Parallella Community

Hello all,

i run the epiphany example e-bandwidth-test on Parallella board and i get the following results:

CASE 1: using e_write()/e_read()
ARM Host --> ERAM write speed = 83.68 MB/s
ARM Host <-- ERAM read speed = 130.22 MB/s

CASE 2: using memcpy()
ARM Host <-> DRAM: Copy speed = 432.73 MB/s


Could someone help me to know why the ARM-ERAM speed is much slower than ARM-DRAM speed ,both are moving data on DRAM.
why the two operations(memcpy vs e_read/write) are so different in speeds?

Thanks a lot
Duncan