1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
|
/* Copyright 2019 Andrew Myers, Maxence Thevenet, Weiqun Zhang
*
*
* This file is part of WarpX.
*
* License: BSD-3-Clause-LBNL
*/
#include "BilinearFilter.H"
#include "Utils/WarpXProfilerWrapper.H"
#include <AMReX_Config.H>
#include <AMReX_Dim3.H>
#include <AMReX_GpuContainers.H>
#include <AMReX_GpuDevice.H>
#include <AMReX_IntVect.H>
#include <AMReX_REAL.H>
#include <AMReX_Vector.H>
#include <array>
#include <vector>
using namespace amrex;
namespace {
void compute_stencil(Gpu::DeviceVector<Real> &stencil, unsigned int npass)
{
Vector<Real> old_s(1u+npass,0.);
Vector<Real> new_s(1u+npass,0.);
old_s.at(0) = 1._rt;
int jmax = 1;
// Convolve the filter with itself npass times
int const lastpass = static_cast<int>(npass+1u);
for(int ipass=1; ipass< lastpass; ipass++){
// element 0 has to be treated in its own way
new_s.at(0) = 0.5_rt * old_s.at(0);
if (1<jmax) new_s.at(0) += 0.5_rt * old_s.at(1);
amrex::Real loc = 0._rt;
// For each element j, apply the filter to
// old_s to get new_s[j]. loc stores the tmp
// filtered value.
for(int j=1; j<jmax+1; j++){
loc = 0.5_rt * old_s[j];
loc += 0.25_rt * old_s[j-1];
if (j<jmax) loc += 0.25_rt * old_s.at(j+1);
new_s.at(j) = loc;
}
// copy new_s into old_s
old_s = new_s;
// extend the stencil length for next iteration
jmax += 1;
}
// we use old_s here to make sure the stencil
// is corrent even when npass = 0
old_s.at(0) *= 0.5_rt; // because we will use it twice
stencil.resize(old_s.size());
Gpu::copyAsync(Gpu::hostToDevice,old_s.begin(),old_s.end(),stencil.begin());
amrex::Gpu::synchronize();
}
}
void BilinearFilter::ComputeStencils(){
WARPX_PROFILE("BilinearFilter::ComputeStencils()");
int i = 0;
for (auto el : npass_each_dir )
stencil_length_each_dir[i++] = el;
stencil_length_each_dir += 1.;
#if defined(WARPX_DIM_3D)
// npass_each_dir = npass_x npass_y npass_z
stencil_x.resize( 1u + npass_each_dir[0] );
stencil_y.resize( 1u + npass_each_dir[1] );
stencil_z.resize( 1u + npass_each_dir[2] );
compute_stencil(stencil_x, npass_each_dir[0]);
compute_stencil(stencil_y, npass_each_dir[1]);
compute_stencil(stencil_z, npass_each_dir[2]);
#elif defined(WARPX_DIM_XZ) || defined(WARPX_DIM_RZ)
// npass_each_dir = npass_x npass_z
stencil_x.resize( 1u + npass_each_dir[0] );
stencil_z.resize( 1u + npass_each_dir[1] );
compute_stencil(stencil_x, npass_each_dir[0]);
compute_stencil(stencil_z, npass_each_dir[1]);
#elif defined(WARPX_DIM_1D_Z)
// npass_each_dir = npass_z
stencil_z.resize( 1u + npass_each_dir[0] );
compute_stencil(stencil_z, npass_each_dir[0]);
#endif
slen = stencil_length_each_dir.dim3();
#if defined(WARPX_DIM_XZ) || defined(WARPX_DIM_RZ)
slen.z = 1;
#endif
#if defined(WARPX_DIM_1D_Z)
slen.y = 1;
slen.z = 1;
#endif
}
|