high performance computing on graphics processing units: hgpu.org

As a standard C++ programming model, SYCL has gained popularity on incorporating various parallel computing frameworks. With the development of hardware technologies, low-level computing devices are becoming increasingly varied and thus result in the great heterogeneity of hardware. Although many computing frameworks, such as OpenCL, OpenMP and CUDA, can benefit to heterogeneous computing, they increase the complexity of cross-platform deployment and reduce productivity due to low portability and miscellaneous features. By comparison, SYCL allows programmers to write high-performance parallel applications in the standard C++ syntax and execute them across vendor-specific hardware without diving into low-level technologies. However, despite the popularity of SYCL on Windows and Linux, there is little research on porting SYCL to QNX, a real-time operating system (RTOS). Therefore, we choose two SYCL implementations and conduct corresponding experiments. In particular, we build a new path of calling OpenCL APIs in SYCL-GTX and significantly reduce the time of compiling SYCL kernels. Although the overall performance of SYCL-GTX on QNX is evaluated on Linux, our experiments demonstrate that there are many possible optimizations that can improve SYCL-GTX on QNX.