[font=“Courier New”]
Hi ,everyone
I want to solve the 3D wave equ and poisson pressure equ in CUDA and want to use conjugate gradient method.But im not sure whether it will be convergence in these two problem without preconditioner(or with jacobi preconditioner:because its simple to implement)
On the side,i implemented one sparse symmetric matrix multiply vector with 1024x1024 data. The cost as down:
method=[ xmul ] gputime=[ 555.264 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 563.168 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 553.792 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 554.944 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 555.072 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 578.144 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 574.304 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 568.704 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 563.936 ] cputime=[ 4.749 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 553.536 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 565.408 ] cputime=[ 4.470 ] occupancy=[ 1.000 ]
method=[ xmul ] gputime=[ 586.080 ] cputime=[ 5.308 ] occupancy=[ 1.000 ]
My question is:Whether or not fast enough?
My GPU: 8800GTS 320MB
thanks!!
[/font]