The management policies are arranged into two categories:
intra-set and inter-set remapping. By moving data of a working
set within a cache set, intra-set policies tries increasing the
write-utilization of SRAM lines and regulating other write
requests over STT-RAM lines for uniform wear-out. Then
using non-uniformity of write requests across various sets of a
cache, inter-set policies use this observation to dynamically
merge a super write-intensive set (with large write working set,
WWS) with one temporary poorly-utilized set (with small
WWS). Hence, frequently-written data from STT-RAM blocks
of the former can now move to SRAM blocks of the latter.