Parallelization of an Unsteady ALE Solver with Deforming Mesh Using OpenACC

Table 3

CPU times, GPU times, and speedups on four test meshes.

Mesh	CTMA(s)	GTMA(s)	SPMA	CTB(s)	GTB(s)	SPB	OSP
1	0.0379	0.0021	18.04x	0.2476	0.0304	8.14x	8.78x
2	0.4902	0.0219	22.38x	3.6404	0.3023	12.04x	12.74x
3	0.3805	0.0244	15.59x	2.5815	0.2392	10.79x	11.23x
4	0.8923	0.0537	16.62x	5.1088	0.3582	14.26x	14.56x

CTMA: CPU time of matrix assembly; GTMA: GPU time of matrix assembly; SPMA: speedup for matrix assembly; CTB: CPU time of p-bicgstab solver; GTB: GPU time of p-bicgstab solver; SPB: speedup for p-bicgstab solver; OSP: overall speedup.