NVENC Reconstructed Frame Output

Gaming and Visualization Technologies Video Processing & Optical Flow
1

It reads "Starting SDK 12.1, NVENCODE API supports reconstructed frame output for H.264, HEVC and AV1 encode for Turing and later GPUs. "

I am now use OpenGL library for rendering and texture management.
Does reconstructed frame output still work on OpenGL ?

GLuint nv12Texture;
glGenTextures(1, &nv12Texture);
glBindTexture(GL_TEXTURE_2D, nv12Texture);

                                    // Set parameters to avoid filtering artifacts
                                    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
                                    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
                                    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
                                    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);

                                    // Allocate memory for NV12 data (Y + interleaved UV)
                                    // Height = original_height * 1.5 (e.g., 1080 → 1620 for Y + UV planes)
                                    glPixelStorei(GL_UNPACK_ALIGNMENT, 1); // Disable padding
                                    glTexImage2D(
                                        GL_TEXTURE_2D,
                                        0,
                                        GL_R8, // Single-channel (Y and UV stored sequentially)
                                        frameSize.w,
                                        frameSize.h * 3 / 2, // Total height = 1.5x original
                                        0,
                                        GL_RED,
                                        GL_UNSIGNED_BYTE,
                                        nullptr);

                                    NV_ENC_REGISTER_RESOURCE registerParams = {0};
                                    registerParams.version = NV_ENC_REGISTER_RESOURCE_VER;
                                    registerParams.resourceType = NV_ENC_INPUT_RESOURCE_TYPE_OPENGL_TEX;
                                    registerParams.width = frameSize.w;
                                    registerParams.height = frameSize.h; // Original height (not 1.5x)
                                    registerParams.pitch = frameSize.w;  // Assume no padding
                                    registerParams.bufferFormat = NV_ENC_BUFFER_FORMAT_NV12;
                                    registerParams.bufferUsage = NV_ENC_OUTPUT_RECON;

                                    NV_ENC_INPUT_RESOURCE_OPENGL_TEX texParams = {0};
                                    texParams.texture = nv12Texture;
                                    texParams.target = GL_TEXTURE_2D;
                                    registerParams.resourceToRegister = &texParams;

                                    status = pEncFn.nvEncRegisterResource(encoder, &registerParams);
                                    if (status != NV_ENC_SUCCESS)
                                    {
                                        std::cerr << "Failed to register NV12 texture!" << std::endl;
                                    }

when I set encParams.outputReconBuffer = recon_mapParams.mappedResource;

I can not run nvEncEncodePicture ,but If I delete " encParams.outputReconBuffer = recon_mapParams.mappedResource;" nvEncEncodePicture is OK ?
by the way, can you tell me how I can get PSNR better, when I use nvEncEncodePicture ?