Today, many big data applications require massively parallel tasks to compute complicated mathematical operations. To perform parallel tasks, platforms like CUDA (Compute Unified Device Architecture) and OpenCL (Open Computing Language) are widely used and developed to enhance the throughput of massively parallel tasks. There is also a need for high-level abstractions and platform-independence over those massively parallel computing platforms. Recently, Khronos group announced SYCL (C++ Single-source Heterogeneous Programming for OpenCL), a new cross-platform abstraction layer, to provide an efficient way for single-source heterogeneous computing, with C++-template-level abstractions. However, since there has been no official implementation of SYCL, we currently have several different implementations from various vendors. In this paper, we analyse the characteristics of those SYCL implementations. We also show performance measures of those SYCL implementations, especially for well-known massively parallel tasks. We show that each implementation has its own strength in computing different types of mathematical operations, along with different sizes of data. Our analysis is available for fundamental measurements of the abstract-level cost-effective use of massively parallel computations, especially for big-data applications.
Recently, performance of mobile device's processor has advanced dramatically. There are many researches to utilize the processor computing ability to maximize application performance. In the field of augmented reality (AR), video stream processing is an application that requires large amount of computing powers, from the real-time processing point of view. However, due to nature of mobile platforms, unnecessary overheads should be minimized to acquire higher performance. In this paper, we introduce OpenCL based video frame segmentation to minimize unnecessary overhead in video processing applications, with dynamically changing the local work group size. Our prototype system show at most 2.3 times speed-up compared to well-known, de-facto image processing library OpenCV.
The LiDAR technology is a data capturing tool that measures distances by illuminating the targets with laser lights. Many geometric applications use this LiDAR technology to build terrains and 3D models of buildings, cultural heritages and variety of objects. Efficiently handling point clouds from the LiDAR scanners is critical for data handling. In this paper, we present a LiDAR-based terrain editing tools, with height maps and texture maps to address all the requirements. Its performance and efficiency gains are also presented.