mirror of
https://github.com/OPM/ResInsight.git
synced 2026-10-24 22:27:55 -05:00
Kristian Bendiksen
5ee62c61e1
computeFaultDistances faked a nearest-point query with cvf::BoundingBoxTree, which only supports box overlap: it grew a cell-sized box until it overlapped a fault face. For cells far from any fault this did many full tree traversals and then brute-forced the distance to a large, over-collected candidate set, and the answer was only approximate. This was slow on grids with ~10 million cells. The metric is the distance from each cell center to the nearest fault face center, which is exactly a nearest-point search. Replace the bounding-box tree and the grow-the-box loop with a nanoflann KD-tree built over the fault face centers and a single nearest-neighbor query per active cell. A zero-copy adaptor reads the existing std::vector<cvf::Vec3d> directly. The OpenMP loop is retained since nanoflann queries are read-only. The result is O(F log F) build plus O(N log F) queries, the full grid-node array copy is gone, and distances are now exact rather than approximate. Add nanoflann as a vcpkg dependency, wired in like Clipper2 as an imported target. Refs: https://github.com/OPM/ResInsight/issues/14128
116 lines
4.7 KiB
C++
116 lines
4.7 KiB
C++
/////////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// Copyright (C) 2026- Equinor ASA
|
|
//
|
|
// ResInsight is free software: you can redistribute it and/or modify
|
|
// it under the terms of the GNU General Public License as published by
|
|
// the Free Software Foundation, either version 3 of the License, or
|
|
// (at your option) any later version.
|
|
//
|
|
// ResInsight is distributed in the hope that it will be useful, but WITHOUT ANY
|
|
// WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
// FITNESS FOR A PARTICULAR PURPOSE.
|
|
//
|
|
// See the GNU General Public License at <http://www.gnu.org/licenses/gpl.html>
|
|
// for more details.
|
|
//
|
|
/////////////////////////////////////////////////////////////////////////////////
|
|
|
|
#include "RigFaultDistanceCalculator.h"
|
|
|
|
#include "RigActiveCellInfo.h"
|
|
#include "RigCell.h"
|
|
#include "RigFault.h"
|
|
#include "RigMainGrid.h"
|
|
|
|
#include <nanoflann.hpp>
|
|
|
|
#include <cmath>
|
|
#include <limits>
|
|
|
|
namespace
|
|
{
|
|
//--------------------------------------------------------------------------------------------------
|
|
/// Adaptor exposing the collected fault-face-center points to nanoflann without copying them.
|
|
/// cvf::Vec3d stores its three doubles contiguously, so ptr()[dim] is a direct coordinate lookup.
|
|
//--------------------------------------------------------------------------------------------------
|
|
struct FaceCenterCloud
|
|
{
|
|
const std::vector<cvf::Vec3d>& points;
|
|
|
|
inline size_t kdtree_get_point_count() const { return points.size(); }
|
|
inline double kdtree_get_pt( size_t idx, size_t dim ) const { return points[idx].ptr()[dim]; }
|
|
|
|
template <class BBOX>
|
|
bool kdtree_get_bbox( BBOX& ) const
|
|
{
|
|
return false;
|
|
}
|
|
};
|
|
|
|
using FaceCenterKdTree = nanoflann::KDTreeSingleIndexAdaptor<nanoflann::L2_Simple_Adaptor<double, FaceCenterCloud>, FaceCenterCloud, 3>;
|
|
} // namespace
|
|
|
|
//--------------------------------------------------------------------------------------------------
|
|
///
|
|
//--------------------------------------------------------------------------------------------------
|
|
void RigFaultDistanceCalculator::computeFaultDistances( const RigMainGrid* mainGrid,
|
|
const RigActiveCellInfo* activeCellInfo,
|
|
const std::vector<const RigFault*>& faultsToInclude,
|
|
std::vector<double>& resultValues )
|
|
{
|
|
if ( !mainGrid || !activeCellInfo ) return;
|
|
|
|
const auto activeCells = activeCellInfo->activeReservoirCellIndices();
|
|
if ( activeCells.empty() ) return;
|
|
|
|
if ( resultValues.size() < activeCells.size() )
|
|
{
|
|
resultValues.resize( activeCells.size(), std::numeric_limits<double>::infinity() );
|
|
}
|
|
|
|
// Collect fault face centers from the requested subset of faults.
|
|
std::vector<cvf::Vec3d> faultFaceCenters;
|
|
for ( const RigFault* fault : faultsToInclude )
|
|
{
|
|
if ( !fault ) continue;
|
|
for ( const RigFault::FaultFace& faultFace : fault->faultFaces() )
|
|
{
|
|
if ( faultFace.m_nativeReservoirCellIndex >= mainGrid->cellCount() ) continue;
|
|
const RigCell& cell = mainGrid->cell( faultFace.m_nativeReservoirCellIndex );
|
|
if ( cell.isInvalid() ) continue;
|
|
faultFaceCenters.push_back( cell.faceCenter( faultFace.m_nativeFace ) );
|
|
}
|
|
}
|
|
|
|
if ( faultFaceCenters.empty() ) return;
|
|
|
|
// Build a KD-tree over the fault face centers and query the nearest one for every active cell.
|
|
// This is an exact nearest-point search in O(log N) per cell, replacing the previous expanding
|
|
// bounding-box heuristic.
|
|
FaceCenterCloud cloud{ faultFaceCenters };
|
|
FaceCenterKdTree kdTree( 3, cloud, nanoflann::KDTreeSingleIndexAdaptorParams( 10 /* max leaf size */ ) );
|
|
kdTree.buildIndex();
|
|
|
|
#pragma omp parallel for
|
|
for ( int activeIndex = 0; activeIndex < static_cast<int>( activeCells.size() ); activeIndex++ )
|
|
{
|
|
auto cellIdx = activeCells[activeIndex];
|
|
if ( cellIdx.value() == cvf::UNDEFINED_SIZE_T ) continue;
|
|
|
|
const RigCell& cell = mainGrid->cell( cellIdx.value() );
|
|
if ( cell.isInvalid() ) continue;
|
|
|
|
const cvf::Vec3d cellCenter = cell.center();
|
|
const double queryPoint[] = { cellCenter.x(), cellCenter.y(), cellCenter.z() };
|
|
|
|
size_t nearestIndex = 0;
|
|
double nearestDistanceSquared = std::numeric_limits<double>::infinity();
|
|
nanoflann::KNNResultSet<double> resultSet( 1 );
|
|
resultSet.init( &nearestIndex, &nearestDistanceSquared );
|
|
kdTree.findNeighbors( resultSet, queryPoint );
|
|
|
|
resultValues[activeIndex] = std::sqrt( nearestDistanceSquared );
|
|
}
|
|
}
|