Cyclic ordering improvement (#14784)

src/libslic3r/GCode.cpp

@@ -3127,7 +3127,8 @@ void GCode::_do_export(Print& print, GCodeOutputStream &file, ThumbnailsGenerato
         // Therefore initialize the printing extruders from there.
         this->set_extruders(tool_ordering.all_extruders());
         print_object_instances_ordering =
-            // By default, order object instances using a nearest neighbor search.
+            // By default, order object instances using nearest-neighbor chaining plus
+            // 2-opt and crossing-removal post-processing.
             (print.config().print_order == PrintOrder::Default ? chain_print_object_instances(print)
             // Snake: serpentine row traversal + 2-opt
             : (print.config().print_order == PrintOrder::Snake ? chain_print_object_instances_snake(print)
@@ -5969,63 +5970,128 @@ LayerResult GCode::process_layer(
     if (m_farthest_point_timelapse.enabled)
         compute_farthest_point(layers, most_used_extruder, support_filaments);
 
-    std::map<unsigned int, std::vector<InstanceToPrint>> filament_to_print_instances;
+    // Per filament: instances to print, and the visit sequence over them. Island-level ordering
+    // may visit an instance more than once per layer; otherwise one visit per instance.
+    std::map<unsigned int, std::pair<std::vector<InstanceToPrint>, std::vector<InstanceVisit>>> filament_to_print_instances;
     {
+        // Order individual islands rather than whole instances. Off for by-object sequencing,
+        // sequential printing, and the explicit AsObjectList order, which tour whole instances.
+        const bool island_level_ordering = print.config().print_sequence != PrintSequence::ByObject &&
+            single_object_instance_idx == size_t(-1) &&
+            print.config().print_order != PrintOrder::AsObjectList;
         for (unsigned int filament_id : layer_tools.extruders) {
             auto objects_by_extruder_it = by_extruder.find(filament_id);
             if (objects_by_extruder_it == by_extruder.end()) continue;
 
+            auto &filament_plan = filament_to_print_instances[filament_id];
+
+            if (!island_level_ordering) {
+                // One visit per instance, printing all of its islands.
+                filament_plan.first = sort_print_object_instances(objects_by_extruder_it->second, layers, ordering, single_object_instance_idx);
+                filament_plan.second.reserve(filament_plan.first.size());
+                for (size_t i = 0; i < filament_plan.first.size(); ++i)
+                    filament_plan.second.push_back({i, {}, true});
+                continue;
+            }
+
             int   plate_idx = print.get_plate_index();
             Point wt_pos(print.config().wipe_tower_x.get_at(plate_idx), print.config().wipe_tower_y.get_at(plate_idx));
 
+            // Build the instances and one tour node per non-empty island (a single node for
+            // instances without chainable islands). Positions quantized to 1 mm so small
+            // centroid drift between layers still hits the tour cache below.
             std::vector<GCode::ObjectByExtruder> &objects_by_extruder = objects_by_extruder_it->second;
-            std::vector<const PrintObject *>      print_objects;
-            for (int obj_idx = 0; obj_idx < objects_by_extruder.size(); obj_idx++) {
-                auto &object_by_extruder = objects_by_extruder[obj_idx];
+            std::vector<InstanceToPrint> &instances = filament_plan.first;
+            std::vector<IslandOrderNode> nodes;
+            std::vector<size_t>          node_instances;
+            auto quantize_to_mm = [](const Point &pt) -> Point {
+                const coord_t grid = coord_t(scale_(1.));
+                // Round to the nearest 1 mm symmetrically (integer division truncates toward
+                // zero, which would make the bucket straddling the origin twice as wide).
+                auto q = [grid](coord_t v) -> coord_t {
+                    return ((v >= 0 ? v + grid / 2 : v - grid / 2) / grid) * grid;
+                };
+                return Point(q(pt.x()), q(pt.y()));
+            };
+            for (ObjectByExtruder &object_by_extruder : objects_by_extruder) {
                 if (object_by_extruder.islands.empty() && (object_by_extruder.support == nullptr || object_by_extruder.support->empty())) continue;
 
-                print_objects.push_back(print.get_object(obj_idx));
+                const size_t       layer_id     = &object_by_extruder - objects_by_extruder.data();
+                const PrintObject *print_object = layers[layer_id].original_object;
+                if (print_object == nullptr)
+                    continue;
+                const Layer *obj_layer = layers[layer_id].object_layer;
+                std::vector<ObjectByExtruder::Island> &islands = object_by_extruder.islands;
+                const bool islands_chainable = obj_layer != nullptr && islands.size() == obj_layer->lslices.size() + 1;
+                for (size_t instance_id = 0; instance_id < print_object->instances().size(); ++instance_id) {
+                    const size_t instance_idx = instances.size();
+                    instances.emplace_back(object_by_extruder, layer_id, *print_object, instance_id,
+                                           print_object->instances()[instance_id].model_instance->get_labeled_id());
+                    const Point &shift = print_object->instances()[instance_id].shift;
+                    const size_t first_node = nodes.size();
+                    if (islands_chainable)
+                        for (size_t i = 0; i + 1 < islands.size(); ++i)
+                            if (!islands[i].by_region.empty()) {
+                                nodes.push_back({print_object->id(), instance_id, i,
+                                                 quantize_to_mm(obj_layer->lslices[i].contour.centroid() + shift)});
+                                node_instances.emplace_back(instance_idx);
+                            }
+                    if (nodes.size() == first_node) {
+                        // No chainable islands: tour the whole instance as one stop.
+                        nodes.push_back({print_object->id(), instance_id, size_t(-1), quantize_to_mm(shift)});
+                        node_instances.emplace_back(instance_idx);
+                    }
+                }
             }
 
-            // Build cache key from sorted object IDs.
-            std::vector<ObjectID> obj_ids;
-            for (const PrintObject* po : print_objects)
-                obj_ids.emplace_back(po->id());
-            std::sort(obj_ids.begin(), obj_ids.end());
-
-            // Check cache: reuse ordering if filament + object set unchanged.
+            // Reuse the cached tour while this filament's island layout is unchanged.
             auto &cache_entry = m_ordering_cache[filament_id];
-            bool cache_hit = (cache_entry.first == obj_ids);
+            if (!(cache_entry.first == nodes)) {
+                cache_entry.first = nodes;
+                Points node_points;
+                node_points.reserve(nodes.size());
+                for (const IslandOrderNode &node : nodes)
+                    node_points.emplace_back(node.pos);
+                std::vector<size_t> tour = order_points_with_strategy(node_points, print.config().print_order, &wt_pos);
+                // Chained starting near the wipe tower, reversed so the layer ends near it.
+                std::reverse(tour.begin(), tour.end());
 
-            if (!cache_hit) {
-                // Compute fresh ordering and store in cache.
-                cache_entry.first = obj_ids;
-                cache_entry.second =
-                    print.config().print_order == PrintOrder::Snake
-                            ? chain_print_object_instances_snake(print_objects, &wt_pos)
-                            : (print.config().print_order == PrintOrder::BestOfStrategies
-                                ? chain_print_object_instances_best_of(print_objects, &wt_pos)
-                                    : chain_print_object_instances(print_objects, &wt_pos));
-            }
-
-            // Reverse a local copy; keep cached value intact for reuse.
-            std::vector<const PrintInstance *> new_ordering = cache_entry.second;
-            std::reverse(new_ordering.begin(), new_ordering.end());
-
-            if (print.config().print_sequence == PrintSequence::ByObject) {
-                filament_to_print_instances[filament_id] = sort_print_object_instances(objects_by_extruder_it->second, layers, ordering, single_object_instance_idx);
-            } else {
-                
-                // PrintSequence::ByLayer to use global ordering ( per object ordering ) if intra-layer order PrintOrder::AsObjectList is specified while keeping behaviour of PrintSequence::ByLayer
-                const std::vector<const PrintInstance*>* ordering_for_filament = (print.config().print_order == PrintOrder::AsObjectList && ordering != nullptr) ? ordering: &new_ordering;
-                filament_to_print_instances[filament_id] = sort_print_object_instances(objects_by_extruder_it->second, layers, ordering_for_filament, single_object_instance_idx);
+                // Group consecutive tour stops of the same instance into visits.
+                std::vector<InstanceVisit> visits;
+                std::vector<bool> instance_seen(instances.size(), false);
+                std::vector<int>  last_visit_of_instance(instances.size(), -1);
+                for (size_t node_idx : tour) {
+                    const size_t instance_idx = node_instances[node_idx];
+                    if (visits.empty() || visits.back().instance_idx != instance_idx) {
+                        visits.push_back({instance_idx, {}, !instance_seen[instance_idx]});
+                        instance_seen[instance_idx] = true;
+                    }
+                    if (nodes[node_idx].island_idx != size_t(-1))
+                        visits.back().islands.emplace_back(nodes[node_idx].island_idx);
+                    last_visit_of_instance[instance_idx] = int(visits.size()) - 1;
+                }
+                // The trailing catch-all island has no geometry to chain by; append it to the
+                // instance's last visit.
+                for (size_t i = 0; i < instances.size(); ++i) {
+                    if (last_visit_of_instance[i] < 0)
+                        continue;
+                    InstanceVisit &last_visit = visits[size_t(last_visit_of_instance[i])];
+                    if (last_visit.islands.empty())
+                        // A visit without explicit islands already prints everything.
+                        continue;
+                    std::vector<ObjectByExtruder::Island> &islands = instances[i].object_by_extruder.islands;
+                    if (!islands.back().by_region.empty())
+                        last_visit.islands.emplace_back(islands.size() - 1);
+                }
+                cache_entry.second = std::move(visits);
             }
+            filament_plan.second = cache_entry.second;
         }
     }
 
     std::set<size_t> layer_object_label_ids;
     for (auto iter = filament_to_print_instances.begin(); iter != filament_to_print_instances.end(); ++iter) {
-        for (const InstanceToPrint &instance : iter->second) {
+        for (const InstanceToPrint &instance : iter->second.first) {
             layer_object_label_ids.insert(instance.label_object_id);
         }
     }
@@ -6095,7 +6161,7 @@ LayerResult GCode::process_layer(
 
         if (print.config().print_sequence == PrintSequence::ByLayer && m_enable_exclude_object && print.config().support_object_skip_flush.value) {
             std::vector<size_t> filament_instances_id;
-            for (InstanceToPrint &instance : filament_to_print_instances[extruder_id]) filament_instances_id.emplace_back(instance.label_object_id);
+            for (InstanceToPrint &instance : filament_to_print_instances[extruder_id].first) filament_instances_id.emplace_back(instance.label_object_id);
             m_filament_instances_code = _encode_label_ids_to_base64(filament_instances_id);
         }
 
@@ -6176,7 +6242,9 @@ LayerResult GCode::process_layer(
         if (layer_tools.has_wipe_tower && m_wipe_tower)
             m_last_processor_extrusion_role = erWipeTower;
 
-        std::vector<InstanceToPrint> &instances_to_print = filament_to_print_instances[extruder_id];
+        auto &filament_plan = filament_to_print_instances[extruder_id];
+        std::vector<InstanceToPrint>     &instances_to_print = filament_plan.first;
+        const std::vector<InstanceVisit> &instance_visits    = filament_plan.second;
 
         // We are almost ready to print. However, we must go through all the objects twice to print the overridden extrusions first (infill/perimeter wiping feature):
         std::vector<ObjectByExtruder::Island::Region> by_region_per_copy_cache;
@@ -6184,10 +6252,11 @@ LayerResult GCode::process_layer(
             if (is_anything_overridden && print_wipe_extrusions == 0)
                 gcode+="; PURGING FINISHED\n";
 
-            for (InstanceToPrint &instance_to_print : instances_to_print) {
+            for (const InstanceVisit &visit : instance_visits) {
+                InstanceToPrint &instance_to_print = instances_to_print[visit.instance_idx];
                 const auto& inst = instance_to_print.print_object.instances()[instance_to_print.instance_id];
                 const LayerToPrint &layer_to_print = layers[instance_to_print.layer_id];
-                if (print_wipe_extrusions == (is_anything_overridden ? 1 : 0)) {
+                if (visit.first_visit && print_wipe_extrusions == (is_anything_overridden ? 1 : 0)) {
                     gcode += generate_object_skirt_group(print, instance_to_print.print_object, instance_to_print.instance_id, layer_tools, layer, extruder_id);
                     gcode += generate_object_brim(print, instance_to_print.print_object, instance_to_print.instance_id, first_layer);
                 }
@@ -6240,7 +6309,7 @@ LayerResult GCode::process_layer(
                     m_avoid_crossing_perimeters.use_external_mp_once();
                 m_last_obj_copy = this_object_copy;
                 this->set_origin(unscale(offset));
-                if (instance_to_print.object_by_extruder.support != nullptr) {
+                if (visit.first_visit && instance_to_print.object_by_extruder.support != nullptr) {
                     m_layer = layers[instance_to_print.layer_id].support_layer;
                     m_object_layer_over_raft = false;
 
@@ -6274,9 +6343,42 @@ LayerResult GCode::process_layer(
                     m_layer = layer_to_print.layer();
                     m_object_layer_over_raft = object_layer_over_raft;
                 }
-                //FIXME order islands?
                 // Sequential tool path ordering of multiple parts within the same object, aka. perimeter tracking (#5511)
-                for (ObjectByExtruder::Island &island : instance_to_print.object_by_extruder.islands) {
+                // Island print order. Use the islands the tour assigned to this visit; if none,
+                // chain all islands nearest-neighbor from the current nozzle position (last_pos(),
+                // in this instance's frame after set_origin() above). Empty islands are skipped;
+                // the trailing catch-all island has no centroid to chain by and always goes last.
+                std::vector<ObjectByExtruder::Island> &islands = instance_to_print.object_by_extruder.islands;
+                std::vector<size_t> island_order = visit.islands;
+                if (island_order.empty()) {
+                    island_order.reserve(islands.size());
+                    if (layer_to_print.object_layer != nullptr && islands.size() == layer_to_print.object_layer->lslices.size() + 1) {
+                        for (size_t i = 0; i + 1 < islands.size(); ++i)
+                            if (!islands[i].by_region.empty())
+                                island_order.emplace_back(i);
+                        if (island_order.size() > 1) {
+                            Points island_centroids;
+                            island_centroids.reserve(island_order.size());
+                            for (size_t i : island_order)
+                                island_centroids.emplace_back(layer_to_print.object_layer->lslices[i].contour.centroid());
+                            const Point start_near = this->last_pos();
+                            std::vector<size_t> chain = chain_points(island_centroids, this->last_pos_defined() ? &start_near : nullptr);
+                            std::vector<size_t> ordered;
+                            ordered.reserve(island_order.size());
+                            for (size_t k : chain)
+                                ordered.emplace_back(island_order[k]);
+                            island_order = std::move(ordered);
+                        }
+                        if (!islands.back().by_region.empty())
+                            island_order.emplace_back(islands.size() - 1);
+                    } else {
+                        // Unexpected islands layout, keep the stored order.
+                        for (size_t i = 0; i < islands.size(); ++i)
+                            island_order.emplace_back(i);
+                    }
+                }
+                for (size_t island_idx : island_order) {
+                    ObjectByExtruder::Island &island = islands[island_idx];
                     const auto& by_region_specific = is_anything_overridden ? island.by_region_per_copy(by_region_per_copy_cache, static_cast<unsigned int>(instance_to_print.instance_id), extruder_id, print_wipe_extrusions != 0) : island.by_region;
                     // When starting a new object, use the external motion planner for the first travel move.
                     const Point& offset = instance_to_print.print_object.instances()[instance_to_print.instance_id].shift;

src/libslic3r/GCode.hpp

@@ -539,9 +539,38 @@ private:
     // Cache for custom seam enforcers/blockers for each layer.
     SeamPlacer                          m_seam_placer;
 
-    // Cache per-filament ordering to avoid recomputing when object set is unchanged across layers.
-    // Key: filament_id. Value: {sorted ObjectIDs of objects present on this filament, cached ordering}.
-    std::map<unsigned int, std::pair<std::vector<ObjectID>, std::vector<const PrintInstance*>>>
+    // One stop of the island-level tour: consecutive islands of a single instance. An instance
+    // can have several visits per layer when its islands are toured non-consecutively.
+    struct InstanceVisit
+    {
+        // Index into the per-filament InstanceToPrint vector.
+        size_t              instance_idx;
+        // Islands to print, in order (indices into ObjectByExtruder::islands). Empty: print all
+        // islands, ordered at extrusion time.
+        std::vector<size_t> islands;
+        // First visit of this instance this layer; skirt, brim and support are emitted here.
+        bool                first_visit;
+    };
+
+    // One node of the island-level tour, also used as cache key: identity plus quantized position.
+    struct IslandOrderNode
+    {
+        ObjectID object_id;
+        size_t   instance_id;
+        // Index into ObjectByExtruder::islands, or size_t(-1) for an instance without chainable
+        // islands (e.g. support only), which is toured as a single stop.
+        size_t   island_idx;
+        // Island centroid in G-code coordinates, quantized to 1 mm for cache stability.
+        Point    pos;
+        bool operator==(const IslandOrderNode &rhs) const {
+            return object_id == rhs.object_id && instance_id == rhs.instance_id &&
+                   island_idx == rhs.island_idx && pos == rhs.pos;
+        }
+    };
+
+    // Cache the per-filament island tour to avoid recomputing while the layer's island layout is
+    // unchanged. Key: filament_id. Value: {nodes the tour was computed from, resulting visits}.
+    std::map<unsigned int, std::pair<std::vector<IslandOrderNode>, std::vector<InstanceVisit>>>
                                         m_ordering_cache;
 
     ExtrusionQualityEstimator m_extrusion_quality_estimator;

src/libslic3r/GCode/OrderingStrategies.cpp

@@ -109,17 +109,22 @@ bool tsp_remove_crossings(std::vector<size_t>& path, const Points& centers)
     size_t pn = path.size();
     if (pn <= 3) return false;
 
-    size_t n_edges = pn - 1;
+    // Treat path as a cycle: include the closing edge (pn-1 -> 0), consistent with the other
+    // TSP helpers (2-opt, closing-edge rotation) that operate on the full cycle.
+    size_t n_edges = pn;
 
     // Scan for first crossing; returns {i, j} or {npos, npos} if none.
     auto find_crossing = [&]() -> std::pair<size_t, size_t> {
         for (size_t i = 0; i < n_edges; ++i) {
             const Point& ai = centers[path[i]];
-            const Point& bi = centers[path[i + 1]];
+            const Point& bi = centers[path[(i + 1) % pn]];
 
             for (size_t j = i + 2; j < n_edges; ++j) {
+                // Skip the (0, pn-1) pair: edges (0,1) and (pn-1,0) share node 0.
+                if (i == 0 && j == pn - 1) continue;
+
                 const Point& aj = centers[path[j]];
-                const Point& bj = centers[path[j + 1]];
+                const Point& bj = centers[path[(j + 1) % pn]];
 
                 if (!bboxes_overlap(ai, bi, aj, bj)) continue;
                 if (Geometry::segments_intersect(ai, bi, aj, bj))
@@ -374,4 +379,57 @@ std::vector<const PrintInstance*> chain_print_object_instances_best_of(const Pri
     return chain_print_object_instances_best_of(print.objects().vector(), nullptr);
 }
 
+/* ====================================================================
+ * Island-level ordering entry point
+ * ==================================================================== */
+
+std::vector<size_t> order_points_with_strategy(const Points& points, PrintOrder print_order, const Point* start_near)
+{
+    if (points.empty())
+        return {};
+
+    if (print_order != PrintOrder::Snake && print_order != PrintOrder::BestOfStrategies)
+        // Nearest neighbor + post-processing; honours start_near natively.
+        return chain_points_with_postprocessing(points, start_near);
+
+    auto run_snake = [&points, start_near]() {
+        std::vector<size_t> path = snake_core(points);
+        if (start_near != nullptr && !path.empty()) {
+            // Start the cycle at the point closest to start_near.
+            size_t best_start = 0;
+            double best_d2 = std::numeric_limits<double>::max();
+            for (size_t k = 0; k < points.size(); ++k) {
+                double d2 = (points[k].cast<double>() - start_near->cast<double>()).squaredNorm();
+                if (d2 < best_d2) { best_d2 = d2; best_start = k; }
+            }
+            auto it = std::find(path.begin(), path.end(), best_start);
+            if (it != path.begin() && it != path.end())
+                std::rotate(path.begin(), it, path.end());
+        } else {
+            tsp_rotate_minimize_closing(path, points);
+        }
+        return path;
+    };
+
+    if (print_order == PrintOrder::Snake)
+        return run_snake();
+
+    // Best-of: pick the shortest total cycle; tiebreak on smallest max edge.
+    std::vector<std::vector<size_t>> candidates;
+    candidates.emplace_back(chain_points_with_postprocessing(points, start_near));
+    candidates.emplace_back(run_snake());
+
+    size_t best      = 0;
+    double best_len  = std::numeric_limits<double>::max();
+    double best_edge = std::numeric_limits<double>::max();
+    for (size_t i = 0; i < candidates.size(); ++i) {
+        double len  = tsp_cycle_path_length(candidates[i], points);
+        double edge = tsp_max_edge_length(candidates[i], points);
+        if (len < best_len || (len == best_len && edge < best_edge)) {
+            best_len = len; best_edge = edge; best = i;
+        }
+    }
+    return candidates[best];
+}
+
 } // namespace Slic3r

src/libslic3r/GCode/OrderingStrategies.hpp

@@ -136,6 +136,11 @@ std::vector<const PrintInstance*> chain_print_object_instances_snake(const Print
 std::vector<const PrintInstance*> chain_print_object_instances_best_of(const std::vector<const PrintObject*>& print_objects, const Point* start_near);
 std::vector<const PrintInstance*> chain_print_object_instances_best_of(const Print& print);
 
+// Order raw points with the selected strategy, returning an index permutation. Island-level
+// counterpart of the chain_print_object_instances_* helpers. The returned cycle starts at the
+// point closest to start_near; orders without a dedicated strategy use nearest-neighbor chaining.
+std::vector<size_t> order_points_with_strategy(const Points& points, PrintOrder print_order, const Point* start_near);
+
 #endif // SLIC3R_TEST_HARNESS
 
 } // namespace Slic3r

src/libslic3r/PrintConfig.cpp

@@ -2001,7 +2001,21 @@ void PrintConfigDef::init_fff_params()
 
     def = this->add("print_order", coEnum);
     def->label = L("Intra-layer order");
-    def->tooltip = L("Print order within a single layer.");
+    def->tooltip = L("Order in which object instances are visited within a single layer, which controls how much "
+                     "travel is spent moving between them.\n\n"
+                     "Default: nearest-neighbor chaining, refined with 2-opt and crossing removal. A good general "
+                     "choice.\n"
+                     "As object list: instances are printed in the same order as the object list, without any path "
+                     "optimization. Use it when you need a predictable, manually controlled order.\n"
+                     "Best of all (shortest path): every strategy is evaluated and the shortest one is used. The "
+                     "object instance order is decided once for the whole print, while the ordering of individual "
+                     "islands is decided per layer, so different layers may end up using different strategies. "
+                     "Slightly slower to slice.\n"
+                     "Snake: serpentine row-by-row traversal, refined with 2-opt. Well suited to regular grids of "
+                     "many small parts.\n\n"
+                     "With multiple filaments or tools in the same layer, minimizing tool changes takes priority: "
+                     "objects are grouped by filament first and this setting only orders the instances within each "
+                     "filament group, so the overall sequence may not look like the shortest path across the plate.");
     def->enum_keys_map = &ConfigOptionEnum<PrintOrder>::get_enum_values();
     def->enum_values.push_back("default");
     def->enum_values.push_back("as_obj_list");