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
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
|
/// \file TestNavIndex.cu
/// \author Andrei Gheata (andrei.gheata@cern.ch)
#include <VecGeom/base/Transformation3D.h>
#include <VecGeom/management/GeoManager.h>
#include <VecGeom/management/CudaManager.h>
#include <VecGeom/navigation/NavigationState.h>
#include <VecGeom/volumes/PlacedVolume.h>
#include <VecGeom/base/Stopwatch.h>
#include <iomanip>
#include <cassert>
#include <cstdio>
using namespace vecgeom;
void check_cuda(cudaError_t result, char const *const func, const char *const file, int const line)
{
if (result) {
fprintf(stderr, "CUDA error = %s at %s:%d\n", cudaGetErrorString(result), file, line);
cudaDeviceReset();
exit(1);
}
}
#define checkCudaErrors(val) check_cuda((val), #val, __FILE__, __LINE__)
namespace visitorcuda {
class GlobalToLocalVisitor {
private:
int fError = 0; ///< error code
int fNiter = 0; ///< number of iterations
public:
VECCORE_ATT_HOST_DEVICE
GlobalToLocalVisitor() {}
VECCORE_ATT_HOST_DEVICE
int GetError() const { return fError; }
VECCORE_ATT_HOST_DEVICE
int GetNiter() const { return fNiter; }
VECCORE_ATT_HOST_DEVICE
void apply(NavStatePath *state, NavIndex_t nav_index)
{
unsigned char level = state->GetLevel();
int dind = 0;
NavIndex_t nav_ind = 1;
VPlacedVolume const *pdaughter = nullptr;
for (int i = 1; i < level + 1; ++i) {
pdaughter = state->At(i);
dind = pdaughter->GetChildId();
if (dind < 0) {
fError = 1;
return;
}
nav_ind = NavStateIndex::PushImpl(nav_ind, pdaughter);
}
// Check if navigation index matches input
if (nav_ind != nav_index) {
fError = 2;
return;
}
// Check if the physical volume is correct
if (NavStateIndex::TopImpl(nav_ind) != state->Top()) {
fError = 3;
return;
}
// Check if the current level is valid
if (level != NavStateIndex::GetLevelImpl(nav_ind)) {
fError = 4;
return;
}
// Check if mother navigation index is consistent
if (level > 0 && nav_ind != NavStateIndex::PushImpl(NavStateIndex::PopImpl(nav_ind), pdaughter)) {
fError = 5;
return;
}
// Check if the number of daughters is correct
if (NavStateIndex::GetNdaughtersImpl(nav_ind) != state->Top()->GetDaughters().size()) {
fError = 6;
return;
}
Transformation3D trans, trans_nav_ind;
state->TopMatrix(trans);
NavStateIndex::TopMatrixImpl(nav_ind, trans_nav_ind);
if (!trans.operator==(trans_nav_ind)) {
fError = 7;
return;
}
// success
fError = 0;
fNiter++;
}
};
/// Traverses the geometry tree keeping track of the state context (volume path or navigation state)
/// and applies the injected Visitor
template <typename Visitor>
VECCORE_ATT_DEVICE
int visitAllPlacedVolumesPassNavIndex(VPlacedVolume const *currentvolume, Visitor *visitor, NavStatePath *state,
NavIndex_t nav_ind)
{
const char *errcodes[] = {"incompatible daughter pointer",
"navigation index mismatch",
"top placed volume pointer mismatch",
"level mismatch",
"navigation index inconsistency for Push/Pop",
"number of daughters mismatch",
"transformation matrix mismatch"};
constexpr int maxiter = 100000; // limit the maximum number of iterations (slow on 1 GPU thread)
if (currentvolume != NULL) {
state->Push(currentvolume);
visitor->apply(state, nav_ind);
auto ierr = visitor->GetError();
if (ierr) {
printf("=== EEE === TestNavIndex: %s\n", errcodes[ierr - 1]);
return ierr;
}
if (visitor->GetNiter() > maxiter) return 0;
for (auto daughter : currentvolume->GetDaughters()) {
auto nav_ind_d = NavStateIndex::PushImpl(nav_ind, daughter);
ierr = visitAllPlacedVolumesPassNavIndex(daughter, visitor, state, nav_ind_d);
if (ierr > 0) return ierr;
if (visitor->GetNiter() > maxiter) return 0;
}
state->Pop();
}
return 0;
}
} // namespace visitorcuda
__global__ void TestNavIndexGPUKernel(vecgeom::cuda::VPlacedVolume const *const gpu_world,
vecgeom::cuda::NavStatePath *const state, int *ierr)
{
using namespace visitorcuda;
state->Clear();
GlobalToLocalVisitor visitor;
NavIndex_t nav_ind_top = 1; // The navigation index corresponding to the world
*ierr = visitAllPlacedVolumesPassNavIndex(gpu_world, &visitor, state, nav_ind_top);
}
int TestNavIndexGPU(vecgeom::cxx::VPlacedVolume const *const world, int maxdepth)
{
// Load and synchronize the geometry on the GPU
size_t statesize = NavigationState::SizeOfInstance(maxdepth);
vecgeom::cxx::CudaManager::Instance().LoadGeometry(world);
vecgeom::cxx::CudaManager::Instance().Synchronize();
auto gpu_world = vecgeom::cxx::CudaManager::Instance().world_gpu();
assert(gpu_world && "GPU world volume is a null pointer");
char *input_buffer = nullptr;
checkCudaErrors(cudaMallocManaged((void **)&input_buffer, statesize));
auto state = NavStatePath::MakeInstanceAt(maxdepth, (void *)(input_buffer));
int ierr;
int *d_ierr;
cudaMalloc(&d_ierr, sizeof(int));
Stopwatch timer;
timer.Start();
TestNavIndexGPUKernel<<<1, 1>>>(gpu_world, state, d_ierr);
cudaMemcpy(&ierr, d_ierr, sizeof(int), cudaMemcpyDeviceToHost);
cudaFree(d_ierr);
checkCudaErrors(cudaGetLastError());
checkCudaErrors(cudaDeviceSynchronize());
checkCudaErrors(cudaFree(input_buffer));
auto tvalidate = timer.Stop();
if (!ierr) std::cout << "=== Info navigation table validation on GPU took: " << tvalidate << " sec.\n";
return ierr;
}
|
