Power and Energy Model

In this page, we will explain how we design and implement power and energy model of hardware.

Embedded Core

We used McPAT [1] library for estimating power and energy consumption of embedded core. We modified McPAT (for windows support and interfacing SimpleSSD), you can download our version of McPAT here.

For McPAT interfacing, see SimpleSSD::CPU::CPU::calculatePower function defined in cpu/cpu.cc.

DRAM

We used DRAMPower [2] library for DRAM power and energy consumption estimation. As gem5 uses DRAMPower, you can find DRAMPower source at ext/drampower of SimpleSSD-FullSystem repository. For standalone version, lib/drampower contains same DRAMPower with full system.

For DRAMPower interfacing, see SimpleSSD::DRAM::SimpleDRAM::read and write functions defined in dram/simple.cc.

NAND flash array

As there is no NAND power estimation library, we created simple power/energy model for NAND flash based on NANDFlashSim [3].

From the NAND flash datasheet, we get current values that NAND flash consumes when idle or perform I/O operations. These values are multiplied by die active time (at each state, which is collected by PAL statistics) to get energy consumption (of course, the voltage value also multiplied).

[1]Sheng Li, et. al., “McPAT: An integrated power, area, and timing modeling framework for multicore and manycore architectures, ” in the 42nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO 42), 2009
[2]Karthik Chandrasekar, et. al., “DRAMPower: Open-source DRAM Power & Energy Estimation Tool, ” URL: http://www.drampower.info
[3]Myoungsoo Jung, et. al., “NANDFlashSim: High-Fidelity, Micro-Architecture-Aware NAND Flash Memory Simulation, ” in ACM Transactions on Storage (ToS), vol. 12 issue 2, no. 6, Feb. 2016