phynexis-deps/mfem
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Remove erroneous MFEM_HOST_DEVICE pragmas

Browse Source
This commit is contained in:
Sebastian Grimberg
2023-06-01 12:23:51 -07:00
parent ae3f95ea65
commit 9c0eae5e3e
5 changed files with 726 additions and 808 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fem/integ/bilininteg_diffusion_kernels.hpp
+157 -176
View File
@@ -24,14 +24,13 @@ namespace internal
// OCCA 2D Assemble kernel
#ifdef MFEM_USE_OCCA
MFEM_HOST_DEVICE inline
void OccaPADiffusionSetup2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &op)
inline void OccaPADiffusionSetup2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &op)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -53,14 +52,13 @@ void OccaPADiffusionSetup2D(const int D1D,
OccaDiffSetup2D_ker.at(id)(NE, o_W, o_J, o_C, o_op, const_c);
}
MFEM_HOST_DEVICE inline
void OccaPADiffusionSetup3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &op)
inline void OccaPADiffusionSetup3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &op)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -85,7 +83,6 @@ void OccaPADiffusionSetup3D(const int D1D,
// PA Diffusion Assemble 2D kernel
template<const int T_SDIM>
MFEM_HOST_DEVICE inline
void PADiffusionSetup2D(const int Q1D,
const int coeffDim,
const int NE,
@@ -95,14 +92,13 @@ void PADiffusionSetup2D(const int Q1D,
Vector &d);
template<>
MFEM_HOST_DEVICE inline
void PADiffusionSetup2D<2>(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
inline void PADiffusionSetup2D<2>(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
{
const bool symmetric = (coeffDim != 4);
const bool const_c = c.Size() == 1;
@@ -163,14 +159,13 @@ void PADiffusionSetup2D<2>(const int Q1D,
// PA Diffusion Assemble 2D kernel with 3D node coords
template<>
MFEM_HOST_DEVICE inline
void PADiffusionSetup2D<3>(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
inline void PADiffusionSetup2D<3>(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
{
MFEM_VERIFY(coeffDim == 1, "Matrix and vector coefficients not supported");
constexpr int DIM = 2;
@@ -209,14 +204,13 @@ void PADiffusionSetup2D<3>(const int Q1D,
}
// PA Diffusion Assemble 3D kernel
MFEM_HOST_DEVICE inline
void PADiffusionSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
inline void PADiffusionSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
const Vector &c,
Vector &d)
{
const bool symmetric = (coeffDim != 9);
const bool const_c = c.Size() == 1;
@@ -330,17 +324,16 @@ void PADiffusionSetup3D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PADiffusionSetup(const int dim,
const int sdim,
const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &D)
inline void PADiffusionSetup(const int dim,
const int sdim,
const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &W,
const Vector &J,
const Vector &C,
Vector &D)
{
if (dim == 1) { MFEM_ABORT("dim==1 not supported in PADiffusionSetup"); }
if (dim == 2)
@@ -371,15 +364,14 @@ void PADiffusionSetup(const int dim,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PADiffusionDiagonal2D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
inline void PADiffusionDiagonal2D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -438,15 +430,14 @@ void PADiffusionDiagonal2D(const int NE,
// Shared memory PA Diffusion Diagonal 2D kernel
template<int T_D1D = 0, int T_Q1D = 0, int T_NBZ = 0>
MFEM_HOST_DEVICE inline
void SmemPADiffusionDiagonal2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPADiffusionDiagonal2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -533,15 +524,14 @@ void SmemPADiffusionDiagonal2D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PADiffusionDiagonal3D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
inline void PADiffusionDiagonal3D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
{
constexpr int DIM = 3;
const int D1D = T_D1D ? T_D1D : d1d;
@@ -635,15 +625,14 @@ void PADiffusionDiagonal3D(const int NE,
// Shared memory PA Diffusion Diagonal 3D kernel
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPADiffusionDiagonal3D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPADiffusionDiagonal3D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
constexpr int DIM = 3;
const int D1D = T_D1D ? T_D1D : d1d;
@@ -753,16 +742,15 @@ void SmemPADiffusionDiagonal3D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PADiffusionAssembleDiagonal(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symm,
const Array<double> &B,
const Array<double> &G,
const Vector &D,
Vector &Y)
inline void PADiffusionAssembleDiagonal(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symm,
const Array<double> &B,
const Array<double> &G,
const Vector &D,
Vector &Y)
{
if (dim == 2)
{
@@ -801,17 +789,16 @@ void PADiffusionAssembleDiagonal(const int dim,
#ifdef MFEM_USE_OCCA
// OCCA PA Diffusion Apply 2D kernel
MFEM_HOST_DEVICE inline
void OccaPADiffusionApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void OccaPADiffusionApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -851,17 +838,16 @@ void OccaPADiffusionApply2D(const int D1D,
}
// OCCA PA Diffusion Apply 3D kernel
MFEM_HOST_DEVICE inline
void OccaPADiffusionApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void OccaPADiffusionApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -903,18 +889,17 @@ void OccaPADiffusionApply3D(const int D1D,
// PA Diffusion Apply 2D kernel
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PADiffusionApply2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Array<double> &bt_,
const Array<double> &gt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void PADiffusionApply2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Array<double> &bt_,
const Array<double> &gt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -1026,16 +1011,15 @@ void PADiffusionApply2D(const int NE,
// Shared memory PA Diffusion Apply 2D kernel
template<int T_D1D = 0, int T_Q1D = 0, int T_NBZ = 0>
MFEM_HOST_DEVICE inline
void SmemPADiffusionApply2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPADiffusionApply2D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -1185,17 +1169,16 @@ void SmemPADiffusionApply2D(const int NE,
// PA Diffusion Apply 3D kernel
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PADiffusionApply3D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Array<double> &bt,
const Array<double> &gt,
const Vector &d_,
const Vector &x_,
Vector &y_,
int d1d = 0, int q1d = 0)
inline void PADiffusionApply3D(const int NE,
const bool symmetric,
const Array<double> &b,
const Array<double> &g,
const Array<double> &bt,
const Array<double> &gt,
const Vector &d_,
const Vector &x_,
Vector &y_,
int d1d = 0, int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -1378,16 +1361,15 @@ void PADiffusionApply3D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPADiffusionApply3D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPADiffusionApply3D(const int NE,
const bool symmetric,
const Array<double> &b_,
const Array<double> &g_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -1600,19 +1582,18 @@ void SmemPADiffusionApply3D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PADiffusionApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symm,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void PADiffusionApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symm,
const Array<double> &B,
const Array<double> &G,
const Array<double> &Bt,
const Array<double> &Gt,
const Vector &D,
const Vector &X,
Vector &Y)
{
#ifdef MFEM_USE_OCCA
if (DeviceCanUseOcca())
fem/integ/bilininteg_hcurl_kernels.hpp
+213 -235
View File
@@ -25,15 +25,14 @@ namespace mfem
namespace internal
{
MFEM_HOST_DEVICE inline
void PAHcurlMassAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
inline void PAHcurlMassAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
{
constexpr static int VDIM = 2;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -83,15 +82,14 @@ void PAHcurlMassAssembleDiagonal2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHcurlMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
inline void PAHcurlMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
{
constexpr static int MAX_D1D = HCURL_MAX_D1D;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -157,15 +155,14 @@ void PAHcurlMassAssembleDiagonal3D(const int D1D,
}
template<int T_D1D = HCURL_MAX_D1D, int T_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPAHcurlMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
inline void SmemPAHcurlMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Vector &pa_data,
Vector &diag)
{
MFEM_VERIFY(D1D <= HCURL_MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= HCURL_MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -280,18 +277,17 @@ void SmemPAHcurlMassAssembleDiagonal3D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHcurlMassApply2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlMassApply2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HCURL_MAX_D1D;
@@ -411,18 +407,17 @@ void PAHcurlMassApply2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHcurlMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
constexpr static int MAX_D1D = HCURL_MAX_D1D;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -608,18 +603,17 @@ void PAHcurlMassApply3D(const int D1D,
}
template<int T_D1D = HCURL_MAX_D1D, int T_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPAHcurlMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void SmemPAHcurlMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= HCURL_MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= HCURL_MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -800,13 +794,12 @@ void SmemPAHcurlMassApply3D(const int D1D,
}
// PA H(curl) curl-curl assemble 2D kernel
MFEM_HOST_DEVICE inline
void PACurlCurlSetup2D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff,
Vector &op)
inline void PACurlCurlSetup2D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff,
Vector &op)
{
const int NQ = Q1D*Q1D;
auto W = w.Read();
@@ -828,14 +821,13 @@ void PACurlCurlSetup2D(const int Q1D,
}
// PA H(curl) curl-curl assemble 3D kernel
MFEM_HOST_DEVICE inline
void PACurlCurlSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff,
Vector &op)
inline void PACurlCurlSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff,
Vector &op)
{
const int NQ = Q1D*Q1D*Q1D;
const bool symmetric = (coeffDim != 9);
@@ -928,14 +920,13 @@ void PACurlCurlSetup3D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PACurlCurlAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
inline void PACurlCurlAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
{
constexpr static int VDIM = 2;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -984,17 +975,16 @@ void PACurlCurlAssembleDiagonal2D(const int D1D,
}
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PACurlCurlAssembleDiagonal3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &go,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
inline void PACurlCurlAssembleDiagonal3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &go,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
{
constexpr static int VDIM = 3;
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
@@ -1171,17 +1161,16 @@ void PACurlCurlAssembleDiagonal3D(const int D1D,
}
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPACurlCurlAssembleDiagonal3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &go,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
inline void SmemPACurlCurlAssembleDiagonal3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &go,
const Array<double> &gc,
const Vector &pa_data,
Vector &diag)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -1362,17 +1351,16 @@ void SmemPACurlCurlAssembleDiagonal3D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PACurlCurlApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PACurlCurlApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HCURL_MAX_D1D;
@@ -1484,20 +1472,19 @@ void PACurlCurlApply2D(const int D1D,
}
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PACurlCurlApply3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PACurlCurlApply3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -1998,20 +1985,19 @@ void PACurlCurlApply3D(const int D1D,
}
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPACurlCurlApply3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void SmemPACurlCurlApply3D(const int D1D,
const int Q1D,
const bool symmetric,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -2030,7 +2016,7 @@ void SmemPACurlCurlApply3D(const int D1D,
const int s = symmetric ? 6 : 9;
auto device_kernel = [=] MFEM_DEVICE (int e)
auto device_kernel = [=] MFEM_HOST_DEVICE (int e)
{
constexpr int VDIM = 3;
@@ -2320,12 +2306,11 @@ void SmemPACurlCurlApply3D(const int D1D,
}
// PA H(curl)-L2 assemble 2D kernel
MFEM_HOST_DEVICE inline
void PAHcurlL2Setup2D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff,
Vector &op)
inline void PAHcurlL2Setup2D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff,
Vector &op)
{
const int NQ = Q1D*Q1D;
auto W = w.Read();
@@ -2341,13 +2326,12 @@ void PAHcurlL2Setup2D(const int Q1D,
}
// PA H(curl) Mass Assemble 3D kernel
MFEM_HOST_DEVICE inline
void PAHcurlL2Setup3D(const int NQ,
const int coeffDim,
const int NE,
const Array<double> &w,
Vector &coeff,
Vector &op)
inline void PAHcurlL2Setup3D(const int NQ,
const int coeffDim,
const int NE,
const Array<double> &w,
Vector &coeff,
Vector &op)
{
auto W = w.Read();
auto C = Reshape(coeff.Read(), coeffDim, NQ, NE);
@@ -2365,18 +2349,17 @@ void PAHcurlL2Setup3D(const int NQ,
});
}
MFEM_HOST_DEVICE inline
void PAHcurlL2Apply2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &bt,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x, // trial = H(curl)
Vector &y) // test = L2 or H1
inline void PAHcurlL2Apply2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &bt,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x, // trial = H(curl)
Vector &y) // test = L2 or H1
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HCURL_MAX_D1D;
@@ -2481,18 +2464,17 @@ void PAHcurlL2Apply2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHcurlL2ApplyTranspose2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &b,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x, // trial = H(curl)
Vector &y) // test = L2 or H1
inline void PAHcurlL2ApplyTranspose2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bot,
const Array<double> &b,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x, // trial = H(curl)
Vector &y) // test = L2 or H1
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HCURL_MAX_D1D;
@@ -2597,19 +2579,18 @@ void PAHcurlL2ApplyTranspose2D(const int D1D,
// Apply to x corresponding to DOFs in H(curl) (trial), whose curl is
// integrated against H(curl) test functions corresponding to y.
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PAHcurlL2Apply3D(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlL2Apply3D(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -2963,17 +2944,16 @@ void PAHcurlL2Apply3D(const int D1D,
// Apply to x corresponding to DOFs in H(curl) (trial), whose curl is
// integrated against H(curl) test functions corresponding to y.
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPAHcurlL2Apply3D(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void SmemPAHcurlL2Apply3D(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -2985,7 +2965,7 @@ void SmemPAHcurlL2Apply3D(const int D1D,
auto X = Reshape(x.Read(), 3*(D1D-1)*D1D*D1D, NE);
auto Y = Reshape(y.ReadWrite(), 3*(D1D-1)*D1D*D1D, NE);
auto device_kernel = [=] MFEM_DEVICE (int e)
auto device_kernel = [=] MFEM_HOST_DEVICE (int e)
{
constexpr int VDIM = 3;
constexpr int maxCoeffDim = 9;
@@ -3274,19 +3254,18 @@ void SmemPAHcurlL2Apply3D(const int D1D,
// Apply to x corresponding to DOFs in H(curl) (trial), integrated against curl
// of H(curl) test functions corresponding to y.
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PAHcurlL2Apply3DTranspose(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlL2Apply3DTranspose(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
// See PAHcurlL2Apply3D for comments.
@@ -3638,17 +3617,16 @@ void PAHcurlL2Apply3DTranspose(const int D1D,
}
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void SmemPAHcurlL2Apply3DTranspose(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void SmemPAHcurlL2Apply3DTranspose(const int D1D,
const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -3660,7 +3638,7 @@ void SmemPAHcurlL2Apply3DTranspose(const int D1D,
auto X = Reshape(x.Read(), 3*(D1D-1)*D1D*D1D, NE);
auto Y = Reshape(y.ReadWrite(), 3*(D1D-1)*D1D*D1D, NE);
auto device_kernel = [=] MFEM_DEVICE (int e)
auto device_kernel = [=] MFEM_HOST_DEVICE (int e)
{
constexpr int VDIM = 3;
constexpr int maxCoeffDim = 9;
fem/integ/bilininteg_hcurlhdiv_kernels.hpp
+68 -74
View File
@@ -25,15 +25,14 @@ namespace internal
// PA H(curl) x H(div) mass assemble 2D kernel, with factor
// dF^{-1} C dF for a vector or matrix coefficient C.
// If transpose, use dF^T C dF^{-T} for H(div) x H(curl).
MFEM_HOST_DEVICE inline
void PAHcurlHdivMassSetup2D(const int Q1D,
const int coeffDim,
const int NE,
const bool transpose,
const Array<double> &w_,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PAHcurlHdivMassSetup2D(const int Q1D,
const int coeffDim,
const int NE,
const bool transpose,
const Array<double> &w_,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const bool symmetric = (coeffDim != 4);
auto W = Reshape(w_.Read(), Q1D, Q1D);
@@ -99,15 +98,14 @@ void PAHcurlHdivMassSetup2D(const int Q1D,
// PA H(curl) x H(div) mass assemble 3D kernel, with factor
// dF^{-1} C dF for a vector or matrix coefficient C.
// If transpose, use dF^T C dF^{-T} for H(div) x H(curl).
MFEM_HOST_DEVICE inline
void PAHcurlHdivMassSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const bool transpose,
const Array<double> &w_,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PAHcurlHdivMassSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const bool transpose,
const Array<double> &w_,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const bool symmetric = (coeffDim != 9);
auto W = Reshape(w_.Read(), Q1D, Q1D, Q1D);
@@ -216,21 +214,20 @@ void PAHcurlHdivMassSetup3D(const int Q1D,
// Mass operator for H(curl) and H(div) functions, using Piola transformations
// u = dF^{-T} \hat{u} in H(curl), v = (1 / det dF) dF \hat{v} in H(div).
MFEM_HOST_DEVICE inline
void PAHcurlHdivMassApply2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const bool scalarCoeff,
const bool trialHcurl,
const bool transpose,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHcurlHdivMassApply2D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const bool scalarCoeff,
const bool trialHcurl,
const bool transpose,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int MAX_D1D = HCURL_MAX_D1D;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -363,21 +360,20 @@ void PAHcurlHdivMassApply2D(const int D1D,
// Mass operator for H(curl) and H(div) functions, using Piola transformations
// u = dF^{-T} \hat{u} in H(curl), v = (1 / det dF) dF \hat{v} in H(div).
MFEM_HOST_DEVICE inline
void PAHcurlHdivMassApply3D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const bool scalarCoeff,
const bool trialHcurl,
const bool transpose,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHcurlHdivMassApply3D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const bool scalarCoeff,
const bool trialHcurl,
const bool transpose,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int MAX_D1D = HCURL_MAX_D1D;
constexpr static int MAX_Q1D = HCURL_MAX_Q1D;
@@ -585,19 +581,18 @@ void PAHcurlHdivMassApply3D(const int D1D,
// Apply to x corresponding to DOFs in H(curl) (trial), whose curl is
// integrated against H(div) test functions corresponding to y.
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PAHcurlHdivApply3D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlHdivApply3D(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gc,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
@@ -942,19 +937,18 @@ void PAHcurlHdivApply3D(const int D1D,
// Apply to x corresponding to DOFs in H(div) (test), integrated against the
// curl of H(curl) trial functions corresponding to y.
template<int MAX_D1D = HCURL_MAX_D1D, int MAX_Q1D = HCURL_MAX_Q1D>
MFEM_HOST_DEVICE inline
void PAHcurlHdivApply3DTranspose(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
inline void PAHcurlHdivApply3DTranspose(const int D1D,
const int D1Dtest,
const int Q1D,
const int NE,
const Array<double> &bo,
const Array<double> &bc,
const Array<double> &bot,
const Array<double> &bct,
const Array<double> &gct,
const Vector &pa_data,
const Vector &x,
Vector &y)
{
MFEM_VERIFY(D1D <= MAX_D1D, "Error: D1D > MAX_D1D");
MFEM_VERIFY(Q1D <= MAX_Q1D, "Error: Q1D > MAX_Q1D");
fem/integ/bilininteg_hdiv_kernels.hpp
+202 -225
View File
@@ -26,14 +26,13 @@ namespace internal
{
// PA H(div) Mass Assemble 2D kernel
MFEM_HOST_DEVICE inline
void PAHdivMassSetup2D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PAHdivMassSetup2D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const bool symmetric = (coeffDim != 4);
const int NQ = Q1D*Q1D;
@@ -91,14 +90,13 @@ void PAHdivMassSetup2D(const int Q1D,
}
// PA H(div) Mass Assemble 3D kernel
MFEM_HOST_DEVICE inline
void PAHdivMassSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PAHdivMassSetup3D(const int Q1D,
const int coeffDim,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const bool symmetric = (coeffDim != 9);
const int NQ = Q1D*Q1D*Q1D;
@@ -179,15 +177,14 @@ void PAHdivMassSetup3D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PAHdivMassAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Vector &op_,
Vector &diag_)
inline void PAHdivMassAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Vector &op_,
Vector &diag_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_Q1D = HDIV_MAX_Q1D;
@@ -236,15 +233,14 @@ void PAHdivMassAssembleDiagonal2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHdivMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Vector &op_,
Vector &diag_)
inline void PAHdivMassAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Vector &op_,
Vector &diag_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -306,18 +302,17 @@ void PAHdivMassAssembleDiagonal3D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHdivMassApply2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivMassApply2D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HDIV_MAX_D1D;
@@ -439,18 +434,17 @@ void PAHdivMassApply2D(const int D1D,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPAHdivMassApply2D(const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAHdivMassApply2D(const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_CONTRACT_VAR(Bot_);
MFEM_CONTRACT_VAR(Bct_);
@@ -608,18 +602,17 @@ void SmemPAHdivMassApply2D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PAHdivMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivMassApply3D(const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -805,18 +798,17 @@ void PAHdivMassApply3D(const int D1D,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPAHdivMassApply3D(const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAHdivMassApply3D(const int NE,
const bool symmetric,
const Array<double> &Bo_,
const Array<double> &Bc_,
const Array<double> &Bot_,
const Array<double> &Bct_,
const Vector &op_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_CONTRACT_VAR(Bot_);
MFEM_CONTRACT_VAR(Bct_);
@@ -1093,19 +1085,18 @@ void SmemPAHdivMassApply3D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PAHdivMassApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo,
const Array<double> &Bc,
const Array<double> &Bot,
const Array<double> &Bct,
const Vector &op,
const Vector &x,
Vector &y)
inline void PAHdivMassApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const bool symmetric,
const Array<double> &Bo,
const Array<double> &Bc,
const Array<double> &Bot,
const Array<double> &Bct,
const Vector &op,
const Vector &x,
Vector &y)
{
const int id = (D1D << 4) | Q1D;
@@ -1139,13 +1130,12 @@ void PAHdivMassApply(const int dim,
// PA H(div) div-div assemble 2D kernel
// NOTE: this is identical to PACurlCurlSetup2D
MFEM_HOST_DEVICE inline
void PADivDivSetup2D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PADivDivSetup2D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const int NQ = Q1D*Q1D;
auto W = w.Read();
@@ -1166,13 +1156,12 @@ void PADivDivSetup2D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PADivDivSetup3D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
inline void PADivDivSetup3D(const int Q1D,
const int NE,
const Array<double> &w,
const Vector &j,
Vector &coeff_,
Vector &op)
{
const int NQ = Q1D*Q1D*Q1D;
auto W = w.Read();
@@ -1201,14 +1190,13 @@ void PADivDivSetup3D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PADivDivAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Vector &op_,
Vector &diag_)
inline void PADivDivAssembleDiagonal2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Vector &op_,
Vector &diag_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_Q1D = HDIV_MAX_Q1D;
@@ -1258,14 +1246,13 @@ void PADivDivAssembleDiagonal2D(const int D1D,
});
}
MFEM_HOST_DEVICE inline
void PADivDivAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Vector &op_,
Vector &diag_)
inline void PADivDivAssembleDiagonal3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Vector &op_,
Vector &diag_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -1325,17 +1312,16 @@ void PADivDivAssembleDiagonal3D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PADivDivApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &Bot_,
const Array<double> &Gct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PADivDivApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &Bot_,
const Array<double> &Gct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HDIV_MAX_D1D;
@@ -1445,17 +1431,16 @@ void PADivDivApply2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PADivDivApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &Bot_,
const Array<double> &Gct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PADivDivApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &Bot_,
const Array<double> &Gct_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -1623,12 +1608,11 @@ void PADivDivApply3D(const int D1D,
}
// PA H(div)-L2 (div u, p) assemble 2D kernel
MFEM_HOST_DEVICE inline
void PAHdivL2Setup2D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff_,
Vector &op)
inline void PAHdivL2Setup2D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff_,
Vector &op)
{
const int NQ = Q1D*Q1D;
auto W = w.Read();
@@ -1643,12 +1627,11 @@ void PAHdivL2Setup2D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PAHdivL2Setup3D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff_,
Vector &op)
inline void PAHdivL2Setup3D(const int Q1D,
const int NE,
const Array<double> &w,
Vector &coeff_,
Vector &op)
{
const int NQ = Q1D*Q1D*Q1D;
auto W = w.Read();
@@ -1664,17 +1647,16 @@ void PAHdivL2Setup3D(const int Q1D,
});
}
MFEM_HOST_DEVICE inline
void PAHdivL2AssembleDiagonal_ADAt_2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &D_,
Vector &diag_)
inline void PAHdivL2AssembleDiagonal_ADAt_2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &D_,
Vector &diag_)
{
constexpr static int VDIM = 2;
@@ -1757,17 +1739,16 @@ void PAHdivL2AssembleDiagonal_ADAt_2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHdivL2AssembleDiagonal_ADAt_3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &D_,
Vector &diag_)
inline void PAHdivL2AssembleDiagonal_ADAt_3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &D_,
Vector &diag_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -1885,17 +1866,16 @@ void PAHdivL2AssembleDiagonal_ADAt_3D(const int D1D,
// Apply to x corresponding to DOFs in H(div) (trial), whose divergence is
// integrated against L_2 test functions corresponding to y.
MFEM_HOST_DEVICE inline
void PAHdivL2Apply3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &L2Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivL2Apply3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &L2Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -2049,17 +2029,16 @@ void PAHdivL2Apply3D(const int D1D,
// Apply to x corresponding to DOFs in H(div) (trial), whose divergence is
// integrated against L_2 test functions corresponding to y.
MFEM_HOST_DEVICE inline
void PAHdivL2Apply2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &L2Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivL2Apply2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &Bo_,
const Array<double> &Gc_,
const Array<double> &L2Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HDIV_MAX_D1D;
@@ -2156,17 +2135,16 @@ void PAHdivL2Apply2D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHdivL2ApplyTranspose3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivL2ApplyTranspose3D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
MFEM_VERIFY(D1D <= HDIV_MAX_D1D, "Error: D1D > HDIV_MAX_D1D");
MFEM_VERIFY(Q1D <= HDIV_MAX_Q1D, "Error: Q1D > HDIV_MAX_Q1D");
@@ -2319,17 +2297,16 @@ void PAHdivL2ApplyTranspose3D(const int D1D,
}); // end of element loop
}
MFEM_HOST_DEVICE inline
void PAHdivL2ApplyTranspose2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
inline void PAHdivL2ApplyTranspose2D(const int D1D,
const int Q1D,
const int L2D1D,
const int NE,
const Array<double> &L2Bo_,
const Array<double> &Gct_,
const Array<double> &Bot_,
const Vector &op_,
const Vector &x_,
Vector &y_)
{
constexpr static int VDIM = 2;
constexpr static int MAX_D1D = HDIV_MAX_D1D;
fem/integ/bilininteg_mass_kernels.hpp
+86 -98
View File
@@ -23,13 +23,12 @@ namespace internal
{
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PAMassAssembleDiagonal2D(const int NE,
const Array<double> &b,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
inline void PAMassAssembleDiagonal2D(const int NE,
const Array<double> &b,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -70,13 +69,12 @@ void PAMassAssembleDiagonal2D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0, int T_NBZ = 0>
MFEM_HOST_DEVICE inline
void SmemPAMassAssembleDiagonal2D(const int NE,
const Array<double> &b_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAMassAssembleDiagonal2D(const int NE,
const Array<double> &b_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -137,13 +135,12 @@ void SmemPAMassAssembleDiagonal2D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PAMassAssembleDiagonal3D(const int NE,
const Array<double> &b,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
inline void PAMassAssembleDiagonal3D(const int NE,
const Array<double> &b,
const Vector &d,
Vector &y,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -207,13 +204,12 @@ void PAMassAssembleDiagonal3D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPAMassAssembleDiagonal3D(const int NE,
const Array<double> &b_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAMassAssembleDiagonal3D(const int NE,
const Array<double> &b_,
const Vector &d_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
const int D1D = T_D1D ? T_D1D : d1d;
const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -293,12 +289,11 @@ void SmemPAMassAssembleDiagonal3D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PAMassAssembleDiagonal(const int dim, const int D1D,
const int Q1D, const int NE,
const Array<double> &B,
const Vector &D,
Vector &Y)
inline void PAMassAssembleDiagonal(const int dim, const int D1D,
const int Q1D, const int NE,
const Array<double> &B,
const Vector &D,
Vector &Y)
{
if (dim == 2)
{
@@ -338,15 +333,14 @@ void PAMassAssembleDiagonal(const int dim, const int D1D,
#ifdef MFEM_USE_OCCA
// OCCA PA Mass Apply 2D kernel
MFEM_HOST_DEVICE inline
void OccaPAMassApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void OccaPAMassApply2D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -384,15 +378,14 @@ void OccaPAMassApply2D(const int D1D,
}
// OCCA PA Mass Apply 3D kernel
MFEM_HOST_DEVICE inline
void OccaPAMassApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void OccaPAMassApply3D(const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
{
occa::properties props;
props["defines/D1D"] = D1D;
@@ -1034,15 +1027,14 @@ void SmemPAMassApply3D_Element(const int e,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PAMassApply2D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void PAMassApply2D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_VERIFY(T_D1D ? T_D1D : d1d <= MAX_D1D, "");
MFEM_VERIFY(T_Q1D ? T_Q1D : q1d <= MAX_Q1D, "");
@@ -1060,15 +1052,14 @@ void PAMassApply2D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0, int T_NBZ = 0>
MFEM_HOST_DEVICE inline
void SmemPAMassApply2D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAMassApply2D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_CONTRACT_VAR(bt_);
const int D1D = T_D1D ? T_D1D : d1d;
@@ -1090,15 +1081,14 @@ void SmemPAMassApply2D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void PAMassApply3D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void PAMassApply3D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_VERIFY(T_D1D ? T_D1D : d1d <= MAX_D1D, "");
MFEM_VERIFY(T_Q1D ? T_Q1D : q1d <= MAX_Q1D, "");
@@ -1116,15 +1106,14 @@ void PAMassApply3D(const int NE,
}
template<int T_D1D = 0, int T_Q1D = 0>
MFEM_HOST_DEVICE inline
void SmemPAMassApply3D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
inline void SmemPAMassApply3D(const int NE,
const Array<double> &b_,
const Array<double> &bt_,
const Vector &d_,
const Vector &x_,
Vector &y_,
const int d1d = 0,
const int q1d = 0)
{
MFEM_CONTRACT_VAR(bt_);
const int D1D = T_D1D ? T_D1D : d1d;
@@ -1143,16 +1132,15 @@ void SmemPAMassApply3D(const int NE,
});
}
MFEM_HOST_DEVICE inline
void PAMassApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
inline void PAMassApply(const int dim,
const int D1D,
const int Q1D,
const int NE,
const Array<double> &B,
const Array<double> &Bt,
const Vector &D,
const Vector &X,
Vector &Y)
{
#ifdef MFEM_USE_OCCA
if (DeviceCanUseOcca())