In my application I keep data in L1 while doing preliminary work on it. I then need to copy this out to SDRAM as part of a larger array that I will have to access later. What I'd like to be able to do is every time I fill up the L1 data, I fire off a DMA transfer to the SDRAM, which is setup as a circular buffer so that I don't have to worry about range checking, etc, the hardware will wrap when necessary. Looking through the docs, it looks like circular buffers are only useful going from/to a peripheral, and not for memory to memory copies, because there is not a way to restart the transfer (and reset the originating address, but not the location in the circular buffer). Right now I'm trying to implement this on a Blackfin 518F. Hopefully someone can point out something I missed in the documentation.