Modern graphics solutions consist of many relatively simple computational units, allowing them to process non-image data. Using the GPU for general-purpose computation has been called GPGPU . One of the most well-known implementations of this concept is the framework OpenCL, which is supported by a wide range of well-known companies.
In particular, support for OpenCL in Mali family of graphics solutions company added ARM. The marketing director of ARM, said that support for GPGPU can not only improve the performance of single-chip systems, but also make them cheaper.
For example, the computing units of the graphics core can be used to process data from an optical camera sensor. Usually used for this purpose a special processor, which, by the way, not only complicates the design of SoC, but also increases the cost and power consumption.
The only question that remains unanswered - when GPGPU will be implemented for mobile devices at the software level.