Initial condition for efficient mapping of level set algorithms on many-core architectures

EURASIP Journal on Advances in Signal Processing

Table 2 Time measurements on NVIDIA GTX 780 GPU compared to Intel core i7 CPU

Data size	Initial condition	${\bar{T}}_{iteration}$ on GPU (µs)	${\bar{T}}_{iteration}$ on CPU (µs)	N _it	Speedup
256×256	1×1	129	1,610	32	12.5
256×256	2×2	126	2,242	59	17
256×256	8×8	140	3,164	20	22
256×256	32×32	143	8,874	8	62
512×512	1×1	317	3,190	64	10
512×512	4×4	167	8,724	40	52
512×512	16×16	157	12,897	25	82
512×512	64×64	123	16,246	18	132
1,024×1,024	1×1	534	6,431	129	12
1,024×1,024	8×8	548	27,461	55	50
1,024×1,024	32×32	590	43,739	32	74
1,024×1,024	128×128	490	84,078	12	171
2,048×2,048	1×1	560	14,972	210	26
2,048×2,048	16×16	703	79,920	79	113
2,048×2,048	64×64	830	198,980	28	239
2,048×2,048	256×256	684	327,541	7	478