Teensy 3.1 DMA memcpy() 1092mbs 16-byte aligned DMA
Teensy 3.1 DMA memcpy, proof of concept
2048 bytes aligned 16 loop set 215.58 mbs 76 us loop copy 910.22 mbs 18 us memset 1365.33 mbs 12 us memcpy 910.22 mbs 18 us memcpy128 1092.27 mbs 15 us DMA errs 0
Obviously, the ARMCC memcpy() is quite fast (unrolled assembler I presume).
You could add IRQ handler if you wanted asynch operations.