glfwWindowShouldClose(gWindow) == GL_FALSE) {
+      GLfloat mat[4][4];
+
+      glfwPollEvents();
+      glClearColor(0.1f, 0.2f, 0.3f, 1.0f);
+      glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+      glEnable(GL_DEPTH_TEST);
+
+      glMatrixMode(GL_MODELVIEW);
+      glLoadIdentity();
+      gluLookAt((GLdouble)eye[0], (GLdouble)eye[1], (GLdouble)eye[2],
+                (GLdouble)lookat[0], (GLdouble)lookat[1], (GLdouble)lookat[2],
+                (GLdouble)up[0], (GLdouble)up[1], (GLdouble)up[2]);
+      build_rotmatrix(mat, curr_quat);
+      glMultMatrixf(&mat[0][0]);
+
+      /* Fit to -1, 1 */
+      glScalef(1.0f / maxExtent, 1.0f / maxExtent, 1.0f / maxExtent);
+
+      /* Centerize object. */
+      glTranslatef(-0.5f * (bmax[0] + bmin[0]), -0.5f * (bmax[1] + bmin[1]),
+                   -0.5f * (bmax[2] + bmin[2]));
+
+      Draw(&gDrawObject);
+
+      glfwSwapBuffers(gWindow);
+    }
+  }
+
+  glfwTerminate();
+}

experimental/viewer.cc

@@ -0,0 +1,612 @@
+//
+// Simple .obj viewer(vertex only)
+//
+#include <vector>
+#include <string>
+#include <cstdio>
+#include <cstdlib>
+#include <iostream>
+#include <limits>
+#include <cmath>
+#include <cassert>
+#include <cstring>	
+#include <algorithm>	
+
+#if defined(ENABLE_ZLIB)
+#include <zlib.h>
+#endif
+
+#include <GL/glew.h>
+
+#ifdef __APPLE__
+#include <OpenGL/glu.h>
+#else
+#include <GL/glu.h>
+#endif
+
+#include <GLFW/glfw3.h>
+
+#include "trackball.h"
+
+#define TINYOBJ_LOADER_OPT_IMPLEMENTATION
+#include "tinyobj_loader_opt.h"
+
+typedef struct {
+  GLuint vb;    // vertex buffer
+  int numTriangles;
+} DrawObject;
+
+std::vector<DrawObject> gDrawObjects;
+
+int width = 768;
+int height = 768;
+
+double prevMouseX, prevMouseY;
+bool mouseLeftPressed;
+bool mouseMiddlePressed;
+bool mouseRightPressed;
+float curr_quat[4];
+float prev_quat[4];
+float eye[3], lookat[3], up[3];
+
+GLFWwindow* window;
+
+void CheckErrors(std::string desc) {
+  GLenum e = glGetError();
+  if (e != GL_NO_ERROR) {
+    fprintf(stderr, "OpenGL error in \"%s\": %d (%d)\n", desc.c_str(), e, e);
+    exit(20);
+  }
+}
+
+void CalcNormal(float N[3], float v0[3], float v1[3], float v2[3]) {
+  float v10[3];
+  v10[0] = v1[0] - v0[0];
+  v10[1] = v1[1] - v0[1];
+  v10[2] = v1[2] - v0[2];
+
+  float v20[3];
+  v20[0] = v2[0] - v0[0];
+  v20[1] = v2[1] - v0[1];
+  v20[2] = v2[2] - v0[2];
+
+  N[0] = v20[1] * v10[2] - v20[2] * v10[1];
+  N[1] = v20[2] * v10[0] - v20[0] * v10[2];
+  N[2] = v20[0] * v10[1] - v20[1] * v10[0];
+
+  float len2 = N[0] * N[0] + N[1] * N[1] + N[2] * N[2];
+  if (len2 > 0.0f) {
+    float len = sqrtf(len2);
+    
+    N[0] /= len;
+    N[1] /= len;
+  }
+}
+
+const char *mmap_file(size_t *len, const char* filename)
+{
+  (*len) = 0;
+#ifdef _WIN64
+  HANDLE file = CreateFileA(filename, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL | FILE_FLAG_SEQUENTIAL_SCAN, NULL);
+    assert(file != INVALID_HANDLE_VALUE);
+
+    HANDLE fileMapping = CreateFileMapping(file, NULL, PAGE_READONLY, 0, 0, NULL);
+    assert(fileMapping != INVALID_HANDLE_VALUE);
+ 
+    LPVOID fileMapView = MapViewOfFile(fileMapping, FILE_MAP_READ, 0, 0, 0);
+    auto fileMapViewChar = (const char*)fileMapView;
+    assert(fileMapView != NULL);
+#else
+
+  FILE* f = fopen(filename, "r" );
+  fseek(f, 0, SEEK_END);
+  long fileSize = ftell(f);
+  fclose(f);
+
+  struct stat sb;
+  char *p;
+  int fd;
+
+  fd = open (filename, O_RDONLY);
+  if (fd == -1) {
+    perror ("open");
+    return NULL;
+  }
+
+  if (fstat (fd, &sb) == -1) {
+    perror ("fstat");
+    return NULL;
+  }
+
+  if (!S_ISREG (sb.st_mode)) {
+    fprintf (stderr, "%s is not a file\n", "lineitem.tbl");
+    return NULL;
+  }
+
+  p = (char*)mmap (0, fileSize, PROT_READ, MAP_SHARED, fd, 0);
+
+  if (p == MAP_FAILED) {
+    perror ("mmap");
+    return NULL;
+  }
+
+  if (close (fd) == -1) {
+    perror ("close");
+    return NULL;
+  }
+
+  (*len) = fileSize;
+
+  return p;
+  
+#endif
+}
+
+bool gz_load(std::vector<char>* buf, const char* filename)
+{
+#ifdef ENABLE_ZLIB
+    gzFile file;
+    file = gzopen (filename, "r");
+    if (! file) {
+        fprintf (stderr, "gzopen of '%s' failed: %s.\n", filename,
+                 strerror (errno));
+        exit (EXIT_FAILURE);
+        return false;
+    }
+    while (1) {
+        int err;                    
+        int bytes_read;
+        unsigned char buffer[1024];
+        bytes_read = gzread (file, buffer, 1024);
+        buf->insert(buf->end(), buffer, buffer + 1024); 
+        //printf ("%s", buffer);
+        if (bytes_read < 1024) {
+            if (gzeof (file)) {
+                break;
+            }
+            else {
+                const char * error_string;
+                error_string = gzerror (file, & err);
+                if (err) {
+                    fprintf (stderr, "Error: %s.\n", error_string);
+                    exit (EXIT_FAILURE);
+                    return false;
+                }
+            }
+        }
+    }
+    gzclose (file);
+    return true;
+#else
+  return false;
+#endif
+}
+
+const char* get_file_data(size_t *len, const char* filename)
+{
+
+  const char *ext = strrchr(filename, '.');
+
+  size_t data_len = 0;
+  const char* data = nullptr;
+
+  if (strcmp(ext, ".gz") == 0) {
+    // gzipped data.
+
+    std::vector<char> buf;
+    bool ret = gz_load(&buf, filename);
+
+    if (ret) {
+      char *p = static_cast<char*>(malloc(buf.size() + 1));  // @fixme { implement deleter }
+      memcpy(p, &buf.at(0), buf.size());
+      p[buf.size()] = '\0';
+      data = p;
+      data_len = buf.size();
+    }
+
+  } else {
+    
+    data = mmap_file(&data_len, filename);
+  }
+
+  (*len) = data_len;
+  return data;
+}
+
+
+bool LoadObjAndConvert(float bmin[3], float bmax[3], const char* filename, int num_threads)
+{
+#if 1
+  tinyobj_opt::attrib_t attrib;
+  std::vector<tinyobj_opt::shape_t> shapes;
+
+  size_t data_len = 0;
+  const char* data = get_file_data(&data_len, filename);
+  if (data == nullptr) {
+    exit(-1);
+    return false;
+  }
+  printf("filesize: %d\n", (int)data_len);
+  tinyobj_opt::LoadOption option;
+  option.req_num_threads = num_threads;
+  bool ret = parseObj(&attrib, &shapes, data, data_len, option);
+
+  bmin[0] = bmin[1] = bmin[2] = std::numeric_limits<float>::max();
+  bmax[0] = bmax[1] = bmax[2] = -std::numeric_limits<float>::max();
+
+  {
+        DrawObject o;
+        std::vector<float> vb; // pos(3float), normal(3float), color(3float)
+        size_t face_offset = 0;
+        for (size_t v = 0; v < attrib.face_num_verts.size(); v++) {
+          assert(attrib.face_num_verts[v] % 3 == 0); // assume all triangle face.
+          for (size_t f = 0; f < attrib.face_num_verts[v] / 3; f++) {
+            tinyobj_opt::vertex_index idx0 = attrib.faces[face_offset+3*f+0];
+            tinyobj_opt::vertex_index idx1 = attrib.faces[face_offset+3*f+1];
+            tinyobj_opt::vertex_index idx2 = attrib.faces[face_offset+3*f+2];
+
+            float v[3][3];
+            for (int k = 0; k < 3; k++) {
+              int f0 = idx0.v_idx;
+              int f1 = idx1.v_idx;
+              int f2 = idx2.v_idx;
+              assert(f0 >= 0);
+              assert(f1 >= 0);
+              assert(f2 >= 0);
+
+              v[0][k] = attrib.vertices[3*f0+k];
+              v[1][k] = attrib.vertices[3*f1+k];
+              v[2][k] = attrib.vertices[3*f2+k];
+              bmin[k] = std::min(v[0][k], bmin[k]);
+              bmin[k] = std::min(v[1][k], bmin[k]);
+              bmin[k] = std::min(v[2][k], bmin[k]);
+              bmax[k] = std::max(v[0][k], bmax[k]);
+              bmax[k] = std::max(v[1][k], bmax[k]);
+              bmax[k] = std::max(v[2][k], bmax[k]);
+            }
+
+            float n[3][3];
+
+            if (attrib.normals.size() > 0) { 
+              int f0 = idx0.vn_idx;
+              int f1 = idx1.vn_idx;
+              int f2 = idx2.vn_idx;
+              assert(f0 >= 0);
+              assert(f1 >= 0);
+              assert(f2 >= 0);
+              assert(3*f0+2 < attrib.normals.size());
+              assert(3*f1+2 < attrib.normals.size());
+              assert(3*f2+2 < attrib.normals.size());
+              for (int k = 0; k < 3; k++) {
+                n[0][k] = attrib.normals[3*f0+k];
+                n[1][k] = attrib.normals[3*f1+k];
+                n[2][k] = attrib.normals[3*f2+k];
+              }
+            } else {
+              // compute geometric normal
+              CalcNormal(n[0], v[0], v[1], v[2]);
+              n[1][0] = n[0][0]; n[1][1] = n[0][1]; n[1][2] = n[0][2];
+              n[2][0] = n[0][0]; n[2][1] = n[0][1]; n[2][2] = n[0][2];
+            }
+
+            for (int k = 0; k < 3; k++) {
+              vb.push_back(v[k][0]);
+              vb.push_back(v[k][1]);
+              vb.push_back(v[k][2]);
+              vb.push_back(n[k][0]);
+              vb.push_back(n[k][1]);
+              vb.push_back(n[k][2]);
+              // Use normal as color.
+              float c[3] = {n[k][0], n[k][1], n[k][2]};
+              float len2 = c[0] * c[0] + c[1] * c[1] + c[2] * c[2];
+              if (len2 > 0.0f) {
+                float len = sqrtf(len2);
+                
+                c[0] /= len;
+                c[1] /= len;
+                c[2] /= len;
+              }
+              vb.push_back(c[0] * 0.5 + 0.5);
+              vb.push_back(c[1] * 0.5 + 0.5);
+              vb.push_back(c[2] * 0.5 + 0.5);
+            }
+          } 
+          face_offset += attrib.face_num_verts[v];
+        }
+
+        o.vb = 0;
+        o.numTriangles = 0;
+        if (vb.size() > 0) {
+          glGenBuffers(1, &o.vb);
+          glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+          glBufferData(GL_ARRAY_BUFFER, vb.size() * sizeof(float), &vb.at(0), GL_STATIC_DRAW);
+          o.numTriangles = vb.size() / 9 / 3;
+        }
+
+        gDrawObjects.push_back(o);
+  }
+  
+  printf("bmin = %f, %f, %f\n", bmin[0], bmin[1], bmin[2]);
+  printf("bmax = %f, %f, %f\n", bmax[0], bmax[1], bmax[2]);
+
+  return true;
+#else
+  return false;
+#endif
+}
+
+void reshapeFunc(GLFWwindow* window, int w, int h)
+{
+  (void)window;
+  // for retinal display.
+  int fb_w, fb_h;
+  glfwGetFramebufferSize(window, &fb_w, &fb_h);
+
+  glViewport(0, 0, fb_w, fb_h);
+  glMatrixMode(GL_PROJECTION);
+  glLoadIdentity();
+  gluPerspective(45.0, (float)w / (float)h, 0.01f, 100.0f);
+  glMatrixMode(GL_MODELVIEW);
+  glLoadIdentity();
+
+  width = w;
+  height = h;
+}
+
+void keyboardFunc(GLFWwindow *window, int key, int scancode, int action, int mods) {
+  (void)window;
+  (void)scancode;
+  (void)mods;
+    if(action == GLFW_PRESS || action == GLFW_REPEAT){
+        // Move camera
+        float mv_x = 0, mv_y = 0, mv_z = 0;
+        if(key == GLFW_KEY_K) mv_x += 1;
+        else if(key == GLFW_KEY_J) mv_x += -1;
+        else if(key == GLFW_KEY_L) mv_y += 1;
+        else if(key == GLFW_KEY_H) mv_y += -1;
+        else if(key == GLFW_KEY_P) mv_z += 1;
+        else if(key == GLFW_KEY_N) mv_z += -1;
+        //camera.move(mv_x * 0.05, mv_y * 0.05, mv_z * 0.05);
+        // Close window
+        if(key == GLFW_KEY_Q || key == GLFW_KEY_ESCAPE) glfwSetWindowShouldClose(window, GL_TRUE);
+
+        //init_frame = true;
+    }
+}
+
+void clickFunc(GLFWwindow* window, int button, int action, int mods){
+  (void)window;
+  (void)mods;
+    if(button == GLFW_MOUSE_BUTTON_LEFT){
+        if(action == GLFW_PRESS){
+            mouseLeftPressed = true;
+            trackball(prev_quat, 0.0, 0.0, 0.0, 0.0);
+        } else if(action == GLFW_RELEASE){
+            mouseLeftPressed = false;
+        }
+    }
+    if(button == GLFW_MOUSE_BUTTON_RIGHT){
+        if(action == GLFW_PRESS){
+            mouseRightPressed = true;
+        } else if(action == GLFW_RELEASE){
+            mouseRightPressed = false;
+        }
+    }
+    if(button == GLFW_MOUSE_BUTTON_MIDDLE){
+        if(action == GLFW_PRESS){
+            mouseMiddlePressed = true;
+        } else if(action == GLFW_RELEASE){
+            mouseMiddlePressed = false;
+        }
+    }
+}
+
+void motionFunc(GLFWwindow* window, double mouse_x, double mouse_y){
+  (void)window;
+  float rotScale = 1.0f;
+  float transScale = 2.0f;
+
+    if(mouseLeftPressed){
+      trackball(prev_quat,
+          rotScale * (2.0f * prevMouseX - width) / (float)width,
+          rotScale * (height - 2.0f * prevMouseY) / (float)height,
+          rotScale * (2.0f * mouse_x - width) / (float)width,
+          rotScale * (height - 2.0f * mouse_y) / (float)height);
+
+      add_quats(prev_quat, curr_quat, curr_quat);
+    } else if (mouseMiddlePressed) {
+      eye[0] -= transScale * (mouse_x - prevMouseX) / (float)width;
+      lookat[0] -= transScale * (mouse_x - prevMouseX) / (float)width;
+      eye[1] += transScale * (mouse_y - prevMouseY) / (float)height;
+      lookat[1] += transScale * (mouse_y - prevMouseY) / (float)height;
+    } else if (mouseRightPressed) {
+      eye[2] += transScale * (mouse_y - prevMouseY) / (float)height;
+      lookat[2] += transScale * (mouse_y - prevMouseY) / (float)height;
+    }
+
+    // Update mouse point
+    prevMouseX = mouse_x;
+    prevMouseY = mouse_y;
+}
+
+void Draw(const std::vector<DrawObject>& drawObjects)
+{
+  glPolygonMode(GL_FRONT, GL_FILL);
+  glPolygonMode(GL_BACK, GL_FILL);
+
+  glEnable(GL_POLYGON_OFFSET_FILL);
+  glPolygonOffset(1.0, 1.0);
+  glColor3f(1.0f, 1.0f, 1.0f);
+  for (size_t i = 0; i < drawObjects.size(); i++) {
+    DrawObject o = drawObjects[i];
+    if (o.vb < 1) {
+      continue;
+    }
+ 
+    glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glEnableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float)*3));
+    glColorPointer(3, GL_FLOAT, 36, (const void*)(sizeof(float)*6));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * o.numTriangles);
+    CheckErrors("drawarrays");
+  }
+
+  // draw wireframe
+  glDisable(GL_POLYGON_OFFSET_FILL);
+  glPolygonMode(GL_FRONT, GL_LINE);
+  glPolygonMode(GL_BACK, GL_LINE);
+
+  glColor3f(0.0f, 0.0f, 0.4f);
+  for (size_t i = 0; i < drawObjects.size(); i++) {
+    DrawObject o = drawObjects[i];
+    if (o.vb < 1) {
+      continue;
+    }
+ 
+    glBindBuffer(GL_ARRAY_BUFFER, o.vb);
+    glEnableClientState(GL_VERTEX_ARRAY);
+    glEnableClientState(GL_NORMAL_ARRAY);
+    glDisableClientState(GL_COLOR_ARRAY);
+    glVertexPointer(3, GL_FLOAT, 36, (const void*)0);
+    glNormalPointer(GL_FLOAT, 36, (const void*)(sizeof(float)*3));
+
+    glDrawArrays(GL_TRIANGLES, 0, 3 * o.numTriangles);
+    CheckErrors("drawarrays");
+  }
+}
+
+static void Init() {
+  trackball(curr_quat, 0, 0, 0, 0);
+
+  eye[0] = 0.0f;
+  eye[1] = 0.0f;
+  eye[2] = 3.0f;
+
+  lookat[0] = 0.0f;
+  lookat[1] = 0.0f;
+  lookat[2] = 0.0f;
+
+  up[0] = 0.0f;
+  up[1] = 1.0f;
+  up[2] = 0.0f;
+}
+
+
+int main(int argc, char **argv)
+{
+  if (argc < 2) {
+    std::cout << "Needs input.obj\n" << std::endl;
+    return 0;
+  }
+
+  bool benchmark_only = false;
+  int num_threads = -1;
+  if (argc > 2) {
+    num_threads = atoi(argv[2]);
+  }
+
+  if (argc > 3) {
+    benchmark_only = true;
+  }
+
+  if (benchmark_only) {
+
+    tinyobj_opt::attrib_t attrib;
+    std::vector<tinyobj_opt::shape_t> shapes;
+
+    size_t data_len = 0;
+    const char* data = get_file_data(&data_len, argv[1]);
+    if (data == nullptr) {
+      exit(-1);
+      return false;
+    }
+    printf("filesize: %d\n", (int)data_len);
+    tinyobj_opt::LoadOption option;
+    option.req_num_threads = num_threads;
+    bool ret = parseObj(&attrib, &shapes, data, data_len, option);
+
+    return ret;
+  }
+
+  Init();
+
+  std::cout << "Initialize GLFW..." << std::endl;
+
+  if(!glfwInit()){
+    std::cerr << "Failed to initialize GLFW." << std::endl;
+    return -1;
+  }
+
+  std::cout << "GLFW OK." << std::endl;
+
+
+  window = glfwCreateWindow(width, height, "Obj viewer", NULL, NULL);
+  if(window == NULL){
+    std::cerr << "Failed to open GLFW window. " << std::endl;
+    glfwTerminate();
+    return 1;
+  }
+
+  glfwMakeContextCurrent(window);
+  glfwSwapInterval(1);
+
+  // Callback
+  glfwSetWindowSizeCallback(window, reshapeFunc);
+  glfwSetKeyCallback(window, keyboardFunc);
+  glfwSetMouseButtonCallback(window, clickFunc);
+  glfwSetCursorPosCallback(window, motionFunc);
+
+  glewExperimental = true;
+  if (glewInit() != GLEW_OK) {
+    std::cerr << "Failed to initialize GLEW." << std::endl;
+    return -1;
+  }
+
+  reshapeFunc(window, width, height);
+
+  float bmin[3], bmax[3];
+  if (false == LoadObjAndConvert(bmin, bmax, argv[1], num_threads)) {
+    printf("failed to load & conv\n");
+    return -1;
+  }
+
+  float maxExtent = 0.5f * (bmax[0] - bmin[0]);
+  if (maxExtent < 0.5f * (bmax[1] - bmin[1])) {
+    maxExtent = 0.5f * (bmax[1] - bmin[1]);
+  }
+  if (maxExtent < 0.5f * (bmax[2] - bmin[2])) {
+    maxExtent = 0.5f * (bmax[2] - bmin[2]);
+  }
+
+  while(glfwWindowShouldClose(window) == GL_FALSE) {
+    glfwPollEvents();
+    glClearColor(0.1f, 0.2f, 0.3f, 1.0f);
+    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+    glEnable(GL_DEPTH_TEST);
+
+    // camera & rotate
+    glMatrixMode(GL_MODELVIEW);
+    glLoadIdentity();
+    GLfloat mat[4][4];
+    gluLookAt(eye[0], eye[1], eye[2], lookat[0], lookat[1], lookat[2], up[0], up[1], up[2]);
+    build_rotmatrix(mat, curr_quat);
+    glMultMatrixf(&mat[0][0]);
+
+    // Fit to -1, 1
+    glScalef(1.0f / maxExtent, 1.0f / maxExtent, 1.0f / maxExtent);
+
+    // Centerize object.
+    glTranslatef(-0.5*(bmax[0] + bmin[0]), -0.5*(bmax[1] + bmin[1]), -0.5*(bmax[2] + bmin[2]));
+  
+    Draw(gDrawObjects);
+
+    glfwSwapBuffers(window);
+  }
+
+  glfwTerminate();
+}

jni/Android.mk

@@ -5,8 +5,8 @@ LOCAL_PATH := $(call my-dir)
 
 include $(CLEAR_VARS)
 LOCAL_MODULE := tinyobjloader
-LOCAL_SRC_FILES := ../tiny_obj_loader.cc ../test.cc
+LOCAL_SRC_FILES := ../tiny_obj_loader.cc
 
 LOCAL_C_INCLUDES := ../
 
-include $(BUILD_EXECUTABLE)
+include $(BUILD_STATIC_LIBRARY)

loader_example.cc

@@ -0,0 +1,382 @@
+//
+// g++ loader_example.cc
+//
+#define TINYOBJLOADER_IMPLEMENTATION
+#include "tiny_obj_loader.h"
+
+#include <cstdio>
+#include <cstdlib>
+#include <cassert>
+#include <iostream>
+#include <sstream>
+#include <fstream>
+
+#ifdef _WIN32
+#ifdef __cplusplus
+extern "C" {
+#endif
+#include <windows.h>
+#include <mmsystem.h>
+#ifdef __cplusplus
+}
+#endif
+#pragma comment(lib, "winmm.lib")
+#else
+#if defined(__unix__) || defined(__APPLE__)
+#include <sys/time.h>
+#else
+#include <ctime>
+#endif
+#endif
+
+class timerutil {
+public:
+#ifdef _WIN32
+  typedef DWORD time_t;
+
+  timerutil() { ::timeBeginPeriod(1); }
+  ~timerutil() { ::timeEndPeriod(1); }
+
+  void start() { t_[0] = ::timeGetTime(); }
+  void end() { t_[1] = ::timeGetTime(); }
+
+  time_t sec() { return (time_t)((t_[1] - t_[0]) / 1000); }
+  time_t msec() { return (time_t)((t_[1] - t_[0])); }
+  time_t usec() { return (time_t)((t_[1] - t_[0]) * 1000); }
+  time_t current() { return ::timeGetTime(); }
+
+#else
+#if defined(__unix__) || defined(__APPLE__)
+  typedef unsigned long int time_t;
+
+  void start() { gettimeofday(tv + 0, &tz); }
+  void end() { gettimeofday(tv + 1, &tz); }
+
+  time_t sec() { return static_cast<time_t>(tv[1].tv_sec - tv[0].tv_sec); }
+  time_t msec() {
+    return this->sec() * 1000 +
+           static_cast<time_t>((tv[1].tv_usec - tv[0].tv_usec) / 1000);
+  }
+  time_t usec() {
+    return this->sec() * 1000000 + static_cast<time_t>(tv[1].tv_usec - tv[0].tv_usec);
+  }
+  time_t current() {
+    struct timeval t;
+    gettimeofday(&t, NULL);
+    return static_cast<time_t>(t.tv_sec * 1000 + t.tv_usec);
+  }
+
+#else // C timer
+  // using namespace std;
+  typedef clock_t time_t;
+
+  void start() { t_[0] = clock(); }
+  void end() { t_[1] = clock(); }
+
+  time_t sec() { return (time_t)((t_[1] - t_[0]) / CLOCKS_PER_SEC); }
+  time_t msec() { return (time_t)((t_[1] - t_[0]) * 1000 / CLOCKS_PER_SEC); }
+  time_t usec() { return (time_t)((t_[1] - t_[0]) * 1000000 / CLOCKS_PER_SEC); }
+  time_t current() { return (time_t)clock(); }
+
+#endif
+#endif
+
+private:
+#ifdef _WIN32
+  DWORD t_[2];
+#else
+#if defined(__unix__) || defined(__APPLE__)
+  struct timeval tv[2];
+  struct timezone tz;
+#else
+  time_t t_[2];
+#endif
+#endif
+};
+
+static void PrintInfo(const tinyobj::attrib_t &attrib, const std::vector<tinyobj::shape_t>& shapes, const std::vector<tinyobj::material_t>& materials)
+{
+  std::cout << "# of vertices  : " << (attrib.vertices.size() / 3) << std::endl;
+  std::cout << "# of normals   : " << (attrib.normals.size() / 3) << std::endl;
+  std::cout << "# of texcoords : " << (attrib.texcoords.size() / 2) << std::endl;
+
+  std::cout << "# of shapes    : " << shapes.size() << std::endl;
+  std::cout << "# of materials : " << materials.size() << std::endl;
+
+  for (size_t v = 0; v < attrib.vertices.size() / 3; v++) {
+    printf("  v[%ld] = (%f, %f, %f)\n", static_cast<long>(v),
+      static_cast<const double>(attrib.vertices[3*v+0]),
+      static_cast<const double>(attrib.vertices[3*v+1]),
+      static_cast<const double>(attrib.vertices[3*v+2]));
+  }
+
+  for (size_t v = 0; v < attrib.normals.size() / 3; v++) {
+    printf("  n[%ld] = (%f, %f, %f)\n", static_cast<long>(v),
+      static_cast<const double>(attrib.normals[3*v+0]),
+      static_cast<const double>(attrib.normals[3*v+1]),
+      static_cast<const double>(attrib.normals[3*v+2]));
+  }
+
+  for (size_t v = 0; v < attrib.texcoords.size() / 2; v++) {
+    printf("  uv[%ld] = (%f, %f)\n", static_cast<long>(v),
+      static_cast<const double>(attrib.texcoords[2*v+0]),
+      static_cast<const double>(attrib.texcoords[2*v+1]));
+  }
+
+  // For each shape
+  for (size_t i = 0; i < shapes.size(); i++) {
+    printf("shape[%ld].name = %s\n", static_cast<long>(i), shapes[i].name.c_str());
+    printf("Size of shape[%ld].indices: %lu\n", static_cast<long>(i), static_cast<unsigned long>(shapes[i].mesh.indices.size()));
+
+	size_t index_offset = 0;
+
+	assert(shapes[i].mesh.num_face_vertices.size() == shapes[i].mesh.material_ids.size());
+
+	printf("shape[%ld].num_faces: %lu\n", static_cast<long>(i), static_cast<unsigned long>(shapes[i].mesh.num_face_vertices.size()));
+
+	// For each face
+    for (size_t f = 0; f < shapes[i].mesh.num_face_vertices.size(); f++) {
+		size_t fnum = shapes[i].mesh.num_face_vertices[f];
+
+		printf("  face[%ld].fnum = %ld\n", static_cast<long>(f), static_cast<unsigned long>(fnum));
+
+		// For each vertex in the face
+		for (size_t v = 0; v < fnum; v++) {
+			tinyobj::index_t idx = shapes[i].mesh.indices[index_offset + v];
+			printf("    face[%ld].v[%ld].idx = %d/%d/%d\n", static_cast<long>(f), static_cast<long>(v), idx.vertex_index, idx.normal_index, idx.texcoord_index);
+		}
+
+		printf("  face[%ld].material_id = %d\n", static_cast<long>(f), shapes[i].mesh.material_ids[f]);
+
+		index_offset += fnum;
+    }
+
+    printf("shape[%ld].num_tags: %lu\n", static_cast<long>(i), static_cast<unsigned long>(shapes[i].mesh.tags.size()));
+    for (size_t t = 0; t < shapes[i].mesh.tags.size(); t++) {
+      printf("  tag[%ld] = %s ", static_cast<long>(t), shapes[i].mesh.tags[t].name.c_str());
+      printf(" ints: [");
+      for (size_t j = 0; j < shapes[i].mesh.tags[t].intValues.size(); ++j)
+      {
+          printf("%ld", static_cast<long>(shapes[i].mesh.tags[t].intValues[j]));
+          if (j < (shapes[i].mesh.tags[t].intValues.size()-1))
+          {
+              printf(", ");
+          }
+      }
+      printf("]");
+
+      printf(" floats: [");
+      for (size_t j = 0; j < shapes[i].mesh.tags[t].floatValues.size(); ++j)
+      {
+          printf("%f", static_cast<const double>(shapes[i].mesh.tags[t].floatValues[j]));
+          if (j < (shapes[i].mesh.tags[t].floatValues.size()-1))
+          {
+              printf(", ");
+          }
+      }
+      printf("]");
+
+      printf(" strings: [");
+      for (size_t j = 0; j < shapes[i].mesh.tags[t].stringValues.size(); ++j)
+      {
+          printf("%s", shapes[i].mesh.tags[t].stringValues[j].c_str());
+          if (j < (shapes[i].mesh.tags[t].stringValues.size()-1))
+          {
+              printf(", ");
+          }
+      }
+      printf("]");
+      printf("\n");
+    }
+  }
+
+  for (size_t i = 0; i < materials.size(); i++) {
+    printf("material[%ld].name = %s\n", static_cast<long>(i), materials[i].name.c_str());
+    printf("  material.Ka = (%f, %f ,%f)\n", static_cast<const double>(materials[i].ambient[0]),       static_cast<const double>(materials[i].ambient[1]),       static_cast<const double>(materials[i].ambient[2]));
+    printf("  material.Kd = (%f, %f ,%f)\n", static_cast<const double>(materials[i].diffuse[0]),       static_cast<const double>(materials[i].diffuse[1]),       static_cast<const double>(materials[i].diffuse[2]));
+    printf("  material.Ks = (%f, %f ,%f)\n", static_cast<const double>(materials[i].specular[0]),      static_cast<const double>(materials[i].specular[1]),      static_cast<const double>(materials[i].specular[2]));
+    printf("  material.Tr = (%f, %f ,%f)\n", static_cast<const double>(materials[i].transmittance[0]), static_cast<const double>(materials[i].transmittance[1]), static_cast<const double>(materials[i].transmittance[2]));
+    printf("  material.Ke = (%f, %f ,%f)\n", static_cast<const double>(materials[i].emission[0]),      static_cast<const double>(materials[i].emission[1]),      static_cast<const double>(materials[i].emission[2]));
+    printf("  material.Ns = %f\n", static_cast<const double>(materials[i].shininess));
+    printf("  material.Ni = %f\n", static_cast<const double>(materials[i].ior));
+    printf("  material.dissolve = %f\n", static_cast<const double>(materials[i].dissolve));
+    printf("  material.illum = %d\n",  materials[i].illum);
+    printf("  material.map_Ka = %s\n", materials[i].ambient_texname.c_str());
+    printf("  material.map_Kd = %s\n", materials[i].diffuse_texname.c_str());
+    printf("  material.map_Ks = %s\n", materials[i].specular_texname.c_str());
+    printf("  material.map_Ns = %s\n", materials[i].specular_highlight_texname.c_str());
+    printf("  material.map_bump = %s\n", materials[i].bump_texname.c_str());
+    printf("  material.map_d = %s\n", materials[i].alpha_texname.c_str());
+    printf("  material.disp = %s\n", materials[i].displacement_texname.c_str());
+    std::map<std::string, std::string>::const_iterator it(materials[i].unknown_parameter.begin());
+    std::map<std::string, std::string>::const_iterator itEnd(materials[i].unknown_parameter.end());
+
+    for (; it != itEnd; it++) {
+      printf("  material.%s = %s\n", it->first.c_str(), it->second.c_str());
+    }
+    printf("\n");
+  }
+}
+
+static bool
+TestLoadObj(
+  const char* filename,
+  const char* basepath = NULL,
+  bool triangulate = true)
+{
+  std::cout << "Loading " << filename << std::endl;
+
+  tinyobj::attrib_t attrib;
+  std::vector<tinyobj::shape_t> shapes;
+  std::vector<tinyobj::material_t> materials;
+
+  timerutil t;
+  t.start();
+  std::string err;
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, filename, basepath, triangulate);
+  t.end();
+  printf("Parsing time: %lu [msecs]\n", t.msec());
+
+  if (!err.empty()) {
+    std::cerr << err << std::endl;
+  }
+
+  if (!ret) {
+    printf("Failed to load/parse .obj.\n");
+    return false;
+  }
+
+  PrintInfo(attrib, shapes, materials);
+
+  return true;
+}
+
+
+static bool
+TestStreamLoadObj()
+{
+  std::cout << "Stream Loading " << std::endl;
+
+  std::stringstream objStream;
+  objStream
+    << "mtllib cube.mtl\n"
+    "\n"
+    "v 0.000000 2.000000 2.000000\n"
+    "v 0.000000 0.000000 2.000000\n"
+    "v 2.000000 0.000000 2.000000\n"
+    "v 2.000000 2.000000 2.000000\n"
+    "v 0.000000 2.000000 0.000000\n"
+    "v 0.000000 0.000000 0.000000\n"
+    "v 2.000000 0.000000 0.000000\n"
+    "v 2.000000 2.000000 0.000000\n"
+    "# 8 vertices\n"
+    "\n"
+    "g front cube\n"
+    "usemtl white\n"
+    "f 1 2 3 4\n"
+    "g back cube\n"
+    "# expects white material\n"
+    "f 8 7 6 5\n"
+    "g right cube\n"
+    "usemtl red\n"
+    "f 4 3 7 8\n"
+    "g top cube\n"
+    "usemtl white\n"
+    "f 5 1 4 8\n"
+    "g left cube\n"
+    "usemtl green\n"
+    "f 5 6 2 1\n"
+    "g bottom cube\n"
+    "usemtl white\n"
+    "f 2 6 7 3\n"
+    "# 6 elements";
+
+std::string matStream(
+    "newmtl white\n"
+    "Ka 0 0 0\n"
+    "Kd 1 1 1\n"
+    "Ks 0 0 0\n"
+    "\n"
+    "newmtl red\n"
+    "Ka 0 0 0\n"
+    "Kd 1 0 0\n"
+    "Ks 0 0 0\n"
+    "\n"
+    "newmtl green\n"
+    "Ka 0 0 0\n"
+    "Kd 0 1 0\n"
+    "Ks 0 0 0\n"
+    "\n"
+    "newmtl blue\n"
+    "Ka 0 0 0\n"
+    "Kd 0 0 1\n"
+    "Ks 0 0 0\n"
+    "\n"
+    "newmtl light\n"
+    "Ka 20 20 20\n"
+    "Kd 1 1 1\n"
+    "Ks 0 0 0");
+
+    using namespace tinyobj;
+    class MaterialStringStreamReader:
+        public MaterialReader
+    {
+        public:
+            MaterialStringStreamReader(const std::string& matSStream): m_matSStream(matSStream) {}
+            virtual ~MaterialStringStreamReader() {}
+            virtual bool operator() (
+              const std::string& matId,
+              std::vector<material_t>* materials,
+              std::map<std::string, int>* matMap,
+              std::string* err)
+            {
+                (void)matId;
+                (void)err;
+                LoadMtl(matMap, materials, &m_matSStream);
+                return true;
+            }
+
+        private:
+            std::stringstream m_matSStream;
+    };
+
+  MaterialStringStreamReader matSSReader(matStream);
+  tinyobj::attrib_t attrib;
+  std::vector<tinyobj::shape_t> shapes;
+  std::vector<tinyobj::material_t> materials;
+  std::string err;
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, &objStream, &matSSReader);    
+  
+  if (!err.empty()) {
+    std::cerr << err << std::endl;
+  }
+
+  if (!ret) {
+    return false;
+  }
+
+  PrintInfo(attrib, shapes, materials);
+    
+  return true;
+}
+
+int
+main(
+  int argc,
+  char **argv)
+{
+  if (argc > 1) {
+    const char* basepath = "models/";
+    if (argc > 2) {
+      basepath = argv[2];
+    }
+    assert(true == TestLoadObj(argv[1], basepath));
+  } else {
+    //assert(true == TestLoadObj("cornell_box.obj"));
+    //assert(true == TestLoadObj("cube.obj"));
+    assert(true == TestStreamLoadObj());
+    assert(true == TestLoadObj("models/catmark_torus_creases0.obj", "models/", false));
+  }
+
+  return 0;
+}

premake4.lua

@@ -1,5 +1,5 @@
 sources = {
-   "test.cc",
+   "loader_example.cc",
    }
 
 -- premake4.lua
@@ -21,9 +21,9 @@ solution "TinyObjLoaderSolution"
       configuration "Debug"
          defines { "DEBUG" } -- -DDEBUG
          flags { "Symbols" }
-         targetname "test_tinyobjloader_debug"
+         targetname "loader_example_debug"
 
       configuration "Release"
          -- defines { "NDEBUG" } -- -NDEBUG
          flags { "Symbols", "Optimize" }
-         targetname "test_tinyobjloader"
+         targetname "loader_example"

tests/Makefile

@@ -0,0 +1,16 @@
+.PHONY: clean
+
+CXX ?= g++
+CXXFLAGS ?= -g -O2
+
+tester: tester.cc
+	$(CXX) $(CXXFLAGS) -o tester tester.cc
+
+all: tester
+
+check: tester
+	./tester	
+
+clean:
+	rm -rf tester
+

tests/README.md

@@ -0,0 +1,37 @@
+# Build&Test
+
+## Use makefile
+
+    $ make check
+
+## Use ninja + kuroga
+
+Assume
+
+* ninja 1.4+
+* python 2.6+
+
+Are installed.
+
+### Linux/MacOSX
+
+    $ python kuroga.py config-posix.py
+    $ ninja
+
+### Windows
+
+Visual Studio 2013 is required to build tester.
+
+On Windows console.
+
+    > python kuroga.py config-msvc.py
+    > vcbuild.bat
+
+
+Or on msys2 bash,
+
+    $ python kuroga.py config-msvc.py
+    $ cmd //c vcbuild.bat
+
+ 
+

tests/config-msvc.py

@@ -0,0 +1,52 @@
+exe = "tester.exe"
+
+toolchain = "msvc"
+
+# optional
+link_pool_depth = 1
+
+# optional
+builddir = {
+    "gnu" :  "build"
+  , "msvc" :  "build"
+  , "clang" :  "build"
+    }
+
+includes = {
+    "gnu" : [ "-I." ]
+  , "msvc" : [ "/I." ]
+  , "clang" : [ "-I." ]
+    }
+
+defines = {
+    "gnu" : [ "-DEXAMPLE=1" ]
+  , "msvc" : [ "/DEXAMPLE=1" ]
+  , "clang" : [ "-DEXAMPLE=1" ]
+    }
+
+cflags = {
+    "gnu" : [ "-O2", "-g" ]
+  , "msvc" : [ "/O2" ]
+  , "clang" : [ "-O2", "-g" ]
+    }
+
+cxxflags = {
+    "gnu" : [ "-O2", "-g" ]
+  , "msvc" : [ "/O2", "/W4", "/EHsc"]
+  , "clang" : [ "-O2", "-g", "-fsanitize=address" ]
+    }
+
+ldflags = {
+    "gnu" : [ ]
+  , "msvc" : [ ]
+  , "clang" : [ "-fsanitize=address" ]
+    }
+
+# optionsl
+cxx_files = [ "tester.cc" ]
+c_files = [ ]
+
+# You can register your own toolchain through register_toolchain function
+def register_toolchain(ninja):
+    pass
+

tests/config-posix.py

@@ -0,0 +1,53 @@
+exe = "tester"
+
+# "gnu" or "clang"
+toolchain = "gnu"
+
+# optional
+link_pool_depth = 1
+
+# optional
+builddir = {
+    "gnu" :  "build"
+  , "msvc" :  "build"
+  , "clang" :  "build"
+    }
+
+includes = {
+    "gnu" : [ "-I." ]
+  , "msvc" : [ "/I." ]
+  , "clang" : [ "-I." ]
+    }
+
+defines = {
+    "gnu" : [ ]
+  , "msvc" : [ ]
+  , "clang" : [  ]
+    }
+
+cflags = {
+    "gnu" : [ "-O2", "-g" ]
+  , "msvc" : [ "/O2" ]
+  , "clang" : [ "-O2", "-g" ]
+    }
+
+# Warn as much as possible: http://qiita.com/MitsutakaTakeda/items/6b9966f890cc9b944d75
+cxxflags = {
+    "gnu" : [ "-O2", "-g", "-pedantic -Wall -Wextra -Wcast-align -Wcast-qual -Wctor-dtor-privacy -Wdisabled-optimization -Wformat=2 -Winit-self -Wmissing-declarations -Wmissing-include-dirs -Wold-style-cast -Woverloaded-virtual -Wredundant-decls -Wshadow -Wsign-conversion -Wsign-promo -Wstrict-overflow=5 -Wswitch-default -Wundef -Werror -Wno-unused", "-fsanitize=address" ]
+  , "msvc" : [ "/O2", "/W4" ]
+  , "clang" : [ "-O2", "-g", "-Werror -Weverything -Wno-c++98-compat -Wno-c++98-compat-pedantic", "-fsanitize=address" ]
+    }
+
+ldflags = {
+    "gnu" : [ "-fsanitize=address" ]
+  , "msvc" : [ ]
+  , "clang" : [ "-fsanitize=address" ]
+    }
+
+cxx_files = [ "tester.cc" ]
+c_files = [ ]
+
+# You can register your own toolchain through register_toolchain function
+def register_toolchain(ninja):
+    pass
+

tests/kuroga.py

@@ -0,0 +1,312 @@
+#!/usr/bin/env python
+
+#
+# Kuroga, single python file meta-build system for ninja
+# https://github.com/lighttransport/kuroga
+#
+# Requirements: python 2.6 or 2.7
+#
+# Usage: $ python kuroga.py input.py
+#
+
+import imp
+import re
+import textwrap
+import glob
+import os
+import sys
+
+# gcc preset
+def add_gnu_rule(ninja):
+    ninja.rule('gnucxx', description='CXX $out',
+        command='$gnucxx -MMD -MF $out.d $gnudefines $gnuincludes $gnucxxflags -c $in -o $out',
+        depfile='$out.d', deps='gcc')
+    ninja.rule('gnucc', description='CC $out',
+        command='$gnucc -MMD -MF $out.d $gnudefines $gnuincludes $gnucflags -c $in -o $out',
+        depfile='$out.d', deps='gcc')
+    ninja.rule('gnulink', description='LINK $out', pool='link_pool',
+        command='$gnuld -o $out $in $libs $gnuldflags')
+    ninja.rule('gnuar', description='AR $out', pool='link_pool',
+        command='$gnuar rsc $out $in')
+    ninja.rule('gnustamp', description='STAMP $out', command='touch $out')
+    ninja.newline()
+
+    ninja.variable('gnucxx', 'g++')
+    ninja.variable('gnucc', 'gcc')
+    ninja.variable('gnuld', '$gnucxx')
+    ninja.variable('gnuar', 'ar')
+    ninja.newline()
+
+# clang preset
+def add_clang_rule(ninja):
+    ninja.rule('clangcxx', description='CXX $out',
+        command='$clangcxx -MMD -MF $out.d $clangdefines $clangincludes $clangcxxflags -c $in -o $out',
+        depfile='$out.d', deps='gcc')
+    ninja.rule('clangcc', description='CC $out',
+        command='$clangcc -MMD -MF $out.d $clangdefines $clangincludes $clangcflags -c $in -o $out',
+        depfile='$out.d', deps='gcc')
+    ninja.rule('clanglink', description='LINK $out', pool='link_pool',
+        command='$clangld -o $out $in $libs $clangldflags')
+    ninja.rule('clangar', description='AR $out', pool='link_pool',
+        command='$clangar rsc $out $in')
+    ninja.rule('clangstamp', description='STAMP $out', command='touch $out')
+    ninja.newline()
+
+    ninja.variable('clangcxx', 'clang++')
+    ninja.variable('clangcc', 'clang')
+    ninja.variable('clangld', '$clangcxx')
+    ninja.variable('clangar', 'ar')
+    ninja.newline()
+
+# msvc preset
+def add_msvc_rule(ninja):
+    ninja.rule('msvccxx', description='CXX $out',
+        command='$msvccxx /TP /showIncludes $msvcdefines $msvcincludes $msvccxxflags -c $in /Fo$out',
+        depfile='$out.d', deps='msvc')
+    ninja.rule('msvccc', description='CC $out',
+        command='$msvccc /TC /showIncludes $msvcdefines $msvcincludes $msvccflags -c $in /Fo$out',
+        depfile='$out.d', deps='msvc')
+    ninja.rule('msvclink', description='LINK $out', pool='link_pool',
+        command='$msvcld $msvcldflags $in $libs /OUT:$out')
+    ninja.rule('msvcar', description='AR $out', pool='link_pool',
+        command='$msvcar $in /OUT:$out')
+    #ninja.rule('msvcstamp', description='STAMP $out', command='touch $out')
+    ninja.newline()
+
+    ninja.variable('msvccxx', 'cl.exe')
+    ninja.variable('msvccc', 'cl.exe')
+    ninja.variable('msvcld', 'link.exe')
+    ninja.variable('msvcar', 'lib.exe')
+    ninja.newline()
+
+# -- from ninja_syntax.py --
+def escape_path(word):
+    return word.replace('$ ', '$$ ').replace(' ', '$ ').replace(':', '$:')
+
+class Writer(object):
+    def __init__(self, output, width=78):
+        self.output = output
+        self.width = width
+
+    def newline(self):
+        self.output.write('\n')
+
+    def comment(self, text, has_path=False):
+        for line in textwrap.wrap(text, self.width - 2, break_long_words=False,
+                                  break_on_hyphens=False):
+            self.output.write('# ' + line + '\n')
+
+    def variable(self, key, value, indent=0):
+        if value is None:
+            return
+        if isinstance(value, list):
+            value = ' '.join(filter(None, value))  # Filter out empty strings.
+        self._line('%s = %s' % (key, value), indent)
+
+    def pool(self, name, depth):
+        self._line('pool %s' % name)
+        self.variable('depth', depth, indent=1)
+
+    def rule(self, name, command, description=None, depfile=None,
+             generator=False, pool=None, restat=False, rspfile=None,
+             rspfile_content=None, deps=None):
+        self._line('rule %s' % name)
+        self.variable('command', command, indent=1)
+        if description:
+            self.variable('description', description, indent=1)
+        if depfile:
+            self.variable('depfile', depfile, indent=1)
+        if generator:
+            self.variable('generator', '1', indent=1)
+        if pool:
+            self.variable('pool', pool, indent=1)
+        if restat:
+            self.variable('restat', '1', indent=1)
+        if rspfile:
+            self.variable('rspfile', rspfile, indent=1)
+        if rspfile_content:
+            self.variable('rspfile_content', rspfile_content, indent=1)
+        if deps:
+            self.variable('deps', deps, indent=1)
+
+    def build(self, outputs, rule, inputs=None, implicit=None, order_only=None,
+              variables=None):
+        outputs = as_list(outputs)
+        out_outputs = [escape_path(x) for x in outputs]
+        all_inputs = [escape_path(x) for x in as_list(inputs)]
+
+        if implicit:
+            implicit = [escape_path(x) for x in as_list(implicit)]
+            all_inputs.append('|')
+            all_inputs.extend(implicit)
+        if order_only:
+            order_only = [escape_path(x) for x in as_list(order_only)]
+            all_inputs.append('||')
+            all_inputs.extend(order_only)
+
+        self._line('build %s: %s' % (' '.join(out_outputs),
+                                     ' '.join([rule] + all_inputs)))
+
+        if variables:
+            if isinstance(variables, dict):
+                iterator = iter(variables.items())
+            else:
+                iterator = iter(variables)
+
+            for key, val in iterator:
+                self.variable(key, val, indent=1)
+
+        return outputs
+
+    def include(self, path):
+        self._line('include %s' % path)
+
+    def subninja(self, path):
+        self._line('subninja %s' % path)
+
+    def default(self, paths):
+        self._line('default %s' % ' '.join(as_list(paths)))
+
+    def _count_dollars_before_index(self, s, i):
+        """Returns the number of '$' characters right in front of s[i]."""
+        dollar_count = 0
+        dollar_index = i - 1
+        while dollar_index > 0 and s[dollar_index] == '$':
+            dollar_count += 1
+            dollar_index -= 1
+        return dollar_count
+
+    def _line(self, text, indent=0):
+        """Write 'text' word-wrapped at self.width characters."""
+        leading_space = '  ' * indent
+        while len(leading_space) + len(text) > self.width:
+            # The text is too wide; wrap if possible.
+
+            # Find the rightmost space that would obey our width constraint and
+            # that's not an escaped space.
+            available_space = self.width - len(leading_space) - len(' $')
+            space = available_space
+            while True:
+                space = text.rfind(' ', 0, space)
+                if (space < 0 or
+                    self._count_dollars_before_index(text, space) % 2 == 0):
+                    break
+
+            if space < 0:
+                # No such space; just use the first unescaped space we can find.
+                space = available_space - 1
+                while True:
+                    space = text.find(' ', space + 1)
+                    if (space < 0 or
+                        self._count_dollars_before_index(text, space) % 2 == 0):
+                        break
+            if space < 0:
+                # Give up on breaking.
+                break
+
+            self.output.write(leading_space + text[0:space] + ' $\n')
+            text = text[space+1:]
+
+            # Subsequent lines are continuations, so indent them.
+            leading_space = '  ' * (indent+2)
+
+        self.output.write(leading_space + text + '\n')
+
+    def close(self):
+        self.output.close()
+
+
+def as_list(input):
+    if input is None:
+        return []
+    if isinstance(input, list):
+        return input
+    return [input]
+
+# -- end from ninja_syntax.py --
+
+def gen(ninja, toolchain, config):
+    
+    ninja.variable('ninja_required_version', '1.4')
+    ninja.newline()
+
+    if hasattr(config, "builddir"):
+        builddir = config.builddir[toolchain]
+        ninja.variable(toolchain + 'builddir', builddir)
+    else:
+        builddir = ''
+
+    ninja.variable(toolchain + 'defines', config.defines[toolchain] or [])
+    ninja.variable(toolchain + 'includes', config.includes[toolchain] or [])
+    ninja.variable(toolchain + 'cflags', config.cflags[toolchain] or [])
+    ninja.variable(toolchain + 'cxxflags', config.cxxflags[toolchain] or [])
+    ninja.variable(toolchain + 'ldflags', config.ldflags[toolchain] or [])
+    ninja.newline()
+
+    if hasattr(config, "link_pool_depth"):
+        ninja.pool('link_pool', depth=config.link_pool_depth)
+    else:
+        ninja.pool('link_pool', depth=4)
+    ninja.newline()
+
+    # Add default toolchain(gnu, clang and msvc)
+    add_gnu_rule(ninja)
+    add_clang_rule(ninja)
+    add_msvc_rule(ninja)
+
+    obj_files = []
+
+    cc = toolchain + 'cc'
+    cxx = toolchain + 'cxx'
+    link = toolchain + 'link'
+    ar = toolchain + 'ar'
+    
+    if hasattr(config, "cxx_files"):
+        for src in config.cxx_files:
+            srcfile = src
+            obj = os.path.splitext(srcfile)[0] + '.o'
+            obj = os.path.join(builddir, obj);
+            obj_files.append(obj)
+            ninja.build(obj, cxx, srcfile)
+        ninja.newline()
+
+    if hasattr(config, "c_files"):
+        for src in config.c_files:
+            srcfile = src
+            obj = os.path.splitext(srcfile)[0] + '.o'
+            obj = os.path.join(builddir, obj);
+            obj_files.append(obj)
+            ninja.build(obj, cc, srcfile)
+        ninja.newline()
+
+    targetlist = []
+    if hasattr(config, "exe"):
+        ninja.build(config.exe, link, obj_files)
+        targetlist.append(config.exe)
+
+    if hasattr(config, "staticlib"):
+        ninja.build(config.staticlib, ar, obj_files)
+        targetlist.append(config.staticlib)
+
+    ninja.build('all', 'phony', targetlist)
+    ninja.newline()
+        
+    ninja.default('all')
+
+def main():
+    if len(sys.argv) < 2:
+        print("Usage: python kuroga.py config.py")
+        sys.exit(1)
+
+    config = imp.load_source("config", sys.argv[1])
+
+    f = open('build.ninja', 'w')
+    ninja = Writer(f)
+
+    if hasattr(config, "register_toolchain"):
+        config.register_toolchain(ninja)
+        
+    gen(ninja, config.toolchain, config)
+    f.close()
+
+main()

tests/tester.cc

@@ -1,5 +1,8 @@
 #define TINYOBJLOADER_IMPLEMENTATION
-#include "tiny_obj_loader.h"
+#include "../tiny_obj_loader.h"
+
+#define CATCH_CONFIG_MAIN  // This tells Catch to provide a main() - only do this in one cpp file
+#include "catch.hpp"
 
 #include <cstdio>
 #include <cstdlib>
@@ -8,11 +11,35 @@
 #include <sstream>
 #include <fstream>
 
-static void PrintInfo(const std::vector<tinyobj::shape_t>& shapes, const std::vector<tinyobj::material_t>& materials, bool triangulate = true)
+static void PrintInfo(const tinyobj::attrib_t &attrib, const std::vector<tinyobj::shape_t>& shapes, const std::vector<tinyobj::material_t>& materials, bool triangulate = true)
 {
+  std::cout << "# of vertices  : " << (attrib.vertices.size() / 3) << std::endl;
+  std::cout << "# of normals   : " << (attrib.normals.size() / 3) << std::endl;
+  std::cout << "# of texcoords : " << (attrib.texcoords.size() / 2) << std::endl;
+
   std::cout << "# of shapes    : " << shapes.size() << std::endl;
   std::cout << "# of materials : " << materials.size() << std::endl;
 
+  for (size_t v = 0; v < attrib.vertices.size() / 3; v++) {
+    printf("  v[%ld] = (%f, %f, %f)\n", v,
+      static_cast<const double>(attrib.vertices[3*v+0]),
+      static_cast<const double>(attrib.vertices[3*v+1]),
+      static_cast<const double>(attrib.vertices[3*v+2]));
+  }
+
+  for (size_t v = 0; v < attrib.normals.size() / 3; v++) {
+    printf("  n[%ld] = (%f, %f, %f)\n", v,
+      static_cast<const double>(attrib.normals[3*v+0]),
+      static_cast<const double>(attrib.normals[3*v+1]),
+      static_cast<const double>(attrib.normals[3*v+2]));
+  }
+
+  for (size_t v = 0; v < attrib.texcoords.size() / 2; v++) {
+    printf("  uv[%ld] = (%f, %f)\n", v,
+      static_cast<const double>(attrib.texcoords[2*v+0]),
+      static_cast<const double>(attrib.texcoords[2*v+1]));
+  }
+
   for (size_t i = 0; i < shapes.size(); i++) {
     printf("shape[%ld].name = %s\n", i, shapes[i].name.c_str());
     printf("Size of shape[%ld].indices: %ld\n", i, shapes[i].mesh.indices.size());
@@ -22,15 +49,23 @@ static void PrintInfo(const std::vector<tinyobj::shape_t>& shapes, const std::ve
         printf("Size of shape[%ld].material_ids: %ld\n", i, shapes[i].mesh.material_ids.size());
         assert((shapes[i].mesh.indices.size() % 3) == 0);
         for (size_t f = 0; f < shapes[i].mesh.indices.size() / 3; f++) {
-          printf("  idx[%ld] = %d, %d, %d. mat_id = %d\n", f, shapes[i].mesh.indices[3*f+0], shapes[i].mesh.indices[3*f+1], shapes[i].mesh.indices[3*f+2], shapes[i].mesh.material_ids[f]);
+          tinyobj::index_t i0 = shapes[i].mesh.indices[3*f+0];
+          tinyobj::index_t i1 = shapes[i].mesh.indices[3*f+1];
+          tinyobj::index_t i2 = shapes[i].mesh.indices[3*f+2];
+          printf("  idx[%ld] = %d/%d/%d, %d/%d/%d, %d/%d/%d. mat_id = %d\n", f,
+            i0.vertex_index, i0.normal_index, i0.texcoord_index,
+            i1.vertex_index, i1.normal_index, i1.texcoord_index,
+            i2.vertex_index, i2.normal_index, i2.texcoord_index,
+            shapes[i].mesh.material_ids[f]);
         }
     } else {
       for (size_t f = 0; f < shapes[i].mesh.indices.size(); f++) {
-        printf("  idx[%ld] = %d\n", f, shapes[i].mesh.indices[f]);
+        tinyobj::index_t idx = shapes[i].mesh.indices[f];
+        printf("  idx[%ld] = %d/%d/%d\n", f, idx.vertex_index, idx.normal_index, idx.texcoord_index);
       }
 
       printf("Size of shape[%ld].material_ids: %ld\n", i, shapes[i].mesh.material_ids.size());
-      assert(shapes[i].mesh.material_ids.size() == shapes[i].mesh.num_vertices.size());
+      assert(shapes[i].mesh.material_ids.size() == shapes[i].mesh.num_face_vertices.size());
       for (size_t m = 0; m < shapes[i].mesh.material_ids.size(); m++) {
         printf("  material_id[%ld] = %d\n", m,
           shapes[i].mesh.material_ids[m]);
@@ -38,20 +73,20 @@ static void PrintInfo(const std::vector<tinyobj::shape_t>& shapes, const std::ve
 
     }
 
-    printf("shape[%ld].num_faces: %ld\n", i, shapes[i].mesh.num_vertices.size());
-    for (size_t v = 0; v < shapes[i].mesh.num_vertices.size(); v++) {
+    printf("shape[%ld].num_faces: %ld\n", i, shapes[i].mesh.num_face_vertices.size());
+    for (size_t v = 0; v < shapes[i].mesh.num_face_vertices.size(); v++) {
       printf("  num_vertices[%ld] = %ld\n", v,
-        static_cast<long>(shapes[i].mesh.num_vertices[v]));
+        static_cast<long>(shapes[i].mesh.num_face_vertices[v]));
     }
 
-    printf("shape[%ld].vertices: %ld\n", i, shapes[i].mesh.positions.size());
-    assert((shapes[i].mesh.positions.size() % 3) == 0);
-    for (size_t v = 0; v < shapes[i].mesh.positions.size() / 3; v++) {
-      printf("  v[%ld] = (%f, %f, %f)\n", v,
-        shapes[i].mesh.positions[3*v+0],
-        shapes[i].mesh.positions[3*v+1],
-        shapes[i].mesh.positions[3*v+2]);
-    }
+    //printf("shape[%ld].vertices: %ld\n", i, shapes[i].mesh.positions.size());
+    //assert((shapes[i].mesh.positions.size() % 3) == 0);
+    //for (size_t v = 0; v < shapes[i].mesh.positions.size() / 3; v++) {
+    //  printf("  v[%ld] = (%f, %f, %f)\n", v,
+    //    static_cast<const double>(shapes[i].mesh.positions[3*v+0]),
+    //    static_cast<const double>(shapes[i].mesh.positions[3*v+1]),
+    //    static_cast<const double>(shapes[i].mesh.positions[3*v+2]));
+    //}
 
     printf("shape[%ld].num_tags: %ld\n", i, shapes[i].mesh.tags.size());
     for (size_t t = 0; t < shapes[i].mesh.tags.size(); t++) {
@@ -70,7 +105,7 @@ static void PrintInfo(const std::vector<tinyobj::shape_t>& shapes, const std::ve
       printf(" floats: [");
       for (size_t j = 0; j < shapes[i].mesh.tags[t].floatValues.size(); ++j)
       {
-          printf("%f", shapes[i].mesh.tags[t].floatValues[j]);
+          printf("%f", static_cast<const double>(shapes[i].mesh.tags[t].floatValues[j]));
           if (j < (shapes[i].mesh.tags[t].floatValues.size()-1))
           {
               printf(", ");
@@ -94,15 +129,15 @@ static void PrintInfo(const std::vector<tinyobj::shape_t>& shapes, const std::ve
 
   for (size_t i = 0; i < materials.size(); i++) {
     printf("material[%ld].name = %s\n", i, materials[i].name.c_str());
-    printf("  material.Ka = (%f, %f ,%f)\n", materials[i].ambient[0], materials[i].ambient[1], materials[i].ambient[2]);
-    printf("  material.Kd = (%f, %f ,%f)\n", materials[i].diffuse[0], materials[i].diffuse[1], materials[i].diffuse[2]);
-    printf("  material.Ks = (%f, %f ,%f)\n", materials[i].specular[0], materials[i].specular[1], materials[i].specular[2]);
-    printf("  material.Tr = (%f, %f ,%f)\n", materials[i].transmittance[0], materials[i].transmittance[1], materials[i].transmittance[2]);
-    printf("  material.Ke = (%f, %f ,%f)\n", materials[i].emission[0], materials[i].emission[1], materials[i].emission[2]);
-    printf("  material.Ns = %f\n", materials[i].shininess);
-    printf("  material.Ni = %f\n", materials[i].ior);
-    printf("  material.dissolve = %f\n", materials[i].dissolve);
-    printf("  material.illum = %d\n", materials[i].illum);
+    printf("  material.Ka = (%f, %f ,%f)\n", static_cast<const double>(materials[i].ambient[0]),       static_cast<const double>(materials[i].ambient[1]),       static_cast<const double>(materials[i].ambient[2]));
+    printf("  material.Kd = (%f, %f ,%f)\n", static_cast<const double>(materials[i].diffuse[0]),       static_cast<const double>(materials[i].diffuse[1]),       static_cast<const double>(materials[i].diffuse[2]));
+    printf("  material.Ks = (%f, %f ,%f)\n", static_cast<const double>(materials[i].specular[0]),      static_cast<const double>(materials[i].specular[1]),      static_cast<const double>(materials[i].specular[2]));
+    printf("  material.Tr = (%f, %f ,%f)\n", static_cast<const double>(materials[i].transmittance[0]), static_cast<const double>(materials[i].transmittance[1]), static_cast<const double>(materials[i].transmittance[2]));
+    printf("  material.Ke = (%f, %f ,%f)\n", static_cast<const double>(materials[i].emission[0]),      static_cast<const double>(materials[i].emission[1]),      static_cast<const double>(materials[i].emission[2]));
+    printf("  material.Ns = %f\n", static_cast<const double>(materials[i].shininess));
+    printf("  material.Ni = %f\n", static_cast<const double>(materials[i].ior));
+    printf("  material.dissolve = %f\n", static_cast<const double>(materials[i].dissolve));
+    printf("  material.illum = %d\n",  materials[i].illum);
     printf("  material.map_Ka = %s\n", materials[i].ambient_texname.c_str());
     printf("  material.map_Kd = %s\n", materials[i].diffuse_texname.c_str());
     printf("  material.map_Ks = %s\n", materials[i].specular_texname.c_str());
@@ -128,11 +163,12 @@ TestLoadObj(
 {
   std::cout << "Loading " << filename << std::endl;
 
+  tinyobj::attrib_t attrib;
   std::vector<tinyobj::shape_t> shapes;
   std::vector<tinyobj::material_t> materials;
 
   std::string err;
-  bool ret = tinyobj::LoadObj(shapes, materials, err, filename, basepath, triangulate);
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, filename, basepath, triangulate);
 
   if (!err.empty()) {
     std::cerr << err << std::endl;
@@ -143,7 +179,7 @@ TestLoadObj(
     return false;
   }
 
-  PrintInfo(shapes, materials, triangulate);
+  PrintInfo(attrib, shapes, materials, triangulate);
 
   return true;
 }
@@ -223,13 +259,13 @@ std::string matStream(
             virtual ~MaterialStringStreamReader() {}
             virtual bool operator() (
               const std::string& matId,
-              std::vector<material_t>& materials,
-              std::map<std::string, int>& matMap,
-              std::string& err)
+              std::vector<material_t>* materials,
+              std::map<std::string, int>* matMap,
+              std::string* err)
             {
                 (void)matId;
                 (void)err;
-                LoadMtl(matMap, materials, m_matSStream);
+                LoadMtl(matMap, materials, &m_matSStream);
                 return true;
             }
 
@@ -238,10 +274,11 @@ std::string matStream(
     };
 
   MaterialStringStreamReader matSSReader(matStream);
+  tinyobj::attrib_t attrib;
   std::vector<tinyobj::shape_t> shapes;
   std::vector<tinyobj::material_t> materials;
   std::string err;
-  bool ret = tinyobj::LoadObj(shapes, materials, err, objStream, matSSReader);    
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, &objStream, &matSSReader);    
   
   if (!err.empty()) {
     std::cerr << err << std::endl;
@@ -251,11 +288,42 @@ std::string matStream(
     return false;
   }
 
-  PrintInfo(shapes, materials);
+  PrintInfo(attrib, shapes, materials);
     
   return true;
 }
 
+const char* gMtlBasePath = "../models";
+
+TEST_CASE("cornell_box", "[Loader]") {
+
+    REQUIRE(true == TestLoadObj("../models/cornell_box.obj", gMtlBasePath));
+}
+
+TEST_CASE("catmark_torus_creases0", "[Loader]") {
+
+  tinyobj::attrib_t attrib;
+  std::vector<tinyobj::shape_t> shapes;
+  std::vector<tinyobj::material_t> materials;
+
+  std::string err;
+  bool ret = tinyobj::LoadObj(&attrib, &shapes, &materials, &err, "../models/catmark_torus_creases0.obj", gMtlBasePath, /*triangulate*/false);
+
+  if (!err.empty()) {
+    std::cerr << err << std::endl;
+  }
+
+  REQUIRE(true == ret);
+
+  REQUIRE(1 == shapes.size());
+  REQUIRE(8 == shapes[0].mesh.tags.size());
+}
+
+TEST_CASE("stream_load", "[Stream]") {
+    REQUIRE(true == TestStreamLoadObj());
+}
+
+#if 0
 int
 main(
   int argc,
@@ -276,3 +344,4 @@ main(
 
   return 0;
 }
+#endif

tests/vcbuild.bat

@@ -0,0 +1,4 @@
+chcp 437
+python kuroga.py config-msvc.py
+call "C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\vcvarsall.bat" x86_amd64
+ninja

wercker.yml

@@ -6,7 +6,5 @@ build:
             name: build
             code: |
                   git clone https://github.com/syoyo/orebuildenv.git
-                  chmod +x ./orebuildenv/build/linux/bin/premake4
-                  ./orebuildenv/build/linux/bin/premake4 gmake
-                  make
-                  ./test_tinyobjloader
+                  cd tests
+                  make check