> What do you think the reason we are seeing maximum of 8KB on
> copyin\out although we are telling dd to use 128KB block size. we are
> reading from SATA disk (SCSI implementation) or USB disk .

It's because block I/O scheme does.

And I'm anxious that you're pursuing the wrong direction to enhance
I/O performance by adding DMA HW for copyin/out.  You will find the
reason as deeper you learn.

Toru Nishimura/ALKYL Technology