For example, the following use-cases are difficult to develop and if one implements directly with for example OpenCL, the application will have to be optimized for the certain target hardware. What happens if the project priorities change and all of a sudden you need to port your application to a new hardware? Rocksolid Compute eliminates the need for costly and cumbersome rewriting of your code.
Rocksolid® Compute
Rocksolid Compute provides a C++ compute API. It is under development but already several libraries and implementations are available for licensing.
Rocksolid Compute will provide a full library of cross-platform accelerated math operations and reference implementations of the basic building blocks for computer vision, inferencing and various ADAS functionalities.
It is a hardware agnostic API that enables a higher level of productivity and easier code portability to new environments. The basic idea is to hide the various back ends from developers so they don’t have to care about those and can just focus on algorithm development.
- Image alignment/stitching
- Sensor fusion / Extended Kalman filter
- Least square optimization using computation graph
- SLAM (mapping and tracking)
- Inverse kinematics
Rocksolid Compute Benefits
Easy to use API to implement common compute functionality
Enables deployment of AI based applications in a wide variety of target hardware
Utilize platform capabilities in an optimal way, e.g. GPU load balancing between graphics and compute tasks