This document is to introduce a detailed method for optimizing OpenCL code on the Adreno GPUs. Wealth of information provided in the documentation OpenCL to help developers understand the basic structure and Adreno, and most importantly, to master the OpenCL optimization skills.
OpenCL optimization often has a lot to try and challenging and requires trial and error. Because each vendor may have his own best practices for the same task, so read through this document, and Adreno GPUs optimization criteria and methods have in-depth understanding is very important. Many seem minor factors can have a significant impact on performance. Unfortunately, not hands-on practice and it would be difficult to solve these problems (such as found in these secondary factors).
Due to time constraints and other reasons, there are some ways not covered. Adreno GPUs support a large number of extensions that can significantly enhance the performance and added many other features. For example, recent Adreno GPU support some proprietary image format, means that obtaining the original picture and audio on a single processor from the picture, compressed into this picture format, then direct and effective treatment. These operations can save some artificial, also can improve bandwidth usage.
After the release of the document might contain more themes.