Currently, the aarch64 port simply implements this stub routine by calling StubRoutines::_jbyte_arraycopy. Just like x86 & sparc port does, we can make this faster by checking the alignment of the src, dest, and count, and call some better arraycopy stubs, (say StubRoutines::_jshort_arraycopy). I see some 1%+ performance improvement on Spark Terasort benchmark by doing this.

BTW: we are missing _unsafe_arraycopy and _generic_arraycopy stubs for jdk8u, I would like to backport them from jdk9 after this improvement is approved and merged.