OpenGL + CUDA互操作 - 图像不显示在窗口中 [英] OpenGL + CUDA interop - image not displaying in window
本文介绍了OpenGL + CUDA互操作 - 图像不显示在窗口中的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！
问题描述
背景：我使用OpenCV从磁盘读取图像，使用CUDA将其传递到GPU，现在，我试图让OpenGL渲染图像。
Background: I read an image from disk using OpenCV, passed it to the GPU using CUDA, and now, I am trying to get OpenGL to render the image.
我没有在这里使用GLUT，因为我编译我的代码，并得到32位Windows创建一个新窗口，我将渲染图像。现在，当我将 flipped.data 传递给 glTexImage2D（）函数。但是，当我使用CUDA + OpenGL时，没有渲染同一个图像。
I am not using GLUT here because I compile my code and get 32-bit Windows to create a new window, inside which I will render the image. Now, I flipped the OpenCV image and got OpenGL to render the image nicely when I simply passed flipped.data to the glTexImage2D() function. However, the same image is not being rendered when I use CUDA + OpenGL.
我的实际图片大于目前的图片。我使用OpenGL像素缓冲对象，和OpenGL纹理来渲染图像。使用纹理允许我指定要显示的图像的部分。我的灰度图像具有尺寸 w1024 x h256 ，它具有8位深度（unsigned char / GL_UNSIGNED_BYTE）。
My actual images are bigger than the current one. I am using the OpenGL pixel buffer object, and the OpenGL texture to render the image. Utilizing the texture allows me to specify the part of the image I want to display. My grayscale image has dimensions w1024xh256, and it has an 8-bit depth (unsigned char/GL_UNSIGNED_BYTE).
问题：我无法弄清楚我的代码出了什么问题。我试图仔细遵循CUDA C编程指南，并注册/映射CUDA资源与PBO和纹理以及实际输入数据。由于我的输入图像数据来自OpenCV，我简单地将将的数据复制到设备指针 dev_inp 。 I（正确？）使用 cudaGraphicsResourceGetMappedPointer（）将 dev_inp 映射到CUDA资源。然而，窗口不显示任何东西，并保持黑色。没有视口更改，我在 glBegin（）.. glEnd（）指定的坐标是正确的，因为它们正确映射 flipped 的数据到纹理。
Question: I can't quite figure out what is going wrong in my code. I tried to carefully follow the CUDA C programming guide, and register/map the CUDA resource with the PBO and the texture as well as with the actual input data. Since my input image data comes from OpenCV, I simply copied flipped's data into the device pointer dev_inp. I (correctly?) mapped the dev_inp to the CUDA resource using cudaGraphicsResourceGetMappedPointer() as well. Yet, the window does not display anything, and remains black. There are no viewport changes, and the coordinates that I specify at glBegin().. glEnd() are correct as they properly map flipped's data to the texture.
我在这里缺少其他东西吗？我是否将CUDA资源不正确地映射到PBO或设备指针？
OpenGL + CUDA互操作部分：此部分特别是在我的代码中只有CUDA + OpenGL互操作。从 WindProc（）方法调用函数 DrawOpenGLScene（）
OpenGL + CUDA interop portion: This portion is specifically only the CUDA + OpenGL interoperation in my code. The function DrawOpenGLScene() is called from the WindProc() method.
void DrawOpenGLScene()
{
initCUDADevice();
Mat image, flipped;
image = imread("K:/Ultrasound experiment images/PA_160.png", CV_LOAD_IMAGE_GRAYSCALE); // Read the file from disk
if(!image.data) // Check for invalid input
{
cout << "Could not open or find the image" << std::endl ;
}
cv::flip(image, flipped, 0);
imshow("flip", image); // displays output
//cout << "depth: " << flipped.depth() << endl;
// ===================================================================================
// opengl setup
// first, the context was created
// now, clear the window with the rendering context
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glLoadIdentity();
// ====================================================================================
// generate the pixel buffer object (PBO)
// Generate a buffer ID called a PBO (Pixel Buffer Object)
glGenBuffers(1, &pbo);
// Make this the current UNPACK buffer (OpenGL is state-based)
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo);
// Allocate data for the buffer. 4-channel 8-bit image
glBufferData(GL_PIXEL_UNPACK_BUFFER, sizeof(unsigned char) * flipped.rows * flipped.cols, NULL, GL_DYNAMIC_COPY);
//gpuErrchk(cudaGLRegisterBufferObject( pbo ));
gpuErrchk(cudaGraphicsGLRegisterBuffer(&cuda_resource, pbo, cudaGraphicsMapFlagsNone));
// ====================================================================================
// create the texture object
// enable 2D texturing
glEnable(GL_TEXTURE_2D);
// bind the texture
glGenTextures(1, &tex);
glBindTexture(GL_TEXTURE_2D, tex);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
//glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
// put flipped.data at the end, and it'll work for normal texturing
glTexImage2D(GL_TEXTURE_2D, 0, GL_LUMINANCE, image.cols, image.rows, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE, NULL);
// put tex at the end, and it'll work for normal texturing
glBindTexture(GL_TEXTURE_2D, 0);
// ====================================================================================
// copy data from openCV
unsigned char *dev_inp;
gpuErrchk( cudaMalloc((void**)&dev_inp, sizeof(unsigned char)*flipped.rows*flipped.cols) );
//cudaGLMapBufferObject((void**)dev_inp, pbo);
gpuErrchk( cudaGraphicsMapResources(1, &cuda_resource, 0) );
size_t size; // = sizeof(unsigned char)*flipped.rows*flipped.cols;
gpuErrchk( cudaGraphicsResourceGetMappedPointer((void **)&dev_inp, &size, cuda_resource) );
gpuErrchk( cudaMemcpy(dev_inp, flipped.data, sizeof(unsigned char)*flipped.rows*flipped.cols, cudaMemcpyHostToDevice) );
//cudaGLUnmapBufferObject(pbo);
gpuErrchk( cudaGraphicsUnmapResources(1, &cuda_resource, 0) );
// ====================================================================================
// bind pbo and texture to render data now
glBindBuffer( GL_PIXEL_UNPACK_BUFFER, pbo);
glBindTexture(GL_TEXTURE_2D, tex);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, flipped.cols, flipped.rows, GL_LUMINANCE, GL_UNSIGNED_BYTE, NULL);
glBegin(GL_QUADS);
glTexCoord2f(0.0f, 0.0f); glVertex3f(-1.0f, -1.0f, 1.0f); // Bottom Left Of The Texture and Quad
glTexCoord2f(1.0f, 0.0f); glVertex3f( 1.0f, -1.0f, 1.0f); // Bottom Right Of The Texture and Quad
glTexCoord2f(1.0f, 1.0f); glVertex3f( 1.0f, 1.0f, 1.0f); // Top Right Of The Texture and Quad
glTexCoord2f(0.0f, 1.0f); glVertex3f(-1.0f, 1.0f, 1.0f); // Top Left Of The Texture and Quad
glEnd();
glFlush(); // force rendering to happen
//glBindTexture(GL_TEXTURE_2D, 0);
}
整个代码 p>
Entire code:
LRESULT CALLBACK WndProc(HWND, UINT, WPARAM, LPARAM);
void DrawOpenGLScene(void);
HGLRC SetUpOpenGLContext(HWND hWnd);
GLuint tex;
GLuint pbo;
struct cudaGraphicsResource *cuda_resource;
int WINAPI WinMain (HINSTANCE hInstance, HINSTANCE hPrevInstance,
LPSTR lpszCmdLine, int nCmdShow)
{
static char szClassName[] = "Myclass";
static char szTitle[]="A Simple Win32 API OpenGL Program";
WNDCLASS wc;
MSG msg;
HWND hWnd;
wc.style = CS_HREDRAW | CS_VREDRAW;
wc.lpfnWndProc = (WNDPROC)WndProc;
wc.cbClsExtra = 0;
wc.cbWndExtra = 0;
wc.hInstance = hInstance;
wc.hIcon = NULL;
wc.hCursor = LoadCursor(NULL, IDC_ARROW);
wc.hbrBackground = (HBRUSH)GetStockObject (BLACK_BRUSH);
wc.lpszMenuName = NULL;
wc.lpszClassName = szClassName;
if (!RegisterClass (&wc))
return 0;
hWnd = CreateWindow(szClassName, szTitle,
WS_OVERLAPPEDWINDOW |
// NEED THESE for OpenGL calls to work!
WS_CLIPCHILDREN | WS_CLIPSIBLINGS,
0, 0, 1024, 256,
NULL, NULL, hInstance, NULL);
ShowWindow(hWnd, nCmdShow);
UpdateWindow( hWnd );
while (GetMessage(&msg, NULL, 0, 0))
{
TranslateMessage( &msg );
DispatchMessage( &msg );
}
return(msg.wParam);
}
LRESULT CALLBACK WndProc( HWND hWnd, UINT msg,
WPARAM wParam, LPARAM lParam )
{
HDC hDC;
static HGLRC hRC; // Note this is STATIC!
PAINTSTRUCT ps;
switch (msg)
{
case WM_CREATE:
// Select a pixel format and create a rendering context
hRC = SetUpOpenGLContext(hWnd);
break;
case WM_PAINT:
// Draw the scene
// Get a DC, make RC current & associate it with this DC
hDC = BeginPaint(hWnd, &ps);
wglMakeCurrent(hDC, hRC);
DrawOpenGLScene(); // Draw
// We're done with the RC, so deselect it
wglMakeCurrent(NULL, NULL);
EndPaint(hWnd, &ps);
break;
case WM_DESTROY:
//cudaGLUnregisterBufferObject(pbo);
cudaGraphicsUnregisterResource(cuda_resource);
// Clean up and terminate
wglDeleteContext(hRC);
PostQuitMessage(0);
break;
default:
return DefWindowProc(hWnd, msg, wParam, lParam);
}
return (0);
}
//*******************************************************
// SetUpOpenGL sets the pixel format and a rendering
// context then returns the RC
//*******************************************************
HGLRC SetUpOpenGLContext(HWND hWnd)
{
static PIXELFORMATDESCRIPTOR pfd = {
sizeof (PIXELFORMATDESCRIPTOR), // strcut size
1, // Version number
PFD_DRAW_TO_WINDOW | // Flags, draw to a window,
PFD_SUPPORT_OPENGL, // use OpenGL
PFD_TYPE_RGBA, // RGBA pixel values
24, // 24-bit color
0, 0, 0, // RGB bits & shift sizes.
0, 0, 0, // Don't care about them
0, 0, // No alpha buffer info
0, 0, 0, 0, 0, // No accumulation buffer
32, // 32-bit depth buffer
0, // No stencil buffer
0, // No auxiliary buffers
PFD_MAIN_PLANE, // Layer type
0, // Reserved (must be 0)
0, // No layer mask
0, // No visible mask
0 // No damage mask
};
int nMyPixelFormatID;
HDC hDC;
HGLRC hRC;
hDC = GetDC(hWnd);
nMyPixelFormatID = ChoosePixelFormat(hDC, &pfd);
SetPixelFormat(hDC, nMyPixelFormatID, &pfd);
hRC = wglCreateContext(hDC);
ReleaseDC(hWnd, hDC);
return hRC;
}
//***********************************************************
// initCUDADevice uses CUDA commands to initiate the CUDA
// enabled graphics card. This is prior to resource mapping,
// and rendering.
//***********************************************************
void initCUDADevice() {
gpuErrchk(cudaGLSetGLDevice( cutGetMaxGflopsDeviceId() ));
}
//********************************************************
// DrawOpenGLScene uses OpenGL commands to draw the scene
// This is where we put the OpenGL drawing commands
//********************************************************
void DrawOpenGLScene()
{
initCUDADevice();
Mat image, flipped;
image = imread("K:/Ultrasound experiment images/PA_160.png", CV_LOAD_IMAGE_GRAYSCALE); // Read the file from disk
if(!image.data) // Check for invalid input
{
cout << "Could not open or find the image" << std::endl ;
}
cv::flip(image, flipped, 0);
imshow("flip", image); // displays output
//cout << "depth: " << flipped.depth() << endl;
// ===================================================================================
// opengl setup
// first, the context was created
// now, clear the window with the rendering context
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glLoadIdentity();
// ====================================================================================
// generate the pixel buffer object (PBO)
// Generate a buffer ID called a PBO (Pixel Buffer Object)
glGenBuffers(1, &pbo);
// Make this the current UNPACK buffer (OpenGL is state-based)
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo);
// Allocate data for the buffer. 4-channel 8-bit image
glBufferData(GL_PIXEL_UNPACK_BUFFER, sizeof(unsigned char) * flipped.rows * flipped.cols, NULL, GL_DYNAMIC_COPY);
//gpuErrchk(cudaGLRegisterBufferObject( pbo ));
gpuErrchk(cudaGraphicsGLRegisterBuffer(&cuda_resource, pbo, cudaGraphicsMapFlagsNone));
// ====================================================================================
// create the texture object
// enable 2D texturing
glEnable(GL_TEXTURE_2D);
// bind the texture
glGenTextures(1, &tex);
glBindTexture(GL_TEXTURE_2D, tex);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
//glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
// put flipped.data at the end, and it'll work for normal texturing
glTexImage2D(GL_TEXTURE_2D, 0, GL_LUMINANCE, image.cols, image.rows, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE, NULL);
// put tex at the end, and it'll work for normal texturing
glBindTexture(GL_TEXTURE_2D, 0);
// ====================================================================================
// copy data from openCV
unsigned char *dev_inp;
gpuErrchk( cudaMalloc((void**)&dev_inp, sizeof(unsigned char)*flipped.rows*flipped.cols) );
//cudaGLMapBufferObject((void**)dev_inp, pbo);
gpuErrchk( cudaGraphicsMapResources(1, &cuda_resource, 0) );
size_t size; // = sizeof(unsigned char)*flipped.rows*flipped.cols;
gpuErrchk( cudaGraphicsResourceGetMappedPointer((void **)&dev_inp, &size, cuda_resource) );
gpuErrchk( cudaMemcpy(dev_inp, flipped.data, sizeof(unsigned char)*flipped.rows*flipped.cols, cudaMemcpyHostToDevice) );
//cudaGLUnmapBufferObject(pbo);
gpuErrchk( cudaGraphicsUnmapResources(1, &cuda_resource, 0) );
// ====================================================================================
// bind pbo and texture to render data now
glBindBuffer( GL_PIXEL_UNPACK_BUFFER, pbo);
glBindTexture(GL_TEXTURE_2D, tex);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, flipped.cols, flipped.rows, GL_LUMINANCE, GL_UNSIGNED_BYTE, NULL);
glBegin(GL_QUADS);
glTexCoord2f(0.0f, 0.0f); glVertex3f(-1.0f, -1.0f, 1.0f); // Bottom Left Of The Texture and Quad
glTexCoord2f(1.0f, 0.0f); glVertex3f( 1.0f, -1.0f, 1.0f); // Bottom Right Of The Texture and Quad
glTexCoord2f(1.0f, 1.0f); glVertex3f( 1.0f, 1.0f, 1.0f); // Top Right Of The Texture and Quad
glTexCoord2f(0.0f, 1.0f); glVertex3f(-1.0f, 1.0f, 1.0f); // Top Left Of The Texture and Quad
glEnd();
glFlush(); // force rendering to happen
//glBindTexture(GL_TEXTURE_2D, 0);
}
推荐答案
别人跑进同样的问题，这个线程可以帮助他们。
我通过改变 DrawOpenGLScene（）中的几个调用来解决我的问题。
In case someone else runs into the same problem, this thread can serve to help them.
I solved my problem by changing only a couple of calls in DrawOpenGLScene().
事实证明， cudaGraphicsResourceGetMappedPointer（）返回一个指向并从OpenGL PBO派生的指针，在 dev_inp 中。它内部为 dev_inp 分配 size = sizeof（unsigned char）* flipped.rows * flipped.cols 以前建立的对 glBufferData（）和 cudaGraphicsGLRegisterBuffer（）的调用。
It turns out that cudaGraphicsResourceGetMappedPointer() returns a pointer to and derived from the OpenGL PBO, and places that pointer in dev_inp. It internally allocates size = sizeof(unsigned char) * flipped.rows * flipped.cols memory for the dev_inp based on the previously established calls to glBufferData() and cudaGraphicsGLRegisterBuffer().
完成后，我之前使用 cudaMalloc（）分配的内存现在不再存在，因为它被 cudaGraphicsResourceGetMappedPointer（）的调用覆盖，将指针放在 dev_inp 中。删除 cudaMalloc（）和 cudaFree（）允许程序按原计划运行。
Once this is done, the memory that I had previously allocated using cudaMalloc() now ceases to exist because it is overwritten by the call to cudaGraphicsResourceGetMappedPointer() that places the pointer in dev_inp. Removing the cudaMalloc() and cudaFree() allowed the program to run as originally intended.
为了释放内存，应该释放PBO，因为OpenGL是内存的所有者，CUDA只是共享对OpenGL拥有的内存的访问。
In order to deallocate the memory, one should deallocate the PBO as OpenGL is the "owner" of the memory, and CUDA just shares access to the memory owned by OpenGL.
修改的 DrawOpenGLScene（）例程贴在下面：
#define GET_PROC_ADDRESS( str ) wglGetProcAddress( str )
PFNGLBINDBUFFERARBPROC glBindBuffer = NULL;
PFNGLDELETEBUFFERSARBPROC glDeleteBuffers = NULL;
PFNGLGENBUFFERSARBPROC glGenBuffers = NULL;
PFNGLBUFFERDATAARBPROC glBufferData = NULL;
void initCUDADevice() {
gpuErrchk(cudaGLSetGLDevice( cutGetMaxGflopsDeviceId() ));
}
//********************************************************
// DrawOpenGLScene uses OpenGL commands to draw the scene
// This is where we put the OpenGL drawing commands
//********************************************************
void DrawOpenGLScene()
{
// Clear Color and Depth Buffers
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Reset transformations
glLoadIdentity();
// ====================================================================================
// initiate GPU by setting it correctly
initCUDADevice();
// ====================================================================================
// read the image that needs to be textured
Mat image, flipped;
image = imread("K:/OCT experiment images/PA_175.png", CV_LOAD_IMAGE_GRAYSCALE); // Read the file from disk
if(!image.data) // Check for invalid input
{
cout << "Could not open or find the image" << std::endl ;
}
cv::flip(image, flipped, 0);
imshow("OpenCV - image", image); // displays output
// ====================================================================================
// allocate the PBO, texture, and CUDA resource
glBindBuffer = (PFNGLBINDBUFFERARBPROC)GET_PROC_ADDRESS("glBindBuffer");
glDeleteBuffers = (PFNGLDELETEBUFFERSARBPROC)GET_PROC_ADDRESS("glDeleteBuffers");
glGenBuffers = (PFNGLGENBUFFERSARBPROC)GET_PROC_ADDRESS("glGenBuffers");
glBufferData = (PFNGLBUFFERDATAARBPROC)GET_PROC_ADDRESS("glBufferData");
// ====================================================================================
// generate the pixel buffer object (PBO)
// Generate a buffer ID called a PBO (Pixel Buffer Object)
glGenBuffers(1, &pbo);
// Make this the current UNPACK buffer (OpenGL is state-based)
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo);
// Allocate data for the buffer. 4-channel 8-bit image
glBufferData(GL_PIXEL_UNPACK_BUFFER, sizeof(unsigned char) * flipped.rows * flipped.cols, NULL, GL_STREAM_DRAW);
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
gpuErrchk(cudaGraphicsGLRegisterBuffer(&cuda_resource, pbo, cudaGraphicsMapFlagsNone));
// ====================================================================================
// create the texture object
// enable 2D texturing
glEnable(GL_TEXTURE_2D);
// generate and bind the texture
glGenTextures(1, &tex);
glBindTexture(GL_TEXTURE_2D, tex);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
// put flipped.data at the end for cpu rendering
glTexImage2D(GL_TEXTURE_2D, 0, GL_LUMINANCE, image.cols, image.rows, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE, 0 );
// put tex at the end for cpu rendering
glBindTexture(GL_TEXTURE_2D, 0);
// ====================================================================================
// copy OpenCV flipped image data into the device pointer
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
unsigned char *dev_inp;
//gpuErrchk( cudaMalloc((void**)&dev_inp, sizeof(unsigned char)*flipped.rows*flipped.cols) );
gpuErrchk( cudaGraphicsMapResources(1, &cuda_resource, 0) );
size_t size;
gpuErrchk( cudaGraphicsResourceGetMappedPointer((void **)&dev_inp, &size, cuda_resource) );
gpuErrchk( cudaMemcpy(dev_inp, flipped.data, sizeof(unsigned char)*flipped.rows*flipped.cols, cudaMemcpyHostToDevice) );
gpuErrchk( cudaGraphicsUnmapResources(1, &cuda_resource, 0) );
// ====================================================================================
// bind pbo and texture to render data now
glBindBuffer( GL_PIXEL_UNPACK_BUFFER, pbo);
//
glBindTexture(GL_TEXTURE_2D, tex);
glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, flipped.cols, flipped.rows, GL_LUMINANCE, GL_UNSIGNED_BYTE, NULL);
gpuErrchk( cudaGraphicsUnregisterResource(cuda_resource));
gpuErrchk( cudaThreadSynchronize());
//gpuErrchk(cudaFree(dev_inp));
// ====================================================================================
// map the texture coords to the vertex coords
glBegin(GL_QUADS);
// Front Face
glTexCoord2f(0.0f, 0.0f); glVertex3f(-1.0f, -1.0f, 1.0f); // Bottom Left Of The Texture and Quad
glTexCoord2f(1.0f, 0.0f); glVertex3f( 1.0f, -1.0f, 1.0f); // Bottom Right Of The Texture and Quad
glTexCoord2f(1.0f, 1.0f); glVertex3f( 1.0f, 1.0f, 1.0f); // Top Right Of The Texture and Quad
glTexCoord2f(0.0f, 1.0f); glVertex3f(-1.0f, 1.0f, 1.0f); // Top Left Of The Texture and Quad
glEnd();
glFlush(); // force rendering
glDisable(GL_TEXTURE_2D);
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
glDeleteBuffers(1, &pbo);
这篇关于OpenGL + CUDA互操作 - 图像不显示在窗口中的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
查看全文
