
186 Energy-Aware Memory Management for EMSs
BB1 BB2 BB3
a [ ]
VWR3
b [ ]
h [ ] f [ ]
g [ ]
VWR1
c [ ]
d [ ]
VWR2
d [ ]
b [ ]
a [ ]
a [ ]
Figure 5.14 VWR COAT domain for Example 5.10 using adapted iVMX
Algorithm 5.5.
3. Indirection logic of iVMX and its corresponding overhead in the algo-
rithm has been removed.
4. Only array variables are considered for VWR allocation (unlike all vari-
ables in [67]).
5. An extra enabling transformation phase has been added to enable effi-
cient mapping on the VWRs.
Algorithm 5.5 shows the adapted version of the iVMX compilation tech-
nique for the VWR architecture. Note that the adapted algorithm is quite
optimized
5
already within ...