12 #include "../tmop.hpp" 14 #include "../linearform.hpp" 15 #include "../../general/forall.hpp" 16 #include "../../linalg/kernels.hpp" 22 const double lim_normal,
38 const bool const_c0 = c0_.
Size() == 1;
40 constexpr
int DIM = 3;
41 const int D1D = T_D1D ? T_D1D : d1d;
42 const int Q1D = T_Q1D ? T_Q1D : q1d;
44 const auto C0 = const_c0 ?
47 const auto LD =
Reshape(lim_dist.
Read(), D1D, D1D, D1D, NE);
57 MFEM_FORALL_3D(e, NE, Q1D, Q1D, Q1D,
59 const int D1D = T_D1D ? T_D1D : d1d;
60 const int Q1D = T_Q1D ? T_Q1D : q1d;
61 constexpr
int MQ1 = T_Q1D ? T_Q1D : T_MAX;
62 constexpr
int MD1 = T_D1D ? T_D1D : T_MAX;
63 constexpr
int MDQ = (MQ1 > MD1) ? MQ1 : MD1;
65 MFEM_SHARED
double B[MQ1*MD1];
66 MFEM_SHARED
double sBLD[MQ1*MD1];
67 kernels::internal::LoadB<MD1,MQ1>(D1D,Q1D,bld,sBLD);
70 MFEM_SHARED
double sm0[MDQ*MDQ*MDQ];
71 MFEM_SHARED
double sm1[MDQ*MDQ*MDQ];
77 MFEM_SHARED
double DDD0[3][MD1*MD1*MD1];
78 MFEM_SHARED
double DDQ0[3][MD1*MD1*MQ1];
79 MFEM_SHARED
double DQQ0[3][MD1*MQ1*MQ1];
80 MFEM_SHARED
double QQQ0[3][MQ1*MQ1*MQ1];
82 MFEM_SHARED
double DDD1[3][MD1*MD1*MD1];
83 MFEM_SHARED
double DDQ1[3][MD1*MD1*MQ1];
84 MFEM_SHARED
double DQQ1[3][MD1*MQ1*MQ1];
85 MFEM_SHARED
double QQQ1[3][MQ1*MQ1*MQ1];
87 kernels::internal::LoadX(e,D1D,LD,DDD);
88 kernels::internal::LoadX<MD1>(e,D1D,X0,DDD0);
89 kernels::internal::LoadX<MD1>(e,D1D,X1,DDD1);
91 kernels::internal::LoadB<MD1,MQ1>(D1D,Q1D,
b,B);
93 kernels::internal::EvalX(D1D,Q1D,BLD,DDD,DDQ);
94 kernels::internal::EvalY(D1D,Q1D,BLD,DDQ,DQQ);
95 kernels::internal::EvalZ(D1D,Q1D,BLD,DQQ,QQQ);
97 kernels::internal::EvalX<MD1,MQ1>(D1D,Q1D,B,DDD0,DDQ0);
98 kernels::internal::EvalY<MD1,MQ1>(D1D,Q1D,B,DDQ0,DQQ0);
99 kernels::internal::EvalZ<MD1,MQ1>(D1D,Q1D,B,DQQ0,QQQ0);
101 kernels::internal::EvalX<MD1,MQ1>(D1D,Q1D,B,DDD1,DDQ1);
102 kernels::internal::EvalY<MD1,MQ1>(D1D,Q1D,B,DDQ1,DQQ1);
103 kernels::internal::EvalZ<MD1,MQ1>(D1D,Q1D,B,DQQ1,QQQ1);
105 MFEM_FOREACH_THREAD(qz,z,Q1D)
107 MFEM_FOREACH_THREAD(qy,y,Q1D)
109 MFEM_FOREACH_THREAD(qx,x,Q1D)
111 double D, p0[3], p1[3];
112 const double *Jtr = &J(0,0,qx,qy,qz,e);
113 const double detJtr = kernels::Det<3>(Jtr);
114 const double weight = W(qx,qy,qz) * detJtr;
115 const double coeff0 = const_c0 ? C0(0,0,0,0) : C0(qx,qy,qz,e);
117 kernels::internal::PullEval(qx,qy,qz,QQQ,D);
118 kernels::internal::PullEval<MQ1>(Q1D,qx,qy,qz,QQQ0,p0);
119 kernels::internal::PullEval<MQ1>(Q1D,qx,qy,qz,QQQ1,p1);
121 const double dist = D;
126 id2 = 0.5 / (dist*dist);
127 dsq = kernels::DistanceSquared<3>(p1,p0) * id2;
128 E(qx,qy,qz,e) = weight * lim_normal * dsq * coeff0;
132 id2 = 1.0 / (dist*dist);
133 dsq = kernels::DistanceSquared<3>(p1,p0) * id2;
134 E(qx,qy,qz,e) = weight * lim_normal * exp(10.0*(dsq-1.0)) * coeff0;
140 return energy * ones;
146 const int D1D =
PA.maps->ndof;
147 const int Q1D =
PA.maps->nqpt;
148 const int id = (D1D << 4 ) | Q1D;
155 MFEM_VERIFY(
PA.maps_lim->ndof == D1D,
"");
156 MFEM_VERIFY(
PA.maps_lim->nqpt == Q1D,
"");
163 const bool exp_lim = (el) ?
true :
false;
165 MFEM_LAUNCH_TMOP_KERNEL(EnergyPA_C0_3D,
id,ln,
LD,
C0,N,J,W,B,BLD,
X0,X,
O,
E,
const T * Read(bool on_dev=true) const
Shortcut for mfem::Read(a.GetMemory(), a.Size(), on_dev).
struct mfem::TMOP_Integrator::@23 PA
int Size() const
Returns the size of the vector.
virtual const double * Read(bool on_dev=true) const
Shortcut for mfem::Read(vec.GetMemory(), vec.Size(), on_dev).
double GetLocalStateEnergyPA_C0_3D(const Vector &) const
const double * Read(bool on_dev=true) const
Shortcut for mfem::Read( GetMemory(), TotalSize(), on_dev).
virtual double * Write(bool on_dev=true)
Shortcut for mfem::Write(vec.GetMemory(), vec.Size(), on_dev).
MFEM_REGISTER_TMOP_KERNELS(void, DatcSize, const int NE, const int ncomp, const int sizeidx, const DenseMatrix &w_, const Array< double > &b_, const Vector &x_, DenseTensor &j_, const int d1d, const int q1d)
A basic generic Tensor class, appropriate for use on the GPU.
Exponential limiter function in TMOP_Integrator.
MFEM_HOST_DEVICE DeviceTensor< sizeof...(Dims), T > Reshape(T *ptr, Dims... dims)
Wrap a pointer as a DeviceTensor with automatically deduced template parameters.
Rank 3 tensor (array of matrices)
TMOP_LimiterFunction * lim_func