TX2 dmatest can olnly up to 215568 KB/s

We do damtest on TX2 to do mem2mem copy.
https://devtalk.nvidia.com/default/topic/1049545/jetson-tx2/how-to-using-gpc-dma-mem2mem-function/

Our test command is:

sudo insmod dmatest channel=dma1chan9 timeout=2000 iterations=1 run=1 test_buf_size=737280 verbose=1 noverify=1

Result is:

[ 1542.192028] dma1chan9-copy0: summary 1 tests, 0 failures 299 iops 215568 KB/s (0)

Seems the mem2mem performance doesn’t meet our expect.
Can we apply dma burst feature for mem2mem, and how?
What is the max mem2mem dma performance we can catch?

Please boost the system to performance mode, also boost the emc clock to max.