Hi there, I am writing a custom video decoding app that process custom stream JPEG frames from USB, without any standard on a STM32H747.
So far, (more or less) so good, I use CM4 for USB operations, and CM7 for JPEG decoding and DMA2D pixel conversion. However, I need three buffers on SDRAM for that operations, one for USB Packets, another for JPEG decoding and a third one for VRAM.
I am wondering if there is a possibility to avoid one of them by transferring data directly from JPEG decoding IP to DMA2D without CPU intervention. That would save me some throughput on FMC bottleneck and release the CPU (indeed it is not very busy).
However, I have no clue on which DMA to use (MDMA or DMA) and what should be the role of each of the peers (should DMA2D pull data from JPEG? or should JPEG push data to DMA2D?).
I did not find any clear explanation or sample about this. Does someone have a hint on that?
TiA
Juan
