SPRUIL1D May 2019 – December 2024 DRA829J , DRA829J-Q1 , DRA829V , DRA829V-Q1 , TDA4VM , TDA4VM-Q1
The SDE core uses the HWA1 node schedulers in the DMPAC HTS.
The scheduler implements 2 consumer sockets:
The scheduler implements 1 producer socket:
Before the start of the stereo disparity processing, the first 16 image lines of both reference and base frame must be aggregated to start the processing of the first row of blocks and then 8 additional image lines in steady state for the subsequent row of blocks. To enable this, the reference and base frame read DMA producer scheduler implements transaction aggregator to aggregate enough number of image lines to start the optical flow operation.
The SDE inherently operates on blocks and reads the inputs and writes the outputs in terms of blocks. The HTS manages the stereo triggers at block level. While the output DMA for stereo disparity happens at block level, the inputs from DDR are read at line level. To handle the difference in pattern between DMA and compute on the input side, there is a pattern adapter implemented inside the reference and base frame read DMA producer scheduler which adapts the lines based growing window fetched from UTC into several blocks to be computed by the SDE. This pattern adaptation is just a way to handle the start of the stereo disparity engine, it does not alter the data organization of these input buffers in SL2.