Remove glew dependency

README.md

@@ -19,10 +19,6 @@ For business inquiries, please contact [researchinquiries@nvidia.com](mailto:res
 
 We do not currently accept outside code contributions in the form of pull requests.
 
-[GLEW](https://github.com/nigels-com/glew) library redistributed under the 
-[Modified BSD License](http://glew.sourceforge.net/glew.txt), the 
-[Mesa 3-D License](http://glew.sourceforge.net/mesa.txt) (MIT) and the 
-[Khronos License](http://glew.sourceforge.net/khronos.txt) (MIT).
 Environment map stored as part of `samples/data/envphong.npz` is derived from a Wave Engine
 [sample material](https://github.com/WaveEngine/Samples/tree/master/Materials/EnvironmentMap/Content/Assets/CubeMap.cubemap)
 originally shared under 

docker/Dockerfile

@@ -12,7 +12,7 @@
 #
 # This file defaults to pytorch/pytorch as it works on slightly older
 # driver versions.
-ARG BASE_IMAGE=pytorch/pytorch:1.6.0-cuda10.1-cudnn7-devel
+ARG BASE_IMAGE=pytorch/pytorch:1.7.1-cuda11.0-cudnn8-devel
 FROM $BASE_IMAGE
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
@@ -44,14 +44,6 @@ ENV PYOPENGL_PLATFORM egl
 
 COPY docker/10_nvidia.json /usr/share/glvnd/egl_vendor.d/10_nvidia.json
 
-# glew installation from source
-RUN curl -L https://downloads.sourceforge.net/project/glew/glew/2.1.0/glew-2.1.0.tgz > /tmp/glew-2.1.0.tgz
-RUN mkdir -p /tmp && \
-    cd /tmp && tar zxf /tmp/glew-2.1.0.tgz && cd glew-2.1.0 && \
-    SYSTEM=linux-egl make && \
-    SYSTEM=linux-egl make install && \
-    rm -rf /tmp/glew-2.1.0.zip /tmp/glew-2.1.0
-
 RUN pip install imageio imageio-ffmpeg
 
 COPY nvdiffrast /tmp/pip/nvdiffrast/

docs/index.html

@@ -994,7 +994,7 @@ severity will be silent.</td></tr></table></div>
 <p>This work is made available under the <a href="https://github.com/NVlabs/nvdiffrast/blob/main/LICENSE.txt">Nvidia Source Code License</a>.</p>
 <p>For business inquiries, please contact <a href="mailto:researchinquiries@nvidia.com">researchinquiries@nvidia.com</a></p>
 <p>We do not currently accept outside contributions in the form of pull requests.</p>
-<p><a href="https://github.com/nigels-com/glew">GLEW</a> library redistributed under the <a href="http://glew.sourceforge.net/glew.txt">Modified BSD License</a>, the <a href="http://glew.sourceforge.net/mesa.txt">Mesa 3-D License</a> (MIT) and the <a href="http://glew.sourceforge.net/khronos.txt">Khronos License</a> (MIT). Environment map stored as part of <code>samples/data/envphong.npz</code> is derived from a Wave Engine <a href="https://github.com/WaveEngine/Samples/tree/master/Materials/EnvironmentMap/Content/Assets/CubeMap.cubemap">sample material</a> originally shared under <a href="https://github.com/WaveEngine/Samples/blob/master/LICENSE.md">MIT License</a>. Mesh and texture stored as part of <code>samples/data/earth.npz</code> are derived from <a href="https://www.turbosquid.com/3d-models/3d-realistic-earth-photorealistic-2k-1279125">3D Earth Photorealistic 2K</a> model originally made available under <a href="https://blog.turbosquid.com/turbosquid-3d-model-license/#3d-model-license">TurboSquid 3D Model License</a>.</p>
+<p>Environment map stored as part of <code>samples/data/envphong.npz</code> is derived from a Wave Engine <a href="https://github.com/WaveEngine/Samples/tree/master/Materials/EnvironmentMap/Content/Assets/CubeMap.cubemap">sample material</a> originally shared under <a href="https://github.com/WaveEngine/Samples/blob/master/LICENSE.md">MIT License</a>. Mesh and texture stored as part of <code>samples/data/earth.npz</code> are derived from <a href="https://www.turbosquid.com/3d-models/3d-realistic-earth-photorealistic-2k-1279125">3D Earth Photorealistic 2K</a> model originally made available under <a href="https://blog.turbosquid.com/turbosquid-3d-model-license/#3d-model-license">TurboSquid 3D Model License</a>.</p>
 <h2 id="citation">Citation</h2>
 <pre><code>@article{Laine2020diffrast,
   title   = {Modular Primitives for High-Performance Differentiable Rendering},

nvdiffrast/__init__.py

@@ -6,4 +6,4 @@
 # distribution of this software and related documentation without an express
 # license agreement from NVIDIA CORPORATION is strictly prohibited.
 
-__version__ = '0.2.4'
+__version__ = '0.2.5'

nvdiffrast/common/framework.h

@@ -42,7 +42,7 @@ using namespace tensorflow::shape_inference;
 #define NVDR_CTX_ARGS int _nvdr_ctx_dummy
 #define NVDR_CTX_PARAMS 0
 #define NVDR_CHECK(COND, ERR) do { TORCH_CHECK(COND, ERR) } while(0)
-#define NVDR_CHECK_CUDA_ERROR(CUDA_CALL) do { cudaError_t err = CUDA_CALL; TORCH_CHECK(err == CUDA_SUCCESS, "Cuda error: ", cudaGetLastError(), "[", #CUDA_CALL, ";]"); } while(0)
+#define NVDR_CHECK_CUDA_ERROR(CUDA_CALL) do { cudaError_t err = CUDA_CALL; TORCH_CHECK(!err, "Cuda error: ", cudaGetLastError(), "[", #CUDA_CALL, ";]"); } while(0)
 #define NVDR_CHECK_GL_ERROR(GL_CALL) do { GL_CALL; GLenum err = glGetError(); TORCH_CHECK(err == GL_NO_ERROR, "OpenGL error: ", getGLErrorString(err), "[", #GL_CALL, ";]"); } while(0)
 #endif
 

nvdiffrast/common/glutil.inl → nvdiffrast/common/glutil.cpp

@@ -6,64 +6,118 @@
 // distribution of this software and related documentation without an express
 // license agreement from NVIDIA CORPORATION is strictly prohibited.
 
-#pragma once
+//------------------------------------------------------------------------
+// Common.
+//------------------------------------------------------------------------
+
 #include "framework.h"
+#include "glutil.h"
 #include <iostream>
 #include <iomanip>
 
+// Create the function pointers.
+#define GLUTIL_EXT(return_type, name, ...) return_type (GLAPIENTRY* name)(__VA_ARGS__) = 0;
+#include "glutil_extlist.h"
+#undef GLUTIL_EXT
+
+// Track initialization status.
+static volatile bool s_glExtInitialized = false;
+
+// Error strings.
+const char* getGLErrorString(GLenum err)
+{
+    switch(err)
+    {
+        case GL_NO_ERROR:                       return "GL_NO_ERROR";
+        case GL_INVALID_ENUM:                   return "GL_INVALID_ENUM";
+        case GL_INVALID_VALUE:                  return "GL_INVALID_VALUE";
+        case GL_INVALID_OPERATION:              return "GL_INVALID_OPERATION";
+        case GL_STACK_OVERFLOW:                 return "GL_STACK_OVERFLOW";
+        case GL_STACK_UNDERFLOW:                return "GL_STACK_UNDERFLOW";
+        case GL_OUT_OF_MEMORY:                  return "GL_OUT_OF_MEMORY";
+        case GL_INVALID_FRAMEBUFFER_OPERATION:  return "GL_INVALID_FRAMEBUFFER_OPERATION";
+        case GL_TABLE_TOO_LARGE:                return "GL_TABLE_TOO_LARGE";
+        case GL_CONTEXT_LOST:                   return "GL_CONTEXT_LOST";
+    }
+    return "Unknown error";
+}
+
 //------------------------------------------------------------------------
 // Windows.
 //------------------------------------------------------------------------
 
 #ifdef _WIN32
-#define NOMINMAX
-#include <windows.h>
-#define GLEW_STATIC
-#include "../lib/glew.h"
-#include <GL/gl.h>
-#include <cuda_gl_interop.h>
 
-//------------------------------------------------------------------------
+static CRITICAL_SECTION getInitializedCriticalSection(void)
+{
+    CRITICAL_SECTION cs;
+    InitializeCriticalSection(&cs);
+    return cs;
+}
+
+static CRITICAL_SECTION s_getProcAddressMutex = getInitializedCriticalSection();
 
-struct GLContext
+static void safeGetProcAddress(const char* name, PROC* pfn)
 {
-    HDC     hdc;
-    HGLRC   hglrc;
-    int     glewInitialized;
-};
+    PROC result = wglGetProcAddress(name);
+    if (!result)
+    {
+        LeaveCriticalSection(&s_getProcAddressMutex); // Prepare for thread exit.
+        LOG(FATAL) << "wglGetProcAddress() failed for '" << name << "'";
+        exit(1); // Should never get here but make sure we exit.
+    }
+    *pfn = result;
+}
 
-//------------------------------------------------------------------------
+static void initializeGLExtensions(void)
+{
+    // Use critical section for thread safety.
+    EnterCriticalSection(&s_getProcAddressMutex);
+
+    // Only dig function pointers if not done already.
+    if (!s_glExtInitialized)
+    {
+        // Generate code to populate the function pointers.
+#define GLUTIL_EXT(return_type, name, ...) safeGetProcAddress(#name, (PROC*)&name);
+#include "glutil_extlist.h"
+#undef GLUTIL_EXT
+
+        // Mark as initialized.
+        s_glExtInitialized = true;
+    }
+
+    // Done.
+    LeaveCriticalSection(&s_getProcAddressMutex);
+    return;
+}
 
-static void setGLContext(GLContext& glctx)
+void setGLContext(GLContext& glctx)
 {
     if (!glctx.hglrc)
         LOG(FATAL) << "setGLContext() called with null gltcx";
     if (!wglMakeCurrent(glctx.hdc, glctx.hglrc))
         LOG(FATAL) << "wglMakeCurrent() failed when setting GL context";
 
-    if (glctx.glewInitialized)
+    if (glctx.extInitialized)
         return;
-    GLenum result = glewInit();
-    if (result != GLEW_OK)
-        LOG(FATAL) << "glewInit() failed, return value = " << result;
-    glctx.glewInitialized = 1;
+    initializeGLExtensions();
+    glctx.extInitialized = 1;
 }
 
-static void releaseGLContext(void)
+void releaseGLContext(void)
 {
     if (!wglMakeCurrent(NULL, NULL))
         LOG(FATAL) << "wglMakeCurrent() failed when releasing GL context";
 }
 
-extern "C" int set_gpu(const char*);
-
-static GLContext createGLContext(int cudaDeviceIdx)
+extern "C" int set_gpu(const char*); // In setgpu.lib
+GLContext createGLContext(int cudaDeviceIdx)
 {
     if (cudaDeviceIdx >= 0)
     {
         char pciBusId[256] = "";
         LOG(INFO) << "Creating GL context for Cuda device " << cudaDeviceIdx;
-        if (cudaDeviceGetPCIBusId(pciBusId, 255, cudaDeviceIdx) != CUDA_SUCCESS)
+        if (cudaDeviceGetPCIBusId(pciBusId, 255, cudaDeviceIdx))
         {
             LOG(INFO) << "PCI bus id query failed";
         }
@@ -115,7 +169,7 @@ static GLContext createGLContext(int cudaDeviceIdx)
     return glctx;
 }
 
-static void destroyGLContext(GLContext& glctx)
+void destroyGLContext(GLContext& glctx)
 {
     if (!glctx.hglrc)
         LOG(FATAL) << "destroyGLContext() called with null gltcx";
@@ -148,31 +202,44 @@ static void destroyGLContext(GLContext& glctx)
 //------------------------------------------------------------------------
 
 #ifdef __linux__
-#define GLEW_NO_GLU
-#define EGL_NO_X11 // X11/Xlib.h has "#define Status int" which breaks Tensorflow. Avoid it.
-#define MESA_EGL_NO_X11_HEADERS
-#if 1
-#   include "../lib/glew.h"    // Use local glew.h
-#else
-#   include <GL/glew.h> // Use system-supplied glew.h
-#endif
-#include <EGL/egl.h>
-#include <EGL/eglext.h>
-#include <GL/gl.h>
-#include <cuda_gl_interop.h>
 
-//------------------------------------------------------------------------
+static pthread_mutex_t s_getProcAddressMutex;
 
-struct GLContext
+typedef void (*PROCFN)();
+
+static void safeGetProcAddress(const char* name, PROCFN* pfn)
 {
-    EGLDisplay  display;
-    EGLContext  context;
-    int         glewInitialized;
-};
+    PROCFN result = eglGetProcAddress(name);
+    if (!result)
+    {
+        pthread_mutex_unlock(&s_getProcAddressMutex); // Prepare for thread exit.
+        LOG(FATAL) << "wglGetProcAddress() failed for '" << name << "'";
+        exit(1); // Should never get here but make sure we exit.
+    }
+    *pfn = result;
+}
 
-//------------------------------------------------------------------------
+static void initializeGLExtensions(void)
+{
+    pthread_mutex_lock(&s_getProcAddressMutex);
 
-static void setGLContext(GLContext& glctx)
+    // Only dig function pointers if not done already.
+    if (!s_glExtInitialized)
+    {
+        // Generate code to populate the function pointers.
+#define GLUTIL_EXT(return_type, name, ...) safeGetProcAddress(#name, (PROCFN*)&name);
+#include "glutil_extlist.h"
+#undef GLUTIL_EXT
+
+        // Mark as initialized.
+        s_glExtInitialized = true;
+    }
+
+    pthread_mutex_unlock(&s_getProcAddressMutex);
+    return;
+}
+
+void setGLContext(GLContext& glctx)
 {
     if (!glctx.context)
         LOG(FATAL) << "setGLContext() called with null gltcx";
@@ -180,16 +247,13 @@ static void setGLContext(GLContext& glctx)
     if (!eglMakeCurrent(glctx.display, EGL_NO_SURFACE, EGL_NO_SURFACE, glctx.context))
         LOG(ERROR) << "eglMakeCurrent() failed when setting GL context";
 
-    if (glctx.glewInitialized)
+    if (glctx.extInitialized)
         return;
-
-    GLenum result = glewInit();
-    if (result != GLEW_OK)
-        LOG(FATAL) << "glewInit() failed, return value = " << result;
-    glctx.glewInitialized = 1;
+    initializeGLExtensions();
+    glctx.extInitialized = 1;
 }
 
-static void releaseGLContext(void)
+void releaseGLContext(void)
 {
     EGLDisplay display = eglGetCurrentDisplay();
     if (display == EGL_NO_DISPLAY)
@@ -229,13 +293,13 @@ static EGLDisplay getCudaDisplay(int cudaDeviceIdx)
     eglQueryDevicesEXT(0, 0, &num_devices);
     if (!num_devices)
         return 0;
-    
+
     EGLDisplay display = 0;
     EGLDeviceEXT* devices = (EGLDeviceEXT*)malloc(num_devices * sizeof(void*));
     eglQueryDevicesEXT(num_devices, devices, &num_devices);
     for (int i=0; i < num_devices; i++)
     {
-        EGLDeviceEXT device = devices[i]; 
+        EGLDeviceEXT device = devices[i];
         intptr_t value = -1;
         if (eglQueryDeviceAttribEXT(device, EGL_CUDA_DEVICE_NV, &value) && value == cudaDeviceIdx)
         {
@@ -248,7 +312,7 @@ static EGLDisplay getCudaDisplay(int cudaDeviceIdx)
     return display;
 }
 
-static GLContext createGLContext(int cudaDeviceIdx)
+GLContext createGLContext(int cudaDeviceIdx)
 {
     EGLDisplay display = 0;
 
@@ -261,7 +325,7 @@ static GLContext createGLContext(int cudaDeviceIdx)
             LOG(INFO) << "Failed, falling back to default display";
     }
 
-    if (!display)  
+    if (!display)
     {
         display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
         if (display == EGL_NO_DISPLAY)
@@ -312,7 +376,7 @@ static GLContext createGLContext(int cudaDeviceIdx)
     return glctx;
 }
 
-static void destroyGLContext(GLContext& glctx)
+void destroyGLContext(GLContext& glctx)
 {
     if (!glctx.context)
         LOG(FATAL) << "destroyGLContext() called with null gltcx";
@@ -332,28 +396,8 @@ static void destroyGLContext(GLContext& glctx)
     memset(&glctx, 0, sizeof(GLContext));
 }
 
-#endif // __linux__
-
-//------------------------------------------------------------------------
-// Common.
 //------------------------------------------------------------------------
 
-static const char* getGLErrorString(GLenum err)
-{
-    switch(err)
-    {
-        case GL_NO_ERROR:                       return "GL_NO_ERROR";
-        case GL_INVALID_ENUM:                   return "GL_INVALID_ENUM";
-        case GL_INVALID_VALUE:                  return "GL_INVALID_VALUE";
-        case GL_INVALID_OPERATION:              return "GL_INVALID_OPERATION";
-        case GL_STACK_OVERFLOW:                 return "GL_STACK_OVERFLOW";
-        case GL_STACK_UNDERFLOW:                return "GL_STACK_UNDERFLOW";
-        case GL_OUT_OF_MEMORY:                  return "GL_OUT_OF_MEMORY";
-        case GL_INVALID_FRAMEBUFFER_OPERATION:  return "GL_INVALID_FRAMEBUFFER_OPERATION";
-        case GL_TABLE_TOO_LARGE:                return "GL_TABLE_TOO_LARGE";
-        case GL_CONTEXT_LOST:                   return "GL_CONTEXT_LOST";
-    }
-    return "Unknown error";
-}
+#endif // __linux__
 
 //------------------------------------------------------------------------

nvdiffrast/common/glutil.h

+// Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.
+//
+// NVIDIA CORPORATION and its licensors retain all intellectual property
+// and proprietary rights in and to this software, related documentation
+// and any modifications thereto.  Any use, reproduction, disclosure or
+// distribution of this software and related documentation without an express
+// license agreement from NVIDIA CORPORATION is strictly prohibited.
+
+#pragma once
+
+//------------------------------------------------------------------------
+// Windows-specific headers and types.
+//------------------------------------------------------------------------
+
+#ifdef _WIN32
+#define NOMINMAX
+#include <windows.h> // Required by gl.h in Windows.
+#define GLAPIENTRY APIENTRY
+
+struct GLContext
+{
+    HDC     hdc;
+    HGLRC   hglrc;
+    int     extInitialized;
+};
+
+#endif // _WIN32
+
+//------------------------------------------------------------------------
+// Linux-specific headers and types.
+//------------------------------------------------------------------------
+
+#ifdef __linux__
+#define EGL_NO_X11 // X11/Xlib.h has "#define Status int" which breaks Tensorflow. Avoid it.
+#define MESA_EGL_NO_X11_HEADERS
+#include <EGL/egl.h>
+#include <EGL/eglext.h>
+#define GLAPIENTRY
+
+struct GLContext
+{
+    EGLDisplay  display;
+    EGLContext  context;
+    int         extInitialized;
+};
+
+#endif // __linux__
+
+//------------------------------------------------------------------------
+// OpenGL, CUDA interop, GL extensions.
+//------------------------------------------------------------------------
+#define GL_GLEXT_LEGACY
+#include <GL/gl.h>
+#include <cuda_gl_interop.h>
+
+// Constants.
+#ifndef GL_VERSION_1_2
+#define GL_CLAMP_TO_EDGE                 0x812F
+#define GL_TEXTURE_3D                    0x806F
+#endif
+#ifndef GL_VERSION_1_5
+#define GL_ARRAY_BUFFER                  0x8892
+#define GL_DYNAMIC_DRAW                  0x88E8
+#define GL_ELEMENT_ARRAY_BUFFER          0x8893
+#endif
+#ifndef GL_VERSION_2_0
+#define GL_FRAGMENT_SHADER               0x8B30
+#define GL_INFO_LOG_LENGTH               0x8B84
+#define GL_LINK_STATUS                   0x8B82
+#define GL_VERTEX_SHADER                 0x8B31
+#endif
+#ifndef GL_VERSION_3_0
+#define GL_MAJOR_VERSION                 0x821B
+#define GL_MINOR_VERSION                 0x821C
+#define GL_RGBA32F                       0x8814
+#define GL_TEXTURE_2D_ARRAY              0x8C1A
+#endif
+#ifndef GL_VERSION_3_2
+#define GL_GEOMETRY_SHADER               0x8DD9
+#endif
+#ifndef GL_ARB_framebuffer_object
+#define GL_COLOR_ATTACHMENT0             0x8CE0
+#define GL_COLOR_ATTACHMENT1             0x8CE1
+#define GL_DEPTH_STENCIL                 0x84F9
+#define GL_DEPTH_STENCIL_ATTACHMENT      0x821A
+#define GL_DEPTH24_STENCIL8              0x88F0
+#define GL_FRAMEBUFFER                   0x8D40
+#define GL_INVALID_FRAMEBUFFER_OPERATION 0x0506
+#define GL_UNSIGNED_INT_24_8             0x84FA
+#endif
+#ifndef GL_ARB_imaging
+#define GL_TABLE_TOO_LARGE               0x8031
+#endif
+#ifndef GL_KHR_robustness
+#define GL_CONTEXT_LOST                  0x0507
+#endif
+
+// Declare function pointers to OpenGL extension functions.
+#define GLUTIL_EXT(return_type, name, ...) extern return_type (GLAPIENTRY* name)(__VA_ARGS__);
+#include "glutil_extlist.h"
+#undef GLUTIL_EXT
+
+//------------------------------------------------------------------------
+// Common functions.
+//------------------------------------------------------------------------
+
+void        setGLContext            (GLContext& glctx);
+void        releaseGLContext        (void);
+GLContext   createGLContext         (int cudaDeviceIdx);
+void        destroyGLContext        (GLContext& glctx);
+const char* getGLErrorString        (GLenum err);
+
+//------------------------------------------------------------------------

nvdiffrast/common/glutil_extlist.h

+// Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.
+//
+// NVIDIA CORPORATION and its licensors retain all intellectual property
+// and proprietary rights in and to this software, related documentation
+// and any modifications thereto.  Any use, reproduction, disclosure or
+// distribution of this software and related documentation without an express
+// license agreement from NVIDIA CORPORATION is strictly prohibited.
+
+#ifndef GL_VERSION_1_2
+GLUTIL_EXT(void,   glTexImage3D,                GLenum target, GLint level, GLint internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const void *pixels);
+#endif
+#ifndef GL_VERSION_1_5
+GLUTIL_EXT(void,   glBindBuffer,                GLenum target, GLuint buffer);
+GLUTIL_EXT(void,   glBufferData,                GLenum target, ptrdiff_t size, const void* data, GLenum usage);
+GLUTIL_EXT(void,   glGenBuffers,                GLsizei n, GLuint* buffers);
+#endif
+#ifndef GL_VERSION_2_0
+GLUTIL_EXT(void,   glAttachShader,              GLuint program, GLuint shader);
+GLUTIL_EXT(void,   glCompileShader,             GLuint shader);
+GLUTIL_EXT(GLuint, glCreateProgram,             void);
+GLUTIL_EXT(GLuint, glCreateShader,              GLenum type);
+GLUTIL_EXT(void,   glDrawBuffers,               GLsizei n, const GLenum* bufs);
+GLUTIL_EXT(void,   glEnableVertexAttribArray,   GLuint index);
+GLUTIL_EXT(void,   glGetProgramInfoLog,         GLuint program, GLsizei bufSize, GLsizei* length, char* infoLog);
+GLUTIL_EXT(void,   glGetProgramiv,              GLuint program, GLenum pname, GLint* param);
+GLUTIL_EXT(void,   glLinkProgram,               GLuint program);
+GLUTIL_EXT(void,   glShaderSource,              GLuint shader, GLsizei count, const char *const* string, const GLint* length);
+GLUTIL_EXT(void,   glUniform2f,                 GLint location, GLfloat v0, GLfloat v1);
+GLUTIL_EXT(void,   glUseProgram,                GLuint program);
+GLUTIL_EXT(void,   glVertexAttribPointer,       GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void* pointer);
+#endif
+#ifndef GL_VERSION_3_2
+GLUTIL_EXT(void,   glFramebufferTexture,        GLenum target, GLenum attachment, GLuint texture, GLint level);
+#endif
+#ifndef GL_ARB_framebuffer_object
+GLUTIL_EXT(void,   glBindFramebuffer,           GLenum target, GLuint framebuffer);
+GLUTIL_EXT(void,   glGenFramebuffers,           GLsizei n, GLuint* framebuffers);
+#endif
+#ifndef GL_ARB_vertex_array_object
+GLUTIL_EXT(void,   glBindVertexArray,           GLuint array);
+GLUTIL_EXT(void,   glGenVertexArrays,           GLsizei n, GLuint* arrays);
+#endif
+#ifndef GL_ARB_multi_draw_indirect
+GLUTIL_EXT(void,   glMultiDrawElementsIndirect, GLenum mode, GLenum type, const void *indirect, GLsizei primcount, GLsizei stride);
+#endif
+
+//------------------------------------------------------------------------

nvdiffrast/common/rasterize.cpp

@@ -7,7 +7,7 @@
 // license agreement from NVIDIA CORPORATION is strictly prohibited.
 
 #include "rasterize.h"
-#include "glutil.inl"
+#include "glutil.h"
 #include <vector>
 #define STRINGIFY_SHADER_SOURCE(x) #x
 

nvdiffrast/common/rasterize.h

@@ -35,10 +35,11 @@ struct RasterizeGradParams
 };
 
 //------------------------------------------------------------------------
-// The rest is for C++ compilation only when using torch.
+// Do not try to include OpenGL stuff when compiling CUDA kernels for torch.
 
-#if !defined(NVDR_TORCH) || !defined(__CUDACC__)
-#include "glutil.inl"
+#if !(defined(NVDR_TORCH) && defined(__CUDACC__))
+#include "framework.h"
+#include "glutil.h"
 
 //------------------------------------------------------------------------
 // Draw command struct used by rasterizer.
@@ -93,4 +94,4 @@ void rasterizeRender(NVDR_CTX_ARGS, RasterizeGLState& s, cudaStream_t stream, co
 void rasterizeCopyResults(NVDR_CTX_ARGS, RasterizeGLState& s, cudaStream_t stream, float** outputPtr, int width, int height, int depth);
 
 //------------------------------------------------------------------------
-#endif // !defined(NVDR_TORCH) || !defined(__CUDACC__)
+#endif // !(defined(NVDR_TORCH) && defined(__CUDACC__))

nvdiffrast/tensorflow/ops.py

@@ -18,8 +18,8 @@ from . import plugin_loader
 # OpenGL-related linker options depending on platform.
 def _get_gl_opts():
     libs = {
-        'posix': ['GL', 'GLEW'],
-        'nt':    ['gdi32', 'glew32s', 'opengl32', 'user32', 'setgpu'],
+        'posix': ['GL', 'EGL'],
+        'nt':    ['gdi32', 'opengl32', 'user32', 'setgpu'],
     }
     return ['-l' + x for x in libs[os.name]]
 

nvdiffrast/tensorflow/plugin_loader.py

@@ -156,7 +156,7 @@ def get_plugin(cuda_file, extra_nvcc_options=[]):
         compile_opts = ''
         if os.name == 'nt':
             compile_opts += '"%s"' % os.path.join(tf.sysconfig.get_lib(), 'python', '_pywrap_tensorflow_internal.lib')
-            compile_opts += ' --library-path="%s"' % (os.path.dirname(__file__) + r"\..\lib") # Find glew32s.lib during compilation.
+            compile_opts += ' --library-path="%s"' % (os.path.dirname(__file__) + r"\..\lib") # Find libraries during compilation.
         elif os.name == 'posix':
             compile_opts += '"%s"' % os.path.join(tf.sysconfig.get_lib(), 'python', '_pywrap_tensorflow_internal.so')
             compile_opts += ' --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=0\''

nvdiffrast/tensorflow/tf_all.cu

@@ -15,6 +15,7 @@
 
 #define __CUDA_INCLUDE_COMPILER_INTERNAL_HEADERS__
 #include "../common/framework.h"
+#include "../common/glutil.cpp"
 
 #include "../common/common.h"
 #include "../common/common.cpp"

nvdiffrast/torch/ops.py

@@ -45,14 +45,15 @@ def _get_plugin():
 
     # Linker options.
     if os.name == 'posix':
-        ldflags = ['-lGL', '-lGLEW', '-lEGL']
+        ldflags = ['-lGL', '-lEGL']
     elif os.name == 'nt':
-        libs = ['gdi32', 'glew32s', 'opengl32', 'user32', 'setgpu']
+        libs = ['gdi32', 'opengl32', 'user32', 'setgpu']
         ldflags = ['/LIBPATH:' + lib_dir] + ['/DEFAULTLIB:' + x for x in libs]
 
     # List of source files.
     source_files = [
         '../common/common.cpp',
+        '../common/glutil.cpp',
         '../common/rasterize.cu',
         '../common/rasterize.cpp',
         '../common/interpolate.cu',