func (m *magnetization) resize() {
backup := m.Buffer().HostCopy()
s2 := Mesh().Size()
resized := data.Resample(backup, s2)
m.buffer_.Free()
m.buffer_ = cuda.NewSlice(VECTOR, s2)
data.Copy(m.buffer_, resized)
}
func (b *thermField) update() {
// we need to fix the time step here because solver will not yet have done it before the first step.
// FixDt as an lvalue that sets Dt_si on change might be cleaner.
if FixDt != 0 {
Dt_si = FixDt
}
if b.generator == 0 {
b.generator = curand.CreateGenerator(curand.PSEUDO_DEFAULT)
b.generator.SetSeed(b.seed)
}
if b.noise == nil {
b.noise = cuda.NewSlice(b.NComp(), b.Mesh().Size())
// when noise was (re-)allocated it's invalid for sure.
B_therm.step = -1
B_therm.dt = -1
}
if Temp.isZero() {
cuda.Memset(b.noise, 0, 0, 0)
b.step = NSteps
b.dt = Dt_si
return
}
// keep constant during time step
if NSteps == b.step && Dt_si == b.dt {
return
}
if FixDt == 0 {
util.Fatal("Finite temperature requires fixed time step. Set FixDt != 0.")
}
N := Mesh().NCell()
k2_VgammaDt := 2 * mag.Kb / (GammaLL * cellVolume() * Dt_si)
noise := cuda.Buffer(1, Mesh().Size())
defer cuda.Recycle(noise)
const mean = 0
const stddev = 1
dst := b.noise
ms := Msat.MSlice()
defer ms.Recycle()
temp := Temp.MSlice()
defer temp.Recycle()
alpha := Alpha.MSlice()
defer alpha.Recycle()
for i := 0; i < 3; i++ {
b.generator.GenerateNormal(uintptr(noise.DevPtr(0)), int64(N), mean, stddev)
cuda.SetTemperature(dst.Comp(i), noise, k2_VgammaDt, ms, temp, alpha)
}
b.step = NSteps
b.dt = Dt_si
}
// allocate storage (not done by init, as mesh size may not yet be known then)
func (m *magnetization) alloc() {
m.buffer_ = cuda.NewSlice(3, m.Mesh().Size())
m.Set(RandomMag()) // sane starting config
}
func (rk *RK23) Step() {
m := M.Buffer()
size := m.Size()
if FixDt != 0 {
Dt_si = FixDt
}
// upon resize: remove wrongly sized k1
if rk.k1.Size() != m.Size() {
rk.Free()
}
// first step ever: one-time k1 init and eval
if rk.k1 == nil {
rk.k1 = cuda.NewSlice(3, size)
torqueFn(rk.k1)
}
// FSAL cannot be used with temperature
if !Temp.isZero() {
torqueFn(rk.k1)
}
t0 := Time
// backup magnetization
m0 := cuda.Buffer(3, size)
defer cuda.Recycle(m0)
data.Copy(m0, m)
k2, k3, k4 := cuda.Buffer(3, size), cuda.Buffer(3, size), cuda.Buffer(3, size)
defer cuda.Recycle(k2)
defer cuda.Recycle(k3)
defer cuda.Recycle(k4)
h := float32(Dt_si * GammaLL) // internal time step = Dt * gammaLL
// there is no explicit stage 1: k1 from previous step
// stage 2
Time = t0 + (1./2.)*Dt_si
cuda.Madd2(m, m, rk.k1, 1, (1./2.)*h) // m = m*1 + k1*h/2
M.normalize()
torqueFn(k2)
// stage 3
Time = t0 + (3./4.)*Dt_si
cuda.Madd2(m, m0, k2, 1, (3./4.)*h) // m = m0*1 + k2*3/4
M.normalize()
torqueFn(k3)
// 3rd order solution
madd4(m, m0, rk.k1, k2, k3, 1, (2./9.)*h, (1./3.)*h, (4./9.)*h)
M.normalize()
// error estimate
Time = t0 + Dt_si
torqueFn(k4)
Err := k2 // re-use k2 as error
// difference of 3rd and 2nd order torque without explicitly storing them first
madd4(Err, rk.k1, k2, k3, k4, (7./24.)-(2./9.), (1./4.)-(1./3.), (1./3.)-(4./9.), (1. / 8.))
// determine error
err := cuda.MaxVecNorm(Err) * float64(h)
// adjust next time step
if err < MaxErr || Dt_si <= MinDt || FixDt != 0 { // mindt check to avoid infinite loop
// step OK
setLastErr(err)
setMaxTorque(k4)
NSteps++
Time = t0 + Dt_si
adaptDt(math.Pow(MaxErr/err, 1./3.))
data.Copy(rk.k1, k4) // FSAL
} else {
// undo bad step
//util.Println("Bad step at t=", t0, ", err=", err)
util.Assert(FixDt == 0)
Time = t0
data.Copy(m, m0)
NUndone++
adaptDt(math.Pow(MaxErr/err, 1./4.))
}
}
func (geometry *geom) setGeom(s Shape) {
SetBusy(true)
defer SetBusy(false)
if s == nil {
// TODO: would be nice not to save volume if entirely filled
s = universe
}
geometry.shape = s
if geometry.Gpu().IsNil() {
geometry.buffer = cuda.NewSlice(1, geometry.Mesh().Size())
}
host := data.NewSlice(1, geometry.Gpu().Size())
array := host.Scalars()
V := host
v := array
n := geometry.Mesh().Size()
c := geometry.Mesh().CellSize()
cx, cy, cz := c[X], c[Y], c[Z]
progress, progmax := 0, n[Y]*n[Z]
var ok bool
for iz := 0; iz < n[Z]; iz++ {
for iy := 0; iy < n[Y]; iy++ {
progress++
util.Progress(progress, progmax, "Initializing geometry")
for ix := 0; ix < n[X]; ix++ {
r := Index2Coord(ix, iy, iz)
x0, y0, z0 := r[X], r[Y], r[Z]
// check if center and all vertices lie inside or all outside
allIn, allOut := true, true
if s(x0, y0, z0) {
allOut = false
} else {
allIn = false
}
if edgeSmooth != 0 { // center is sufficient if we're not really smoothing
for _, Δx := range []float64{-cx / 2, cx / 2} {
for _, Δy := range []float64{-cy / 2, cy / 2} {
for _, Δz := range []float64{-cz / 2, cz / 2} {
if s(x0+Δx, y0+Δy, z0+Δz) { // inside
allOut = false
} else {
allIn = false
}
}
}
}
}
switch {
case allIn:
v[iz][iy][ix] = 1
ok = true
case allOut:
v[iz][iy][ix] = 0
default:
v[iz][iy][ix] = geometry.cellVolume(ix, iy, iz)
ok = ok || (v[iz][iy][ix] != 0)
}
}
}
}
if !ok {
util.Fatal("SetGeom: geometry completely empty")
}
data.Copy(geometry.buffer, V)
// M inside geom but previously outside needs to be re-inited
needupload := false
geomlist := host.Host()[0]
mhost := M.Buffer().HostCopy()
m := mhost.Host()
rng := rand.New(rand.NewSource(0))
for i := range m[0] {
if geomlist[i] != 0 {
mx, my, mz := m[X][i], m[Y][i], m[Z][i]
if mx == 0 && my == 0 && mz == 0 {
needupload = true
rnd := randomDir(rng)
m[X][i], m[Y][i], m[Z][i] = float32(rnd[X]), float32(rnd[Y]), float32(rnd[Z])
}
}
}
if needupload {
data.Copy(M.Buffer(), mhost)
}
M.normalize() // removes m outside vol
}
// rescale and download quantity, save in rescaleBuf
func (ren *render) download() {
InjectAndWait(func() {
if ren.quant == nil { // not yet set, default = m
ren.quant = &M
}
quant := ren.quant
size := quant.Mesh().Size()
// don't slice out of bounds
renderLayer := ren.layer
if renderLayer >= size[Z] {
renderLayer = size[Z] - 1
}
if renderLayer < 0 {
renderLayer = 0
}
// scaling sanity check
if ren.scale < 1 {
ren.scale = 1
}
if ren.scale > maxScale {
ren.scale = maxScale
}
// Don't render too large images or we choke
for size[X]/ren.scale > maxImgSize {
ren.scale++
}
for size[Y]/ren.scale > maxImgSize {
ren.scale++
}
for i := range size {
size[i] /= ren.scale
if size[i] == 0 {
size[i] = 1
}
}
size[Z] = 1 // selects one layer
// make sure buffers are there
if ren.imgBuf.Size() != size {
ren.imgBuf = data.NewSlice(3, size) // always 3-comp, may be re-used
}
buf, r := quant.Slice()
if r {
defer cuda.Recycle(buf)
}
if !buf.GPUAccess() {
ren.imgBuf = Download(quant) // fallback (no zoom)
return
}
// make sure buffers are there (in CUDA context)
if ren.rescaleBuf.Size() != size {
ren.rescaleBuf.Free()
ren.rescaleBuf = cuda.NewSlice(1, size)
}
for c := 0; c < quant.NComp(); c++ {
cuda.Resize(ren.rescaleBuf, buf.Comp(c), renderLayer)
data.Copy(ren.imgBuf.Comp(c), ren.rescaleBuf)
}
})
}
func (rk *RK45DP) Step() {
m := M.Buffer()
size := m.Size()
if FixDt != 0 {
Dt_si = FixDt
}
// upon resize: remove wrongly sized k1
if rk.k1.Size() != m.Size() {
rk.Free()
}
// first step ever: one-time k1 init and eval
if rk.k1 == nil {
rk.k1 = cuda.NewSlice(3, size)
torqueFn(rk.k1)
}
// FSAL cannot be used with finite temperature
if !Temp.isZero() {
torqueFn(rk.k1)
}
t0 := Time
// backup magnetization
m0 := cuda.Buffer(3, size)
defer cuda.Recycle(m0)
data.Copy(m0, m)
k2, k3, k4, k5, k6 := cuda.Buffer(3, size), cuda.Buffer(3, size), cuda.Buffer(3, size), cuda.Buffer(3, size), cuda.Buffer(3, size)
defer cuda.Recycle(k2)
defer cuda.Recycle(k3)
defer cuda.Recycle(k4)
defer cuda.Recycle(k5)
defer cuda.Recycle(k6)
// k2 will be re-used as k7
h := float32(Dt_si * GammaLL) // internal time step = Dt * gammaLL
// there is no explicit stage 1: k1 from previous step
// stage 2
Time = t0 + (1./5.)*Dt_si
cuda.Madd2(m, m, rk.k1, 1, (1./5.)*h) // m = m*1 + k1*h/5
M.normalize()
torqueFn(k2)
// stage 3
Time = t0 + (3./10.)*Dt_si
cuda.Madd3(m, m0, rk.k1, k2, 1, (3./40.)*h, (9./40.)*h)
M.normalize()
torqueFn(k3)
// stage 4
Time = t0 + (4./5.)*Dt_si
madd4(m, m0, rk.k1, k2, k3, 1, (44./45.)*h, (-56./15.)*h, (32./9.)*h)
M.normalize()
torqueFn(k4)
// stage 5
Time = t0 + (8./9.)*Dt_si
madd5(m, m0, rk.k1, k2, k3, k4, 1, (19372./6561.)*h, (-25360./2187.)*h, (64448./6561.)*h, (-212./729.)*h)
M.normalize()
torqueFn(k5)
// stage 6
Time = t0 + (1.)*Dt_si
madd6(m, m0, rk.k1, k2, k3, k4, k5, 1, (9017./3168.)*h, (-355./33.)*h, (46732./5247.)*h, (49./176.)*h, (-5103./18656.)*h)
M.normalize()
torqueFn(k6)
// stage 7: 5th order solution
Time = t0 + (1.)*Dt_si
// no k2
madd6(m, m0, rk.k1, k3, k4, k5, k6, 1, (35./384.)*h, (500./1113.)*h, (125./192.)*h, (-2187./6784.)*h, (11./84.)*h) // 5th
M.normalize()
k7 := k2 // re-use k2
torqueFn(k7) // next torque if OK
// error estimate
Err := cuda.Buffer(3, size) //k3 // re-use k3 as error estimate
defer cuda.Recycle(Err)
madd6(Err, rk.k1, k3, k4, k5, k6, k7, (35./384.)-(5179./57600.), (500./1113.)-(7571./16695.), (125./192.)-(393./640.), (-2187./6784.)-(-92097./339200.), (11./84.)-(187./2100.), (0.)-(1./40.))
// determine error
err := cuda.MaxVecNorm(Err) * float64(h)
// adjust next time step
if err < MaxErr || Dt_si <= MinDt || FixDt != 0 { // mindt check to avoid infinite loop
// step OK
setLastErr(err)
setMaxTorque(k7)
NSteps++
Time = t0 + Dt_si
adaptDt(math.Pow(MaxErr/err, 1./5.))
data.Copy(rk.k1, k7) // FSAL
} else {
// undo bad step
//util.Println("Bad step at t=", t0, ", err=", err)
util.Assert(FixDt == 0)
Time = t0
data.Copy(m, m0)
NUndone++
adaptDt(math.Pow(MaxErr/err, 1./6.))
}
}
