Embedded AI
Introducing Embedded AI: revolutionizing intelligent technology with efficient, resource-conscious solutions tailored for optimal performance in even the most resource-limited environments.
The main goal is to find an optimal mapping of an algorithm to the selected embedded system.
The main steps in porting an algorithm on the embedded system are:
- Algorithm analysis (elementary processing modules, data paths, control paths, mutual dependences)
- Algorithm partitioning (grouping of modules in blocks, organization of modules inside the block, synchronization of blocks and data packets between blocks)
- Algorithm adaptation to target platform (adaptation of arithmetic to target DSP/RISC cores, error analysis)
- Performance analysis (identification of conflicts and stalls, processor load analysis, memory load analysis)