I am trying to setup an image algorithm on BF537 including multiple successive image processing steps. Each step consists of a blockwise processing on the image, hence every time the image is split up into smaller blocks and moved by the DMA to L1 to be processed - for every single step. To make the algorithm work with a high performance, the goal should be to give every single algorithm the chance to use all of the L1 Memory. This results in a small number of blocks to be processed, because they just can be larger. This means that variables for every single part of the algorithm, that are used in L1 should map to the same Memory Space.
How can this be achieved? I guess that one has to be able to generate several sections in the LDF, which all map to the same space, more precisely to L1. Can anybody help?
Thank you, Michel F.