Abstract:

In this paper, an allocation strategy is proposed to optimize energy of memory system. A timeslotted cache conflict graph is proposed to model the behavior of data cache conflict. An integer nonlinear programming (INP) is implemented, which can avoid time-consuming linearization process, to select the most profitable data pages and employ a virtual memory system to remap those data pages that can cause severe cache conflict within a time slot, to the on-chip scratchpad memory (SPM). To minimize the swapping overhead of dynamic SPM allocation, we introduce a novel SPM controller with a tightly coupled DMA to issue the swapping operations without CPU intervention. The proposed method can optimize all data segments including global data, heap and stack data in general, and reduce the total energy consumption by 24.83% on average without any performance degradation.

Key words: time-slotted Cache conflict graph, SPM controller, energy optimization