Merge branch 'master' of https://github.com/prusa3d/PrusaSlicer into et_gcode_viewer

2020-05-22 12:12:56 +02:00
parent 83ea38c2f3 c09d702045
commit 80c2f107c1
28 changed files with 1208 additions and 52 deletions
--- a/src/libslic3r/AABBTreeIndirect.hpp
+++ b/src/libslic3r/AABBTreeIndirect.hpp
@@ -0,0 +1,698 @@
+// AABB tree built upon external data set, referencing the external data by integer indices.
+// The AABB tree balancing and traversal (ray casting, closest triangle of an indexed triangle mesh)
+// were adapted from libigl AABB.{cpp,hpp} Copyright (C) 2015 Alec Jacobson <alecjacobson@gmail.com>
+// while the implicit balanced tree representation and memory optimizations are Vojtech's.
+
+#ifndef slic3r_AABBTreeIndirect_hpp_
+#define slic3r_AABBTreeIndirect_hpp_
+
+#include <algorithm>
+#include <limits>
+#include <type_traits>
+#include <vector>
+
+#include "Utils.hpp" // for next_highest_power_of_2()
+
+extern "C"
+{
+// Ray-Triangle Intersection Test Routines by Tomas Moller, May 2000
+#include <igl/raytri.c>
+}
+// Definition of the ray intersection hit structure.
+#include <igl/Hit.h>
+
+namespace Slic3r {
+namespace AABBTreeIndirect {
+
+// Static balanced AABB tree for raycasting and closest triangle search.
+// The balanced tree is built over a single large std::vector of nodes, where the children of nodes
+// are addressed implicitely using a power of two indexing rule.
+// Memory for a full balanced tree is allocated, but not all nodes at the last level are used.
+// This may seem like a waste of memory, but one saves memory for the node links and there is zero
+// overhead of a memory allocator management (usually the memory allocator adds at least one pointer 
+// before the memory returned). However, allocating memory in a single vector is very fast even
+// in multi-threaded environment and it is cache friendly.
+//
+// A balanced tree is built upon a vector of bounding boxes and their centroids, storing the reference
+// to the source entity (a 3D triangle, a 2D segment etc, a 3D or 2D point etc).
+// The source bounding boxes may have an epsilon applied to fight numeric rounding errors when 
+// traversing the AABB tree.
+template<int ANumDimensions, typename ACoordType>
+class Tree
+{
+public:
+    static constexpr int    NumDimensions = ANumDimensions;
+	using					CoordType     = ACoordType;
+    using 					VectorType 	  = Eigen::Matrix<CoordType, NumDimensions, 1, Eigen::DontAlign>;
+    using  					BoundingBox   = Eigen::AlignedBox<CoordType, NumDimensions>;
+    // Following could be static constexpr size_t, but that would not link in C++11
+    enum : size_t {
+    	// Node is not used.
+        npos  = size_t(-1),
+        // Inner node (not leaf).
+        inner = size_t(-2)
+    };
+
+    // Single node of the implicit balanced AABB tree. There are no links to the children nodes,
+    // as these links are calculated implicitely using a power of two rule.
+    struct Node {
+    	// Index of the external source entity, for which this AABB tree was built, npos for internal nodes.
+        size_t 			idx = npos;
+    	// Bounding box around this entity, possibly with epsilons applied to fight numeric rounding errors
+    	// when traversing the AABB tree.
+        BoundingBox 	bbox;
+
+        bool 	is_valid() const { return this->idx != npos; }
+        bool 	is_inner() const { return this->idx == inner; }
+        bool 	is_leaf()  const { return ! this->is_inner(); }
+
+    	template<typename SourceNode>
+    	void set(const SourceNode &rhs) {
+            this->idx  = rhs.idx();
+            this->bbox = rhs.bbox();
+		}
+    };
+
+	void clear() { m_nodes.clear(); }
+
+	// SourceNode shall implement
+	// size_t SourceNode::idx() const
+	// 		- Index to the outside entity (triangle, edge, point etc).
+	// const VectorType& SourceNode::centroid() const
+	// 		- Centroid of this node. The centroid is used for balancing the tree.
+	// const BoundingBox& SourceNode::bbox() const
+	// 		- Bounding box of this node, likely expanded with epsilon to account for numeric rounding during tree traversal.
+	//        Union of bounding boxes at a single level of the AABB tree is used for deciding the longest axis aligned dimension
+	//        to split around.
+	template<typename SourceNode>
+	void build(std::vector<SourceNode> &&input)
+	{
+        if (input.empty())
+			clear();
+		else {
+			// Allocate enough memory for a full binary tree.
+            m_nodes.assign(next_highest_power_of_2(input.size()) * 2 - 1, Node());
+            build_recursive(input, 0, 0, input.size() - 1);
+		}
+        input.clear();
+	}
+
+	const std::vector<Node>& 	nodes() const { return m_nodes; }
+	const Node& 				node(size_t idx) const { return m_nodes[idx]; }
+	bool 						empty() const { return m_nodes.empty(); }
+
+	// Addressing the child nodes using the power of two rule.
+    static size_t				left_child_idx(size_t idx) { return idx * 2 + 1; }
+    static size_t				right_child_idx(size_t idx) { return left_child_idx(idx) + 1; }
+	const Node&					left_child(size_t idx) const { return m_nodes[left_child_idx(idx)]; }
+	const Node&					right_child(size_t idx) const { return m_nodes[right_child_idx(idx)]; }
+
+	template<typename SourceNode>
+    void build(const std::vector<SourceNode> &input)
+	{
+        std::vector<SourceNode> copy(input);
+        this->build(std::move(copy));
+	}
+
+private:
+	// Build a balanced tree by splitting the input sequence by an axis aligned plane at a dimension.
+	template<typename SourceNode>
+	void build_recursive(std::vector<SourceNode> &input, size_t node, const size_t left, const size_t right)
+	{
+        assert(node < m_nodes.size());
+        assert(left <= right);
+
+		if (left == right) {
+			// Insert a node into the balanced tree.
+			m_nodes[node].set(input[left]);
+			return;
+		}
+
+		// Calculate bounding box of the input.
+        BoundingBox bbox(input[left].bbox());
+        for (size_t i = left + 1; i <= right; ++ i)
+            bbox.extend(input[i].bbox());
+        int dimension = -1;
+        bbox.diagonal().maxCoeff(&dimension);
+
+		// Partition the input to left / right pieces of the same length to produce a balanced tree.
+		size_t center = (left + right) / 2;
+		partition_input(input, size_t(dimension), left, right, center);
+		// Insert an inner node into the tree. Inner node does not reference any input entity (triangle, line segment etc).
+		m_nodes[node].idx  = inner;
+		m_nodes[node].bbox = bbox;
+        build_recursive(input, node * 2 + 1, left, center);
+		build_recursive(input, node * 2 + 2, center + 1, right);
+	}
+
+	// Partition the input m_nodes <left, right> at "k" and "dimension" using the QuickSelect method:
+	// https://en.wikipedia.org/wiki/Quickselect
+	// Items left of the k'th item are lower than the k'th item in the "dimension", 
+	// items right of the k'th item are higher than the k'th item in the "dimension", 
+	template<typename SourceNode>
+	void partition_input(std::vector<SourceNode> &input, const size_t dimension, size_t left, size_t right, const size_t k) const
+	{
+		while (left < right) {
+			size_t center = (left + right) / 2;
+			CoordType pivot;
+			{
+				// Bubble sort the input[left], input[center], input[right], so that a median of the three values
+				// will end up in input[center].
+				CoordType left_value   = input[left  ].centroid()(dimension);
+				CoordType center_value = input[center].centroid()(dimension);
+				CoordType right_value  = input[right ].centroid()(dimension);
+				if (left_value > center_value) {
+					std::swap(input[left], input[center]);
+					std::swap(left_value,  center_value);
+				}
+				if (left_value > right_value) {
+					std::swap(input[left], input[right]);
+					right_value = left_value;
+				}
+				if (center_value > right_value) {
+					std::swap(input[center], input[right]);
+					center_value = right_value;
+				}
+				pivot = center_value;
+			}
+			if (right <= left + 2)
+				// The <left, right> interval is already sorted.
+				break;
+			size_t i = left;
+			size_t j = right - 1;
+			std::swap(input[center], input[j]);
+			// Partition the set based on the pivot.
+			for (;;) {
+				// Skip left points that are already at correct positions.
+				// Search will certainly stop at position (right - 1), which stores the pivot.
+				while (input[++ i].centroid()(dimension) < pivot) ;
+				// Skip right points that are already at correct positions.
+				while (input[-- j].centroid()(dimension) > pivot && i < j) ;
+				if (i >= j)
+					break;
+				std::swap(input[i], input[j]);
+			}
+			// Restore pivot to the center of the sequence.
+			std::swap(input[i], input[right - 1]);
+			// Which side the kth element is in?
+			if (k < i)
+				right = i - 1;
+			else if (k == i)
+				// Sequence is partitioned, kth element is at its place.
+				break;
+			else
+				left = i + 1;
+		}
+	}
+
+	// The balanced tree storage.
+	std::vector<Node> m_nodes;
+};
+
+using Tree2f = Tree<2, float>;
+using Tree3f = Tree<3, float>;
+using Tree2d = Tree<2, double>;
+using Tree3d = Tree<3, double>;
+
+namespace detail {
+	template<typename AVertexType, typename AIndexedFaceType, typename ATreeType, typename AVectorType>
+	struct RayIntersector {
+		using VertexType 		= AVertexType;
+		using IndexedFaceType 	= AIndexedFaceType;
+		using TreeType			= ATreeType;
+		using VectorType 		= AVectorType;
+
+		const std::vector<VertexType> 		&vertices;
+		const std::vector<IndexedFaceType> 	&faces;
+		const TreeType 						&tree;
+
+		const VectorType					 origin;
+		const VectorType 					 dir;
+		const VectorType					 invdir;
+	};
+
+    template<typename VertexType, typename IndexedFaceType, typename TreeType, typename VectorType>
+    struct RayIntersectorHits : RayIntersector<VertexType, IndexedFaceType, TreeType, VectorType> {
+		std::vector<igl::Hit>				 hits;
+	};
+
+	//FIXME implement SSE for float AABB trees with float ray queries.
+	// SSE/SSE2 is supported by any Intel/AMD x64 processor.
+	// SSE support requires 16 byte alignment of the AABB nodes, representing the bounding boxes with 4+4 floats,
+	// storing the node index as the 4th element of the bounding box min value etc.
+	// https://www.flipcode.com/archives/SSE_RayBox_Intersection_Test.shtml
+	template <typename Derivedsource, typename Deriveddir, typename Scalar>
+	inline bool ray_box_intersect_invdir(
+  		const Eigen::MatrixBase<Derivedsource> 	&origin,
+  		const Eigen::MatrixBase<Deriveddir> 	&inv_dir,
+  		Eigen::AlignedBox<Scalar,3> 			 box,
+  		const Scalar 							&t0,
+  		const Scalar 							&t1) {
+		// http://people.csail.mit.edu/amy/papers/box-jgt.pdf
+		// "An Efficient and Robust Ray–Box Intersection Algorithm"
+		if (inv_dir.x() < 0)
+			std::swap(box.min().x(), box.max().x());
+		if (inv_dir.y() < 0)
+			std::swap(box.min().y(), box.max().y());
+        Scalar tmin = (box.min().x() - origin.x()) * inv_dir.x();
+		Scalar tymax = (box.max().y() - origin.y()) * inv_dir.y();
+		if (tmin > tymax)
+			return false;
+        Scalar tmax = (box.max().x() - origin.x()) * inv_dir.x();
+		Scalar tymin = (box.min().y()  - origin.y()) * inv_dir.y();
+		if (tymin > tmax)
+			return false;
+		if (tymin > tmin)
+			tmin = tymin;
+		if (tymax < tmax)
+			tmax = tymax;
+		if (inv_dir.z() < 0)
+			std::swap(box.min().z(), box.max().z());
+		Scalar tzmin = (box.min().z()  - origin.z()) * inv_dir.z();
+		if (tzmin > tmax)
+			return false;
+		Scalar tzmax = (box.max().z() - origin.z()) * inv_dir.z();
+		if (tmin > tzmax)
+			return false;
+		if (tzmin > tmin)
+			tmin = tzmin;
+		if (tzmax < tmax)
+			tmax = tzmax;
+        return tmin < t1 && tmax > t0;
+	}
+
+	template<typename V, typename W>
+    std::enable_if_t<std::is_same<typename V::Scalar, double>::value && std::is_same<typename W::Scalar, double>::value, bool>
+	intersect_triangle(const V &origin, const V &dir, const W &v0, const W &v1, const W v2, double &t, double &u, double &v) {
+        return intersect_triangle1(const_cast<double*>(origin.data()), const_cast<double*>(dir.data()),
+                                   const_cast<double*>(v0.data()), const_cast<double*>(v1.data()), const_cast<double*>(v2.data()),
+                                   &t, &u, &v);
+	}
+
+	template<typename V, typename W>
+    std::enable_if_t<std::is_same<typename V::Scalar, double>::value && !std::is_same<typename W::Scalar, double>::value, bool>
+	intersect_triangle(const V &origin, const V &dir, const W &v0, const W &v1, const W v2, double &t, double &u, double &v) {
+        using Vector = Eigen::Matrix<double, 3, 1>;
+        Vector w0 = v0.template cast<double>();
+        Vector w1 = v1.template cast<double>();
+        Vector w2 = v2.template cast<double>();
+        return intersect_triangle1(const_cast<double*>(origin.data()), const_cast<double*>(dir.data()),
+                                   w0.data(), w1.data(), w2.data(), &t, &u, &v);
+	}
+
+	template<typename V, typename W>
+    std::enable_if_t<! std::is_same<typename V::Scalar, double>::value && std::is_same<typename W::Scalar, double>::value, bool>
+	intersect_triangle(const V &origin, const V &dir, const W &v0, const W &v1, const W v2, double &t, double &u, double &v) {
+        using Vector = Eigen::Matrix<double, 3, 1>;
+        Vector o  = origin.template cast<double>();
+        Vector d  = dir.template cast<double>();
+        return intersect_triangle1(o.data(), d.data(), const_cast<double*>(v0.data()), const_cast<double*>(v1.data()), const_cast<double*>(v2.data()), &t, &u, &v);
+	}
+
+	template<typename V, typename W>
+    std::enable_if_t<! std::is_same<typename V::Scalar, double>::value && ! std::is_same<typename W::Scalar, double>::value, bool>
+	intersect_triangle(const V &origin, const V &dir, const W &v0, const W &v1, const W v2, double &t, double &u, double &v) {
+        using Vector = Eigen::Matrix<double, 3, 1>;
+        Vector o  = origin.template cast<double>();
+        Vector d  = dir.template cast<double>();
+        Vector w0 = v0.template cast<double>();
+        Vector w1 = v1.template cast<double>();
+        Vector w2 = v2.template cast<double>();
+	    return intersect_triangle1(o.data(), d.data(), w0.data(), w1.data(), w2.data(), &t, &u, &v);
+	}
+
+    template<typename RayIntersectorType, typename Scalar>
+	static inline bool intersect_ray_recursive_first_hit(
+        RayIntersectorType 	   &ray_intersector,
+        size_t 				    node_idx,
+        Scalar                  min_t,
+        igl::Hit 			   &hit)
+	{
+        const auto &node = ray_intersector.tree.node(node_idx);
+        assert(node.is_valid());
+		
+        if (! ray_box_intersect_invdir(ray_intersector.origin, ray_intersector.invdir, node.bbox.template cast<Scalar>(), Scalar(0), min_t))
+			return false;
+
+	  	if (node.is_leaf()) {
+		    // shoot ray, record hit
+            auto   face = ray_intersector.faces[node.idx];
+		    double t, u, v;
+		    if (intersect_triangle(
+		    		ray_intersector.origin, ray_intersector.dir, 
+		    		ray_intersector.vertices[face(0)], ray_intersector.vertices[face(1)], ray_intersector.vertices[face(2)], 
+                    t, u, v)
+		    	&& t > 0.) {
+                hit = igl::Hit { int(node.idx), -1, float(u), float(v), float(t) };
+				return true;
+		    } else
+		    	return false;
+	  	} else {
+			// Left / right child node index.
+			size_t left  = node_idx * 2 + 1;
+			size_t right = left + 1;
+			igl::Hit left_hit;
+			igl::Hit right_hit;
+	        bool left_ret = intersect_ray_recursive_first_hit(ray_intersector, left,  min_t, left_hit);
+			if (left_ret && left_hit.t < min_t) {
+	    		min_t = left_hit.t;
+	    		hit   = left_hit;
+	  		} else
+			    left_ret = false;
+	        bool right_ret = intersect_ray_recursive_first_hit(ray_intersector, right, min_t, right_hit);
+			if (right_ret && right_hit.t < min_t)
+				hit = right_hit;
+			else
+				right_ret = false;
+			return left_ret || right_ret;
+		}
+	}
+
+    template<typename RayIntersectorType>
+	static inline void intersect_ray_recursive_all_hits(RayIntersectorType &ray_intersector, size_t node_idx)
+	{
+        using Scalar = typename RayIntersectorType::VectorType::Scalar;
+
+		const auto &node = ray_intersector.tree.node(node_idx);
+		assert(node.is_valid());
+
+        if (! ray_box_intersect_invdir(ray_intersector.origin, ray_intersector.invdir, node.bbox.template cast<Scalar>(),
+    			Scalar(0), std::numeric_limits<Scalar>::infinity()))
+			return;
+
+	  	if (node.is_leaf()) {
+            auto   face = ray_intersector.faces[node.idx];
+		    double t, u, v;
+		    if (intersect_triangle(
+		    		ray_intersector.origin, ray_intersector.dir, 
+		    		ray_intersector.vertices[face(0)], ray_intersector.vertices[face(1)], ray_intersector.vertices[face(2)], 
+                    t, u, v)
+		    	&& t > 0.) {
+                ray_intersector.hits.emplace_back(igl::Hit{ int(node.idx), -1, float(u), float(v), float(t) });
+			}
+	  	} else {
+			// Left / right child node index.
+			size_t left  = node_idx * 2 + 1;
+			size_t right = left + 1;
+		  	intersect_ray_recursive_all_hits(ray_intersector, left);
+		  	intersect_ray_recursive_all_hits(ray_intersector, right);
+		}
+	}
+
+	// Nothing to do with COVID-19 social distancing.
+	template<typename AVertexType, typename AIndexedFaceType, typename ATreeType, typename AVectorType>
+	struct IndexedTriangleSetDistancer {
+		using VertexType 		= AVertexType;
+		using IndexedFaceType 	= AIndexedFaceType;
+		using TreeType			= ATreeType;
+		using VectorType 		= AVectorType;
+
+		const std::vector<VertexType> 		&vertices;
+		const std::vector<IndexedFaceType> 	&faces;
+		const TreeType 						&tree;
+
+		const VectorType					 origin;
+	};
+
+	// Real-time collision detection, Ericson, Chapter 5
+	template<typename Vector>
+	static inline Vector closest_point_to_triangle(const Vector &p, const Vector &a, const Vector &b, const Vector &c)
+	{
+		using Scalar = typename Vector::Scalar;
+		// Check if P in vertex region outside A
+		Vector ab = b - a;
+		Vector ac = c - a;
+		Vector ap = p - a;
+		Scalar d1 = ab.dot(ap);
+		Scalar d2 = ac.dot(ap);
+		if (d1 <= 0 && d2 <= 0)
+		  return a;
+		// Check if P in vertex region outside B
+		Vector bp = p - b;
+		Scalar d3 = ab.dot(bp);
+		Scalar d4 = ac.dot(bp);
+		if (d3 >= 0 && d4 <= d3)
+		  return b;
+		// Check if P in edge region of AB, if so return projection of P onto AB
+		Scalar vc = d1*d4 - d3*d2;
+		if (a != b && vc <= 0 && d1 >= 0 && d3 <= 0) {
+		    Scalar v = d1 / (d1 - d3);
+		    return a + v * ab;
+		}
+		// Check if P in vertex region outside C
+		Vector cp = p - c;
+		Scalar d5 = ab.dot(cp);
+		Scalar d6 = ac.dot(cp);
+		if (d6 >= 0 && d5 <= d6)
+		  return c;
+		// Check if P in edge region of AC, if so return projection of P onto AC
+		Scalar vb = d5*d2 - d1*d6;
+		if (vb <= 0 && d2 >= 0 && d6 <= 0) {
+		  Scalar w = d2 / (d2 - d6);
+		  return a + w * ac;
+		}
+		// Check if P in edge region of BC, if so return projection of P onto BC
+		Scalar va = d3*d6 - d5*d4;
+		if (va <= 0 && (d4 - d3) >= 0 && (d5 - d6) >= 0) {
+		  Scalar w = (d4 - d3) / ((d4 - d3) + (d5 - d6));
+		  return b + w * (c - b);
+		}
+		// P inside face region. Compute Q through its barycentric coordinates (u,v,w)
+		Scalar denom = Scalar(1.0) / (va + vb + vc);
+		Scalar v = vb * denom;
+		Scalar w = vc * denom;
+		return a + ab * v + ac * w; // = u*a + v*b + w*c, u = va * denom = 1.0-v-w
+	};
+
+	template<typename IndexedTriangleSetDistancerType, typename Scalar>
+    static inline Scalar squared_distance_to_indexed_triangle_set_recursive(
+        IndexedTriangleSetDistancerType	&distancer,
+		size_t 							 node_idx,
+		Scalar 							 low_sqr_d,
+  		Scalar 							 up_sqr_d,
+		size_t 							&i,
+  		Eigen::PlainObjectBase<typename IndexedTriangleSetDistancerType::VectorType> &c)
+	{
+		using Vector = typename IndexedTriangleSetDistancerType::VectorType;
+
+  		if (low_sqr_d > up_sqr_d)
+			return low_sqr_d;
+	  
+	  	// Save the best achieved hit.
+        auto set_min = [&i, &c, &up_sqr_d](const Scalar sqr_d_candidate, const size_t i_candidate, const Vector &c_candidate) {
+			if (sqr_d_candidate < up_sqr_d) {
+				i     	 = i_candidate;
+				c     	 = c_candidate;
+				up_sqr_d = sqr_d_candidate;
+			}
+        };
+
+		const auto &node = distancer.tree.node(node_idx);
+		assert(node.is_valid());
+  		if (node.is_leaf()) 
+  		{
+            const auto &triangle = distancer.faces[node.idx];
+            Vector c_candidate = closest_point_to_triangle<Vector>(
+				distancer.origin, 
+                distancer.vertices[triangle(0)].template cast<Scalar>(),
+                distancer.vertices[triangle(1)].template cast<Scalar>(),
+                distancer.vertices[triangle(2)].template cast<Scalar>());
+            set_min((c_candidate - distancer.origin).squaredNorm(), node.idx, c_candidate);
+  		} 
+  		else
+  		{
+			size_t left_node_idx  = node_idx * 2 + 1;
+            size_t right_node_idx = left_node_idx + 1;
+			const auto &node_left  = distancer.tree.node(left_node_idx);
+			const auto &node_right = distancer.tree.node(right_node_idx);
+			assert(node_left.is_valid());
+			assert(node_right.is_valid());
+
+			bool   looked_left    = false;
+			bool   looked_right   = false;
+			const auto &look_left = [&]()
+			{
+                size_t	i_left;
+                Vector 	c_left = c;
+                Scalar	sqr_d_left = squared_distance_to_indexed_triangle_set_recursive(distancer, left_node_idx, low_sqr_d, up_sqr_d, i_left, c_left);
+				set_min(sqr_d_left, i_left, c_left);
+				looked_left = true;
+			};
+			const auto &look_right = [&]()
+			{
+                size_t	i_right;
+                Vector	c_right = c;
+                Scalar	sqr_d_right = squared_distance_to_indexed_triangle_set_recursive(distancer, right_node_idx, low_sqr_d, up_sqr_d, i_right, c_right);
+				set_min(sqr_d_right, i_right, c_right);
+				looked_right = true;
+			};
+
+			// must look left or right if in box
+            using BBoxScalar = typename IndexedTriangleSetDistancerType::TreeType::BoundingBox::Scalar;
+            if (node_left.bbox.contains(distancer.origin.template cast<BBoxScalar>()))
+			  	look_left();
+            if (node_right.bbox.contains(distancer.origin.template cast<BBoxScalar>()))
+			  	look_right();
+			// if haven't looked left and could be less than current min, then look
+			Scalar left_up_sqr_d  = node_left.bbox.squaredExteriorDistance(distancer.origin);
+			Scalar right_up_sqr_d = node_right.bbox.squaredExteriorDistance(distancer.origin);
+			if (left_up_sqr_d < right_up_sqr_d) {
+			  	if (! looked_left && left_up_sqr_d < up_sqr_d)
+			    	look_left();
+			  	if (! looked_right && right_up_sqr_d < up_sqr_d)
+			    	look_right();
+			} else {
+			  	if (! looked_right && right_up_sqr_d < up_sqr_d)
+			    	look_right();
+			  	if (! looked_left && left_up_sqr_d < up_sqr_d)
+			    	look_left();
+			}
+		}
+		return up_sqr_d;
+	}
+
+} // namespace detail
+
+// Build a balanced AABB Tree over an indexed triangles set, balancing the tree
+// on centroids of the triangles.
+// Epsilon is applied to the bounding boxes of the AABB Tree to cope with numeric inaccuracies
+// during tree traversal.
+template<typename VertexType, typename IndexedFaceType>
+inline Tree<3, typename VertexType::Scalar> build_aabb_tree_over_indexed_triangle_set(
+	// Indexed triangle set - 3D vertices.
+	const std::vector<VertexType> 		&vertices, 
+	// Indexed triangle set - triangular faces, references to vertices.
+    const std::vector<IndexedFaceType> 	&faces,
+	//FIXME do we want to apply an epsilon?
+    const typename VertexType::Scalar 	 eps = 0)
+{
+    using 				 TreeType 		= Tree<3, typename VertexType::Scalar>;
+//    using				 CoordType      = typename TreeType::CoordType;
+    using 				 VectorType	    = typename TreeType::VectorType;
+    using 				 BoundingBox 	= typename TreeType::BoundingBox;
+
+	struct InputType {
+        size_t 				idx()       const { return m_idx; }
+        const BoundingBox& 	bbox()      const { return m_bbox; }
+        const VectorType& 	centroid()  const { return m_centroid; }
+
+		size_t 		m_idx;
+		BoundingBox m_bbox;
+        VectorType 	m_centroid;
+	};
+
+	std::vector<InputType> input;
+	input.reserve(faces.size());
+    const VectorType veps(eps, eps, eps);
+	for (size_t i = 0; i < faces.size(); ++ i) {
+        const IndexedFaceType &face = faces[i];
+		const VertexType &v1 = vertices[face(0)];
+		const VertexType &v2 = vertices[face(1)];
+		const VertexType &v3 = vertices[face(2)];
+		InputType n;
+        n.m_idx      = i;
+        n.m_centroid = (1./3.) * (v1 + v2 + v3);
+        n.m_bbox = BoundingBox(v1, v1);
+        n.m_bbox.extend(v2);
+        n.m_bbox.extend(v3);
+        n.m_bbox.min() -= veps;
+        n.m_bbox.max() += veps;
+        input.emplace_back(n);
+	}
+
+	TreeType out;
+	out.build(std::move(input));
+	return out;
+}
+
+// Find a first intersection of a ray with indexed triangle set.
+// Intersection test is calculated with the accuracy of VectorType::Scalar
+// even if the triangle mesh and the AABB Tree are built with floats.
+template<typename VertexType, typename IndexedFaceType, typename TreeType, typename VectorType>
+inline bool intersect_ray_first_hit(
+	// Indexed triangle set - 3D vertices.
+	const std::vector<VertexType> 		&vertices,
+	// Indexed triangle set - triangular faces, references to vertices.
+	const std::vector<IndexedFaceType> 	&faces,
+	// AABBTreeIndirect::Tree over vertices & faces, bounding boxes built with the accuracy of vertices.
+	const TreeType 						&tree,
+	// Origin of the ray.
+	const VectorType					&origin,
+	// Direction of the ray.
+	const VectorType 					&dir,
+	// First intersection of the ray with the indexed triangle set.
+	igl::Hit 							&hit)
+{
+    using Scalar = typename VectorType::Scalar;
+    auto ray_intersector = detail::RayIntersector<VertexType, IndexedFaceType, TreeType, VectorType> {
+		vertices, faces, tree,
+        origin, dir, VectorType(dir.cwiseInverse())
+	};
+	return ! tree.empty() && detail::intersect_ray_recursive_first_hit(
+        ray_intersector, size_t(0), std::numeric_limits<Scalar>::infinity(), hit);
+}
+
+// Find all intersections of a ray with indexed triangle set.
+// Intersection test is calculated with the accuracy of VectorType::Scalar
+// even if the triangle mesh and the AABB Tree are built with floats.
+// The output hits are sorted by the ray parameter.
+// If the ray intersects a shared edge of two triangles, hits for both triangles are returned.
+template<typename VertexType, typename IndexedFaceType, typename TreeType, typename VectorType>
+inline bool intersect_ray_all_hits(
+	// Indexed triangle set - 3D vertices.
+	const std::vector<VertexType> 		&vertices,
+	// Indexed triangle set - triangular faces, references to vertices.
+	const std::vector<IndexedFaceType> 	&faces,
+	// AABBTreeIndirect::Tree over vertices & faces, bounding boxes built with the accuracy of vertices.
+	const TreeType 						&tree,
+	// Origin of the ray.
+	const VectorType					&origin,
+	// Direction of the ray.
+	const VectorType 					&dir,
+	// All intersections of the ray with the indexed triangle set, sorted by parameter t.
+	std::vector<igl::Hit> 				&hits)
+{
+    auto ray_intersector = detail::RayIntersectorHits<VertexType, IndexedFaceType, TreeType, VectorType> {
+		vertices, faces, tree,
+        origin, dir, VectorType(dir.cwiseInverse())
+	};
+	if (! tree.empty()) {
+        ray_intersector.hits.reserve(8);
+		detail::intersect_ray_recursive_all_hits(ray_intersector, 0);
+		std::swap(hits, ray_intersector.hits);
+	    std::sort(hits.begin(), hits.end(), [](const auto &l, const auto &r) { return l.t < r.t; });
+	}
+	return ! hits.empty();
+}
+
+// Finding a closest triangle, its closest point and squared distance to the closest point
+// on a 3D indexed triangle set using a pre-built AABBTreeIndirect::Tree.
+// Closest point to triangle test will be performed with the accuracy of VectorType::Scalar
+// even if the triangle mesh and the AABB Tree are built with floats.
+// Returns squared distance to the closest point or -1 if the input is empty.
+template<typename VertexType, typename IndexedFaceType, typename TreeType, typename VectorType>
+inline typename VectorType::Scalar squared_distance_to_indexed_triangle_set(
+	// Indexed triangle set - 3D vertices.
+	const std::vector<VertexType> 		&vertices,
+	// Indexed triangle set - triangular faces, references to vertices.
+	const std::vector<IndexedFaceType> 	&faces,
+	// AABBTreeIndirect::Tree over vertices & faces, bounding boxes built with the accuracy of vertices.
+	const TreeType 						&tree,
+	// Point to which the closest point on the indexed triangle set is searched for.
+	const VectorType					&point,
+	// Index of the closest triangle in faces.
+	size_t 								&hit_idx_out,
+	// Position of the closest point on the indexed triangle set.
+	Eigen::PlainObjectBase<VectorType>	&hit_point_out)
+{
+    using Scalar = typename VectorType::Scalar;
+    auto distancer = detail::IndexedTriangleSetDistancer<VertexType, IndexedFaceType, TreeType, VectorType>
+        { vertices, faces, tree, point };
+    return tree.empty() ? Scalar(-1) : 
+    	detail::squared_distance_to_indexed_triangle_set_recursive(distancer, size_t(0), Scalar(0), std::numeric_limits<Scalar>::infinity(), hit_idx_out, hit_point_out);
+}
+
+} // namespace AABBTreeIndirect
+} // namespace Slic3r
+
+#endif /* slic3r_AABBTreeIndirect_hpp_ */
--- a/src/libslic3r/Arrange.cpp
+++ b/src/libslic3r/Arrange.cpp
@@ -51,8 +51,8 @@ template<class S> struct NfpImpl<S, NfpLevel::CONVEX_ONLY>
 namespace Slic3r {

 template<class Tout = double, class = FloatingOnly<Tout>, int...EigenArgs>
-inline SLIC3R_CONSTEXPR Eigen::Matrix<Tout, 2, EigenArgs...> unscaled(
-    const ClipperLib::IntPoint &v) SLIC3R_NOEXCEPT
+inline constexpr Eigen::Matrix<Tout, 2, EigenArgs...> unscaled(
+    const ClipperLib::IntPoint &v) noexcept
 {
    return Eigen::Matrix<Tout, 2, EigenArgs...>{unscaled<Tout>(v.X),
                                                unscaled<Tout>(v.Y)};
--- a/src/libslic3r/CMakeLists.txt
+++ b/src/libslic3r/CMakeLists.txt
@@ -306,7 +306,7 @@ if(WIN32)
 endif()

 if(SLIC3R_PROFILE)
-    target_link_libraries(slic3r Shiny)
+    target_link_libraries(libslic3r Shiny)
 endif()

 if (SLIC3R_PCH AND NOT SLIC3R_SYNTAXONLY)
--- a/src/libslic3r/Point.hpp
+++ b/src/libslic3r/Point.hpp
@@ -369,7 +369,7 @@ namespace boost { namespace polygon {
        typedef coord_t coordinate_type;
    
        static inline coordinate_type get(const Slic3r::Point& point, orientation_2d orient) {
-            return (orient == HORIZONTAL) ? (coordinate_type)point(0) : (coordinate_type)point(1);
+            return (coordinate_type)point((orient == HORIZONTAL) ? 0 : 1);
        }
    };
    
@@ -377,16 +377,10 @@ namespace boost { namespace polygon {
    struct point_mutable_traits<Slic3r::Point> {
        typedef coord_t coordinate_type;
        static inline void set(Slic3r::Point& point, orientation_2d orient, coord_t value) {
-            if (orient == HORIZONTAL)
-                point(0) = value;
-            else
-                point(1) = value;
+            point((orient == HORIZONTAL) ? 0 : 1) = value;
        }
        static inline Slic3r::Point construct(coord_t x_value, coord_t y_value) {
-            Slic3r::Point retval;
-            retval(0) = x_value;
-            retval(1) = y_value; 
-            return retval;
+            return Slic3r::Point(x_value, y_value);
        }
    };
 } }
--- a/src/libslic3r/libslic3r.h
+++ b/src/libslic3r/libslic3r.h
@@ -24,37 +24,37 @@

 #if 1
 // Saves around 32% RAM after slicing step, 6.7% after G-code export (tested on PrusaSlicer 2.2.0 final).
-typedef int32_t coord_t;
+using coord_t = int32_t;
 #else
 //FIXME At least FillRectilinear2 requires coord_t to be 32bit.
 typedef int64_t coord_t;
 #endif

-typedef double  coordf_t;
+using coordf_t = double;

 //FIXME This epsilon value is used for many non-related purposes:
 // For a threshold of a squared Euclidean distance,
 // for a trheshold in a difference of radians,
 // for a threshold of a cross product of two non-normalized vectors etc.
-#define EPSILON 1e-4
+static constexpr double EPSILON = 1e-4;
 // Scaling factor for a conversion from coord_t to coordf_t: 10e-6
 // This scaling generates a following fixed point representation with for a 32bit integer:
 // 0..4294mm with 1nm resolution
 // int32_t fits an interval of (-2147.48mm, +2147.48mm)
 // with int64_t we don't have to worry anymore about the size of the int.
-#define SCALING_FACTOR 0.000001
+static constexpr double SCALING_FACTOR = 0.000001;
 // RESOLUTION, SCALED_RESOLUTION: Used as an error threshold for a Douglas-Peucker polyline simplification algorithm.
-#define RESOLUTION 0.0125
-#define SCALED_RESOLUTION (RESOLUTION / SCALING_FACTOR)
-#define PI 3.141592653589793238
+static constexpr double RESOLUTION = 0.0125;
+#define                 SCALED_RESOLUTION (RESOLUTION / SCALING_FACTOR)
+static constexpr double PI = 3.141592653589793238;
 // When extruding a closed loop, the loop is interrupted and shortened a bit to reduce the seam.
-#define LOOP_CLIPPING_LENGTH_OVER_NOZZLE_DIAMETER 0.15
+static constexpr double LOOP_CLIPPING_LENGTH_OVER_NOZZLE_DIAMETER = 0.15;
 // Maximum perimeter length for the loop to apply the small perimeter speed. 
-#define SMALL_PERIMETER_LENGTH (6.5 / SCALING_FACTOR) * 2 * PI
-#define INSET_OVERLAP_TOLERANCE 0.4
+#define                 SMALL_PERIMETER_LENGTH  ((6.5 / SCALING_FACTOR) * 2 * PI)
+static constexpr double INSET_OVERLAP_TOLERANCE = 0.4;
 // 3mm ring around the top / bottom / bridging areas.
 //FIXME This is quite a lot.
-#define EXTERNAL_INFILL_MARGIN 3.
+static constexpr double EXTERNAL_INFILL_MARGIN = 3.;
 //FIXME Better to use an inline function with an explicit return type.
 //inline coord_t scale_(coordf_t v) { return coord_t(floor(v / SCALING_FACTOR + 0.5f)); }
 #define scale_(val) ((val) / SCALING_FACTOR)
@@ -63,14 +63,6 @@ typedef double  coordf_t;

 #define SLIC3R_DEBUG_OUT_PATH_PREFIX "out/"

-#if defined(_MSC_VER) &&  _MSC_VER < 1900
-# define SLIC3R_CONSTEXPR
-# define SLIC3R_NOEXCEPT
-#else
-#define SLIC3R_CONSTEXPR constexpr
-#define SLIC3R_NOEXCEPT  noexcept
-#endif
-
 inline std::string debug_out_path(const char *name, ...)
 {
 	char buffer[2048];
@@ -92,11 +84,6 @@ inline std::string debug_out_path(const char *name, ...)
 #define UNUSED(x) (void)(x)
 #endif /* UNUSED */

-// Detect whether the compiler supports C++11 noexcept exception specifications.
-#if defined(_MSC_VER) && _MSC_VER < 1900
-    #define noexcept throw()
-#endif
-
 // Write slices as SVG images into out directory during the 2D processing of the slices.
 // #define SLIC3R_DEBUG_SLICE_PROCESSING