ARM裸机 - 时钟系统_arm soc clk ctrl

Working on firmware which is a pure bare metal code. use case is only compute task on this SoC. Using this code my task is to run a job or kernel on GPU for example matrix multiplication. Looking out for the architecture details that needs to be looked into for developing this kind of a usecase where job or kernel is running on a baremetal. Do capture the details on how the memory allocated from CPU can be passed to GPU, what kind of page table related configurations need to be done and then configure on GPU configuration registers side, how to trigger a job on GPU etc. Cover all the elements needed in this flow right from start to end.