As a global leader in digital parallel world technology, Par**erse Parallel Cloud focuses on providing cutting-edge decentralized real-time cloud rendering products and services to enterprises and developers. Its real-time cloud rendering product, LarkXR, has been widely used in many commercial scenarios and has become the leading real-time cloud rendering solution in the market.
Par**erse Paracloud is committed to continuously improving the application effect and overall experience of LARKXR in the field of real-time cloud rendering. Since 2023, the company has continued to improve the relevant components and processes of LARKXR products, and has carried out a series of optimizations for LARKXR-related products. These include the continuous improvement of the real-time image quality analysis (IQA) function based on CUDA, the continuous improvement of XLONG, an independent transport library for cloud VR scenes, and the further improvement of the real-time cloud rendering process. These advancements not only improve the performance of LarkXR, but also lay the groundwork for the future development of Par**Erse Paracloud in terms of optimization of real-time cloud rendering technology.
Based oncudaIQA analysis
In the process of rendering and encoding and decoding 3D applications, the graphics engine and codec engine of the GPU assume the main task, but the CUDA computing engine is not fully utilized, resulting in a waste of resources. At the same time, due to the extremely high requirements for real-time performance of cloud rendering products, it is also a challenge to perceive the effect of first-class encoding in real time. LARKXR makes full use of the wasted CUDA computing resources, and completes the quality analysis of the encoded image in the GPU in a closed loop without relying on the CPU computing power, which allows users to perceive the current image quality in real time, thereby further enhancing the user-friendliness of LARKXR, and also laying a solid foundation for further optimizing the transmission control function based on the Bitrate Ladder.
Cloud-orientedvrTransmission optimization
Compared with general real-time cloud rendering scenarios, cloud VR has more challenges and has basic requirements of higher bitrate and lower latency. LarkXR has launched an independent transmission control framework for cloud VR scenarios, Xlong. XLONG breaks the hierarchical division between rendering, encoding and network at the level of transmission control, and is committed to scheduling and controlling the transmission of audio ** as a whole, which can reduce the intra-frame transmission delay as much as possible and further improve the adaptability of LARKXR in cloud VR scenarios. 
Application of xlong in XR scenarios3.Process optimization for LARKXR cloud rendering
Traditional cloud rendering products generally dynamically allocate bitrates and adjust the target bitrate of encoders based on the estimation of network bandwidth. However, the output granularity of the encoder is based on frames, and due to the dynamic change characteristics of the scene, the output bitrate and the allocated bitrate cannot be effectively adapted, resulting in frequent overshoot or undershoot phenomena of the encoder. This phenomenon can cause a series of problems, such as a steep increase in latency or inaccurate bandwidth detection. On the basis of the traditional cloud rendering process, LARKXR has established an integrated monitoring mechanism of encoding, speed queuing, transmission control and bandwidth estimation, which can carry out better target optimization based on the current transmission feedback, speed queuing status, bandwidth estimation results, and encoder output characteristics, and effectively alleviate a series of problems such as increased latency. Through a series of R&D efforts, Par**erse Paracloud has taken an important step forward in deepening the understanding of the nature of real-time cloud rendering problems, enabling more flexibility to adjust and optimize related components and processes to better adapt to different application scenarios and needs. These technological advancements not only demonstrate Par**erse's leading position in the field of real-time cloud rendering technology, but also provide new perspectives and methods for the entire industry to improve and optimize real-time cloud rendering solutions. With the continuous development and application of these technologies, we can expect higher quality and richer visual effects in the exploration and experience of digital parallel worlds.