Energy-Aware High-Performance Computing: Survey of State-of-the-Art Tools, Techniques, and Environments

Table 6

Correlation of proposed metrics with energy/power control methods and devices.

Target metric	Device	Energy/power control method
		Selection of devices/scheduling	DVFS/DFS/DCT	Power capping	Application optimizations	Hybrid
Performance/execution time with power limit	1x CPU		[52]	[24, 42]
	Nx CPU	[68]	[47]	[46, 62]		[30, 48, 70]
	GPU
	Hybrid	[73]
Performance/execution time/energy minimization + thermal aware	1x CPU	[33]			[38]
	Nx CPU	[44, 66, 67]
	GPU
	Hybrid
Performance/execution time/value + energy optimization	1x CPU	[58]	[59]	[42]	[43]	[77]
	Nx CPU	[76]	[14, 45, 63–65, 72]		[53, 57]	[60, 69, 74]
	GPU	[39, 75]				[34]
	Hybrid					[51]
Energy minimization	1x CPU	[35]	[41, 49]	[40]	[54]
	Nx CPU		[55, 56, 61]
	GPU		[37]		[36]
	Hybrid
Product of energy and execution time	1x CPU
	Nx CPU
	GPU				[36]
	Hybrid

Target metrics and energy/control methods according to Table 4 and Table 5. The names of devices from Table 3 are shortened as follows: Single-core/multicore/manycore CPU (1x CPU), multiprocessor system (Nx CPU), GPU/accelerator (GPU), and hybrid (hybrid).