A Strategy for Automatic Performance Tuning of Stencil Computations on GPUs

Table 8

Results for variable work-group size kernels.

	Speedup		Configs		Compile time (s)		Runtime (s)
	NVIDIA	AMD	NVIDIA	AMD	NVIDIA	AMD	NVIDIA	AMD
Fixed size	1.12	1.32	266	266	88.35	53.67	21.51	5.45
Variable size	0.94	1.04	49	68	14.92	14.32	42.15	6.09