论文信息 - OpenCL Optimization and Best Practices for Qualcomm Adreno GPUs

OpenCL Optimization and Best Practices for Qualcomm Adreno GPUs

As the industry's leading mobile graphics processing unit (GPU) core, Adreno™ in Qualcomm®'s Snapdragon™ SOCs has supported the OpenCL™ standard since its A3x family and all through its A4x, A5x families, and the latest A6x family. How to effectively program and optimize OpenCL applications on Adreno OpenCL is of great interest for many OEMs as well as 3rd party app developers. This paper provides a high level overview of Adreno's compute architecture, introduces Adreno's OpenCL support and general guidance and good practices on programming, optimization and profiling, and illustrates how to apply them and achieve good performance through two use case studies.

[1] Guohui Wang,et al. OpenCL-Based Mobile GPGPU Benchmarking: Methods and Challenges , 2016, IWOCL.

[2] David Kaeli,et al. Heterogeneous Computing with OpenCL 2.0 , 2015 .

[3] David A. Patterson,et al. Computer Architecture, Fifth Edition: A Quantitative Approach , 2011 .

[4] David A. Patterson,et al. Computer Architecture: A Quantitative Approach , 1969 .

[5] Joseph R. Cavallaro,et al. Accelerating computer vision algorithms using OpenCL framework on the mobile GPU - A case study , 2013, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing.

[6] Jian Sun,et al. Guided Image Filtering , 2010, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7] Amit Jindal,et al. Symphony: Task Scheduling and Memory Management in Heterogeneous Computing , 2017, IWOCL.

[8] Carsten Rother,et al. Fast cost-volume filtering for visual correspondence and beyond , 2011, CVPR 2011.

[9] Timothy G. Mattson,et al. OpenCL Programming Guide , 2011 .

[10] David A. Patterson,et al. Computer Architecture - A Quantitative Approach, 5th Edition , 1996 .