func (d *dotProduct) EvalTo(dst *data.Slice) {
A := ValueOf(d.a)
defer cuda.Recycle(A)
B := ValueOf(d.b)
defer cuda.Recycle(B)
cuda.Zero(dst)
cuda.AddDotProduct(dst, 1, A, B)
}
func AddAnisotropyEnergyDensity(dst *data.Slice) {
haveUnixial := Ku1.nonZero() || Ku2.nonZero()
haveCubic := Kc1.nonZero() || Kc2.nonZero() || Kc3.nonZero()
if !haveUnixial && !haveCubic {
return
}
buf := cuda.Buffer(B_anis.NComp(), B_anis.Mesh().Size())
defer cuda.Recycle(buf)
// unnormalized magnetization:
Mf, r := M_full.Slice()
if r {
defer cuda.Recycle(Mf)
}
if haveUnixial {
// 1st
cuda.Zero(buf)
addUniaxialAnisotropyFrom(buf, M, Msat, Ku1, sZero, AnisU)
cuda.AddDotProduct(dst, -1./2., buf, Mf)
// 2nd
cuda.Zero(buf)
addUniaxialAnisotropyFrom(buf, M, Msat, sZero, Ku2, AnisU)
cuda.AddDotProduct(dst, -1./4., buf, Mf)
}
if haveCubic {
// 1st
cuda.Zero(buf)
addCubicAnisotropyFrom(buf, M, Msat, Kc1, sZero, sZero, AnisC1, AnisC2)
cuda.AddDotProduct(dst, -1./4., buf, Mf)
// 2nd
cuda.Zero(buf)
addCubicAnisotropyFrom(buf, M, Msat, sZero, Kc2, sZero, AnisC1, AnisC2)
cuda.AddDotProduct(dst, -1./6., buf, Mf)
// 3nd
cuda.Zero(buf)
addCubicAnisotropyFrom(buf, M, Msat, sZero, sZero, Kc3, AnisC1, AnisC2)
cuda.AddDotProduct(dst, -1./8., buf, Mf)
}
}
func AddAnisotropyEnergyDensity(dst *data.Slice) {
haveUnixial := ku1_red.nonZero() || ku2_red.nonZero()
haveCubic := kc1_red.nonZero() || kc2_red.nonZero() || kc3_red.nonZero()
if !haveUnixial && !haveCubic {
return
}
buf := cuda.Buffer(B_anis.NComp(), B_anis.Mesh().Size())
defer cuda.Recycle(buf)
// unnormalized magnetization:
Mf, r := M_full.Slice()
if r {
defer cuda.Recycle(Mf)
}
if haveUnixial {
// 1st
cuda.Zero(buf)
cuda.AddUniaxialAnisotropy(buf, M.Buffer(), ku1_red.gpuLUT1(), zero.gpuLUT1(), AnisU.gpuLUT(), regions.Gpu())
cuda.AddDotProduct(dst, -1./2., buf, Mf)
// 2nd
cuda.Zero(buf)
cuda.AddUniaxialAnisotropy(buf, M.Buffer(), zero.gpuLUT1(), ku2_red.gpuLUT1(), AnisU.gpuLUT(), regions.Gpu())
cuda.AddDotProduct(dst, -1./4., buf, Mf)
}
if haveCubic {
// 1st
cuda.Zero(buf)
cuda.AddCubicAnisotropy(buf, M.Buffer(), kc1_red.gpuLUT1(), zero.gpuLUT1(), zero.gpuLUT1(), AnisC1.gpuLUT(), AnisC2.gpuLUT(), regions.Gpu())
cuda.AddDotProduct(dst, -1./4., buf, Mf)
// 2nd
cuda.Zero(buf)
cuda.AddCubicAnisotropy(buf, M.Buffer(), zero.gpuLUT1(), kc2_red.gpuLUT1(), zero.gpuLUT1(), AnisC1.gpuLUT(), AnisC2.gpuLUT(), regions.Gpu())
cuda.AddDotProduct(dst, -1./6., buf, Mf)
// 3nd
cuda.Zero(buf)
cuda.AddCubicAnisotropy(buf, M.Buffer(), zero.gpuLUT1(), zero.gpuLUT1(), kc3_red.gpuLUT1(), AnisC1.gpuLUT(), AnisC2.gpuLUT(), regions.Gpu())
cuda.AddDotProduct(dst, -1./8., buf, Mf)
}
}
func (d *dotProduct) Slice() (*data.Slice, bool) {
slice := cuda.Buffer(d.NComp(), d.Mesh().Size())
cuda.Zero(slice)
A, r := d.a.Slice()
if r {
defer cuda.Recycle(A)
}
B, r := d.b.Slice()
if r {
defer cuda.Recycle(B)
}
cuda.AddDotProduct(slice, 1, A, B)
return slice, true
}
// returns a function that adds to dst the energy density:
// prefactor * dot (M_full, field)
func makeEdensAdder(field outputField, prefactor float64) func(*data.Slice) {
return func(dst *data.Slice) {
B, r1 := field.Slice()
if r1 {
defer cuda.Recycle(B)
}
m, r2 := M_full.Slice()
if r2 {
defer cuda.Recycle(m)
}
factor := float32(prefactor)
cuda.AddDotProduct(dst, factor, B, m)
}
}