Move data fast inside memory

There are several DMAs in QSys. The "normal" DMA, the Scatter-Gather DMA and modular sgdma (http://www.alterawiki.com/wiki/modular_sgdma).

If you want to make your own DMA, you need at least one Avalon MM slave for control and one Avalon MM master for DMA-ing. I suggest you have a look at modular SGMDA, just to get the basic idea how does DMA work. Modular SGDMA is optimized for speed, so the implementation is quite complicated (with Avalon streaming bus and FIFOs), your DMA does not need them (at least at the beginning).