Answer the Preview's Per-Frame Time Query From a Cached Sum

The G-code preview's cost is linear in the number of toolpath vertices, and on a
tall multi-filament print the wipe tower dominates that count: it emits a roughly
constant 160-180 moves on every layer whatever the object is, measured at 57-61%
of all moves on a three-filament print.

Four places scanned or allocated across the whole vertex array. None of them
needed to.

get_estimated_time_at re-accumulated the estimated time from vertex 0 on every
call, and its caller is the tool marker tooltip, which ImGui re-renders every
frame while the properties panel is unfolded. It now starts from a running sum
kept at each layer's first vertex, built at load in vertex order, and adds only
that layer's vertices: the same additions in the same order, so the float result
is unchanged, at a cost of one float per layer and time mode rather than per
vertex. At the 351k vertices of a 636-layer test print the call scanned the whole
print (238us); it now scans one layer.

update_view_full_range walked from vertex 0 to find where the layer range starts,
on every slider tick. It now starts at the first vertex of that layer. The index
is derived from the vertices rather than from Layers::Item::range, because
Layers::update folds a vertex whose layer_id arrives out of order into whichever
bucket is open, which makes that range the wrong answer in general; the index
costs four bytes per layer, not per vertex.

update_colors_texture allocated one float per vertex of the whole print on every
slider tick. It now reuses a buffer.

render_legend fetched the layer Zs and the per-layer times from inside loops over
the custom G-code items, and built whole vectors only to test them for emptiness.
The times are hoisted, the Zs are built lazily so a print with no colour change
does not pay for them at all, and the emptiness tests use the existing counters.

No rendering behaviour changes.

src/libvgcode/src/ViewerImpl.cpp

@@ -875,6 +875,12 @@ void ViewerImpl::reset()
     m_travels_time = { 0.0f, 0.0f };
     m_vertices.clear();
     m_vertices_colors.clear();
+    // swap rather than clear: these are sized by the print, and a reset means the memory
+    // should go back, not sit reserved until the next load
+    for (std::vector<float>& times : m_layer_start_times)
+        std::vector<float>().swap(times);
+    std::vector<uint32_t>().swap(m_layer_first_vertex);
+    std::vector<float>().swap(m_colors_scratch);
     m_valid_lines_bitset.clear();
 #if VGCODE_ENABLE_COG_AND_TOOL_MARKERS
     m_cog_marker.reset();
@@ -1048,6 +1054,37 @@ void ViewerImpl::load(GCodeInputData&& gcode_data)
         v.layer_duration = m_layers.get_layer_time(m_settings.time_mode, static_cast<size_t>(v.layer_id));
     }
 
+    // Index of the first vertex of each layer, walked back to front so that a layer with no
+    // vertex of its own inherits the next layer's index and the array stays non-decreasing.
+    if (!m_layers.empty()) {
+        const uint32_t vertices_count = static_cast<uint32_t>(m_vertices.size());
+        m_layer_first_vertex.assign(m_layers.count(), vertices_count);
+        for (uint32_t i = vertices_count; i > 0; --i) {
+            const uint32_t layer_id = m_vertices[i - 1].layer_id;
+            if (layer_id < m_layer_first_vertex.size())
+                m_layer_first_vertex[layer_id] = i - 1;
+        }
+        for (size_t i = m_layer_first_vertex.size() - 1; i > 0; --i)
+            m_layer_first_vertex[i - 1] = std::min(m_layer_first_vertex[i - 1], m_layer_first_vertex[i]);
+
+        // the running time at each layer's first vertex, summed in vertex order so that
+        // get_estimated_time_at() matches a full accumulation exactly
+        std::array<float, TIME_MODES_COUNT> running{};
+        for (std::vector<float>& times : m_layer_start_times)
+            times.assign(m_layer_first_vertex.size(), 0.0f);
+        size_t layer = 0;
+        for (size_t i = 0; i <= m_vertices.size(); ++i) {
+            for (; layer < m_layer_first_vertex.size() && m_layer_first_vertex[layer] == i; ++layer) {
+                for (size_t j = 0; j < TIME_MODES_COUNT; ++j)
+                    m_layer_start_times[j][layer] = running[j];
+            }
+            if (i < m_vertices.size()) {
+                for (size_t j = 0; j < TIME_MODES_COUNT; ++j)
+                    running[j] += m_vertices[i].times[j];
+            }
+        }
+    }
+
     if (!m_layers.empty())
         m_layers.set_view_range(0, static_cast<uint32_t>(m_layers.count()) - 1);
 
@@ -1261,7 +1298,10 @@ void ViewerImpl::update_colors_texture()
 
     // Based on current settings and slider position, we might want to render some
     // vertices as dark grey (or darkened, see above). Use either that or the normal color (from the cache).
-    std::vector<float> colors(m_vertices_colors.size());
+    // Reused across calls: this runs on every slider tick, and the allocation alone is
+    // 4 bytes per vertex of the whole print each time.
+    std::vector<float>& colors = m_colors_scratch;
+    colors.resize(m_vertices_colors.size());
     assert(colors.size() == m_vertices.size() && m_vertices_colors.size() == m_vertices.size());
     for (size_t i=0; i<m_vertices.size(); ++i) {
         const PathVertex& v = m_vertices[i];
@@ -1516,8 +1556,19 @@ void ViewerImpl::set_view_visible_range(Interval::value_type min, Interval::valu
 
 float ViewerImpl::get_estimated_time_at(size_t id) const
 {
-    return std::accumulate(m_vertices.begin(), m_vertices.begin() + id + 1, 0.0f, 
-        [this](float a, const PathVertex& v) { return a + v.times[static_cast<size_t>(m_settings.time_mode)]; });
+    const size_t mode = static_cast<size_t>(m_settings.time_mode);
+    if (mode >= TIME_MODES_COUNT || id >= m_vertices.size())
+        return 0.0f;
+    size_t first = 0;
+    float time = 0.0f;
+    const size_t layer = static_cast<size_t>(m_vertices[id].layer_id);
+    if (layer < m_layer_first_vertex.size() && m_layer_first_vertex[layer] <= id) {
+        first = m_layer_first_vertex[layer];
+        time = m_layer_start_times[mode][layer];
+    }
+    for (size_t i = first; i <= id; ++i)
+        time += m_vertices[i].times[mode];
+    return time;
 }
 
 Color ViewerImpl::get_vertex_color(const PathVertex& v) const
@@ -1722,6 +1773,10 @@ size_t ViewerImpl::get_used_cpu_memory() const
     ret += sizeof(m_extrusion_roles_colors);
     ret += sizeof(m_options_colors);
     ret += STDVEC_MEMSIZE(m_vertices, PathVertex);
+    for (const std::vector<float>& times : m_layer_start_times)
+        ret += STDVEC_MEMSIZE(times, float);
+    ret += STDVEC_MEMSIZE(m_layer_first_vertex, uint32_t);
+    ret += STDVEC_MEMSIZE(m_colors_scratch, float);
     ret += m_valid_lines_bitset.size_in_bytes_cpu();
     ret += m_height_range.size_in_bytes_cpu();
     ret += m_width_range.size_in_bytes_cpu();
@@ -1787,7 +1842,11 @@ void ViewerImpl::update_view_full_range()
     const bool travels_visible = m_settings.options_visibility[size_t(EOptionType::Travels)];
     const bool wipes_visible   = m_settings.options_visibility[size_t(EOptionType::Wipes)];
 
+    // every vertex before m_layer_first_vertex[layers_range[0]] has a smaller layer_id, so the loop
+    // below would skip all of them anyway
     auto first_it = m_vertices.begin();
+    if (layers_range[0] < m_layer_first_vertex.size())
+        first_it += m_layer_first_vertex[layers_range[0]];
     while (first_it != m_vertices.end() &&
            (first_it->layer_id < layers_range[0] || !is_visible(*first_it, m_settings))) {
         ++first_it;

src/libvgcode/src/ViewerImpl.hpp

@@ -234,6 +234,20 @@ private:
     //
     std::array<float, TIME_MODES_COUNT> m_total_time{ 0.0f, 0.0f };
     //
+    // Running sum of the vertex estimated times at each layer's first vertex, for each time mode,
+    // so that get_estimated_time_at() only accumulates the vertices of one layer.
+    //
+    std::array<std::vector<float>, TIME_MODES_COUNT> m_layer_start_times;
+    //
+    // For each layer L, the index of the first vertex whose layer_id is >= L (m_vertices.size()
+    // if there is none). Derived from the vertices, so it stays exact whatever order they arrive in.
+    //
+    std::vector<uint32_t> m_layer_first_vertex;
+    //
+    // Scratch buffer for update_colors_texture(), kept alive across slider steps
+    //
+    std::vector<float> m_colors_scratch;
+    //
     // Detected travel moves times
     //
     std::array<float, TIME_MODES_COUNT> m_travels_time{ 0.0f, 0.0f };

src/slic3r/GUI/GCodeViewer.cpp

@@ -1627,7 +1627,7 @@ void GCodeViewer::render_scene(int canvas_width, int canvas_height)
     glsafe(::glEnable(GL_DEPTH_TEST));
     render_shells(canvas_width, canvas_height);
 
-    if (m_viewer.get_extrusion_roles().empty())
+    if (m_viewer.get_extrusion_roles_count() == 0)
         return;
 
     render_toolpaths();
@@ -3426,6 +3426,12 @@ void GCodeViewer::render_legend(float &legend_height, int canvas_width, int canv
         std::vector<std::pair<ColorRGBA, std::pair<double, double>>> ret;
         ret.reserve(custom_gcode_per_print_z.size());
 
+        // Loop invariant, but built lazily: this lambda runs once per extruder on every frame
+        // and most prints reach neither colour change below, so fetching it up front would cost
+        // more than the per-item fetch it replaces.
+        std::vector<float> zs;
+        bool zs_built = false;
+
         for (const auto& item : custom_gcode_per_print_z) {
             if (extruder_id + 1 != static_cast<unsigned char>(item.extruder))
                 continue;
@@ -3433,7 +3439,10 @@ void GCodeViewer::render_legend(float &legend_height, int canvas_width, int canv
             if (item.type != ColorChange)
                 continue;
 
-            const std::vector<float> zs = m_viewer.get_layers_zs();
+            if (!zs_built) {
+                zs = m_viewer.get_layers_zs();
+                zs_built = true;
+            }
             auto lower_b = std::lower_bound(zs.begin(), zs.end(),
                 static_cast<float>(item.print_z - epsilon()));
             if (lower_b == zs.end())
@@ -4582,6 +4591,8 @@ void GCodeViewer::render_legend(float &legend_height, int canvas_width, int canv
 
         // ORCA: Get layer Zs as doubles
         std::vector<double> layer_zs = get_layers_zs();
+        // loop invariant, same reason as the layer Zs above
+        const std::vector<float> layer_times = m_viewer.get_layers_estimated_times();
 
         for (Slic3r::CustomGCode::Item custom_gcode : custom_gcode_per_print_z) {
             ImGui::Dummy({window_padding, window_padding});
@@ -4601,7 +4612,6 @@ void GCodeViewer::render_legend(float &legend_height, int canvas_width, int canv
             imgui.text(buf);
             ImGui::SameLine(max_len * 1.5);
 
-            std::vector<float> layer_times = m_viewer.get_layers_estimated_times();
             float custom_gcode_time = 0;
             if (layer > 0)
             {
@@ -4650,7 +4660,7 @@ void GCodeViewer::render_legend(float &legend_height, int canvas_width, int canv
     std::string print_str = _u8L("Model printing time");
     std::string total_str = _u8L("Total time");
     float max_len = window_padding + 2 * ImGui::GetStyle().ItemSpacing.x;
-    if (m_viewer.get_layers_estimated_times().empty())
+    if (m_viewer.get_layers_count() == 0)
         max_len += ImGui::CalcTextSize(total_str.c_str()).x;
     else {
         if (m_viewer.get_view_type() == libvgcode::EViewType::FeatureType)