L1/L2 cache miss hit ratio for AM335x

That will depend on the spacial/time locality of your code.

There are measurements to check memory read latency based on lmbench benchmarks http://www.bitmover.com/lmbench/lat_mem_rd.8.html

On am335x, a stride of 128 and an array size of 31K gives an average latency of 3ns. Increasing the array size to 1000K gives an avarage of 221ns.