func main() {
nimble.Init()
defer nimble.Cleanup()
nimble.SetOD("demag2.out")
gpu.LockCudaThread()
N0, N1, N2 := 4, 32, 1024
cx, cy, cz := 3e-9, 3.125e-9, 3.125e-9
mesh := nimble.NewMesh(N0, N1, N2, cx, cy, cz)
fmt.Println("mesh:", mesh)
mbox := gpu.NewConst("m", "", mesh, nimble.UnifiedMemory, []float64{1, 0, 0})
m := mbox.Output()
const acc = 4
kernel := mag.BruteKernel(mesh, acc)
conv := gpu.NewConvolution("B", "T", mesh, nimble.UnifiedMemory, kernel, m)
B := conv.Output()
outputc := B.NewReader()
nimble.RunStack()
output := host(outputc.ReadNext(mesh.NCell()))
out0 := core.Reshape(output[0], mesh.Size())
out1 := core.Reshape(output[1], mesh.Size())
out2 := core.Reshape(output[2], mesh.Size())
X, Y, Z := N0/2, N1/2, N2/2
if out0[X][Y][Z] < -0.95 || out0[X][Y][Z] > -0.90 || out1[X][Y][Z] > 0.001 || out2[X][Y][Z] > 0.001 {
fmt.Println("failed, got:", out0[X][Y][Z], out1[X][Y][Z], out2[X][Y][Z])
os.Exit(2)
} else {
fmt.Println("OK")
}
}
// Precision test for the kernel: thin film.
func main() {
nimble.Init()
defer nimble.Cleanup()
nimble.SetOD("kernel-film.out")
gpu.LockCudaThread()
Y, X := core.IntArg(0), core.IntArg(1)
y, x := float64(Y), float64(X)
N0, N1, N2 := 1, 1024/Y, 1024/X
cx, cy, cz := 1e-9, 1e-9*y, 1e-9*x
mesh := nimble.NewMesh(N0, N1, N2, cx, cy, cz)
fmt.Println("mesh:", mesh)
mbox := gpu.NewConst("m", "", mesh, nimble.UnifiedMemory, []float64{1, 0, 0})
m := mbox.Output()
acc := 4.
kernel := mag.BruteKernel(mesh, acc)
conv := gpu.NewConvolution("B", "T", mesh, nimble.UnifiedMemory, kernel, m)
B := conv.Output()
outputc := B.NewReader()
nimble.RunStack()
output := host(outputc.ReadNext(mesh.NCell()))
Bz := core.Reshape(output[0], [3]int{N0, N1, N2})
probe := float64(Bz[N0/2][N1/2][N2/2])
fmt.Println(probe)
want := -1.
if math.Abs(probe-want) > 0.01 {
fmt.Println("FAIL")
os.Exit(2)
} else {
fmt.Println("OK")
}
}
