TIME thru                            Starting, CPU:  0.374 (d= 0.374)
Thrust v1.7, CUDA_VERSION: 6,000, CUDA_ARCH: 0
N=1,000,000,000
TIME thru                   cudaMallocManaged, CPU:  1.716 (d= 1.342)
TIME thru                            for loop, CPU:  3.209 (d= 1.493)
TIME thru                  sort managed array, CPU:  9.463 (d= 6.253)