A Novel GPU-Based Deformation Pipeline

Table 2

Comparison of execution performance.

Grid size	Time for rendering a single frame (in msecs)				Speedup (a/d)	Speedup (b/d)	Speedup (c/d)
	CPU (a)	GPU-CUDA-UOP (b)	GPU-CUDA-OP (c)	GPU-TF (d)
64×64	99.90	4.05	2.557	1.07	93.36	3.79	2.39
128×128	396.82	4.92	3.126	1.47	269.94	3.35	2.13
256×256	1,515.15	6.99	5.742	3.34	453.63	2.09	1.72
512×512	5,000.00	31.06	9.861	11.65	429.18	2.66	0.85
1024×1024	200,000.00	133.69	38.957	54.76	3652.30	2.44	0.71
2048×2048	N/A	699.31	172.408	265.25	N/A	2.63	0.65