update tracy to v0.13.1

- uses a huge global seen-states-cache. Not scalable without more
filtering

.gitignore

@@ -9,3 +9,5 @@ cmake-build-release
 /perf.data
 /perf.data.old
 /clusters.puzzle
+/benchs.json
+/benchs.old.json

CMakeLists.txt

@@ -18,24 +18,27 @@ option(DISABLE_BENCH "Disable building benchmarks" OFF)
 # Headers + Sources (excluding main.cpp)
 set(SOURCES
     src/backward.cpp
+    src/bits.cpp
+    src/cpu_layout_engine.cpp
+    src/cpu_spring_system.cpp
     src/graph_distances.cpp
     src/input_handler.cpp
     src/load_save.cpp
-    src/mass_spring_system.cpp
     src/octree.cpp
     src/orbit_camera.cpp
     src/puzzle.cpp
     src/renderer.cpp
     src/state_manager.cpp
-    src/threaded_physics.cpp
     src/user_interface.cpp
 )
 
 # Libraries
 include(FetchContent)
 find_package(raylib REQUIRED)
+find_package(GLEW REQUIRED)
+find_package(libmorton REQUIRED)
 find_package(Boost COMPONENTS program_options REQUIRED)
-set(LIBS raylib Boost::headers Boost::program_options)
+set(LIBS raylib GLEW::GLEW Boost::headers Boost::program_options)
 set(FLAGS "")
 
 if(WIN32)
@@ -51,12 +54,14 @@ if(NOT DISABLE_BACKWARD)
 
     list(APPEND LIBS Backward::Backward)
     list(APPEND FLAGS BACKWARD)
+
+    message("-- BACKWARD: Enabled")
 endif()
 
 if(NOT DISABLE_TRACY)
     FetchContent_Declare(tracy
                          GIT_REPOSITORY https://github.com/wolfpld/tracy.git
-                         GIT_TAG v0.11.1
+                         GIT_TAG v0.13.1
                          GIT_SHALLOW TRUE
                          GIT_PROGRESS TRUE
     )
@@ -66,6 +71,8 @@ if(NOT DISABLE_TRACY)
 
     list(APPEND LIBS TracyClient)
     list(APPEND FLAGS TRACY)
+
+    message("-- TRACY: Enabled")
 endif()
 
 # Set this after fetching tracy to hide tracy's warnings.
@@ -106,10 +113,12 @@ if(NOT DISABLE_TESTS AND NOT WIN32)
 
     add_executable(tests ${TEST_SOURCES} ${SOURCES})
     target_include_directories(tests PRIVATE include)
-    target_link_libraries(tests Catch2::Catch2WithMain raylib)
+    target_link_libraries(tests Catch2::Catch2WithMain raylib GLEW::GLEW)
 
     include(Catch)
     catch_discover_tests(tests)
+
+    message("-- TESTS: Enabled")
 endif()
 
 # Benchmarking
@@ -122,15 +131,17 @@ if(NOT DISABLE_BENCH AND NOT WIN32)
 
     add_executable(benchmarks ${BENCH_SOURCES} ${SOURCES})
     target_include_directories(benchmarks PRIVATE include)
-    target_link_libraries(benchmarks benchmark raylib)
+    target_link_libraries(benchmarks benchmark raylib GLEW::GLEW)
+
+    message("-- BENCHMARKS: Enabled")
 endif()
 
 # LTO
 include(CheckIPOSupported)
 check_ipo_supported(RESULT supported OUTPUT error)
 if(supported)
-    message(STATUS "IPO / LTO enabled")
+    message("-- IPO/LTO: Enabled")
     set_property(TARGET masssprings PROPERTY INTERPROCEDURAL_OPTIMIZATION TRUE)
 else()
-    message(STATUS "IPO / LTO not supported")
+    message("-- IPO/LTO: Disabled")
 endif()

benchmark/state_space.cpp

@@ -1,6 +1,11 @@
+// ReSharper disable CppTooWideScope
+// ReSharper disable CppDFAUnreadVariable
 #include "puzzle.hpp"
 
+#include <random>
+#include <unordered_set>
 #include <benchmark/benchmark.h>
+#include <boost/unordered/unordered_flat_map.hpp>
 
 static std::vector<std::string> puzzles = {
     // 0: RushHour 1
@@ -35,6 +40,113 @@ static std::vector<std::string> puzzles = {
     "S:[4x5] G:[1,3] M:[F] B:[{_ 2X2 _ _} {1x1 _ _ 1x1} {1x2 2x1 _ 1x2} {_ 2x1 _ _} {1x1 2x1 _ 1x1}]",
 };
 
+template <u8 N>
+struct uint_hasher
+{
+    int64_t nums;
+
+    auto operator()(const std::array<u64, N>& ints) const noexcept -> size_t
+    {
+        size_t h = 0;
+        for (size_t i = 0; i < N; ++i) {
+            puzzle::hash_combine(h, ints[i]);
+        }
+        return h;
+    }
+};
+
+template <u8 N>
+static auto unordered_set_uint64(benchmark::State& state) -> void
+{
+    std::random_device random_device;
+    std::mt19937 generator(random_device());
+    std::uniform_int_distribution<u64> distribution(
+        std::numeric_limits<u64>::min(),
+        std::numeric_limits<u64>::max()
+    );
+
+    std::unordered_set<std::array<u64, N>, uint_hasher<N>> set;
+    std::array<u64, N> ints;
+    for (size_t i = 0; i < N; ++i) {
+        ints[i] = distribution(generator);
+    }
+
+    for (auto _ : state) {
+        for (size_t i = 0; i < 100000; ++i) {
+            set.emplace(ints);
+        }
+
+        benchmark::DoNotOptimize(set);
+    }
+}
+
+template <u8 N>
+static auto unordered_flat_set_uint64(benchmark::State& state) -> void
+{
+    std::random_device random_device;
+    std::mt19937 generator(random_device());
+    std::uniform_int_distribution<u64> distribution(
+        std::numeric_limits<u64>::min(),
+        std::numeric_limits<u64>::max()
+    );
+
+    boost::unordered_flat_set<std::array<u64, N>, uint_hasher<N>> set;
+    std::array<u64, N> ints;
+    for (size_t i = 0; i < N; ++i) {
+        ints[i] = distribution(generator);
+    }
+
+    for (auto _ : state) {
+        for (size_t i = 0; i < 100000; ++i) {
+            set.emplace(ints);
+        }
+
+        benchmark::DoNotOptimize(set);
+    }
+}
+
+static auto unordered_flat_set_block_hasher(benchmark::State& state) -> void
+{
+    blockset set;
+    const block b = block(2, 3, 1, 2, true, false);
+
+    for (auto _ : state) {
+        for (size_t i = 0; i < 100000; ++i) {
+            set.emplace(b);
+        }
+
+        benchmark::DoNotOptimize(set);
+    }
+}
+
+static auto unordered_flat_set_block_hasher2(benchmark::State& state) -> void
+{
+    blockset2 set;
+    const block b = block(2, 3, 1, 2, true, false);
+
+    for (auto _ : state) {
+        for (size_t i = 0; i < 100000; ++i) {
+            set.emplace(b);
+        }
+
+        benchmark::DoNotOptimize(set);
+    }
+}
+
+static auto unordered_flat_set_puzzle_hasher(benchmark::State& state) -> void
+{
+    puzzleset set;
+    const puzzle p = puzzle(puzzles[0]);
+
+    for (auto _ : state) {
+        for (size_t i = 0; i < 100000; ++i) {
+            set.emplace(p);
+        }
+
+        benchmark::DoNotOptimize(set);
+    }
+}
+
 static auto explore_state_space(benchmark::State& state) -> void
 {
     const puzzle p = puzzle(puzzles[state.range(0)]);
@@ -48,43 +160,52 @@ static auto explore_state_space(benchmark::State& state) -> void
 
 static auto explore_rush_hour_puzzle_space(benchmark::State& state) -> void
 {
-    // ReSharper disable once CppTooWideScope
-    constexpr uint8_t max_blocks = 5;
+    constexpr u8 max_blocks = 5;
 
-    constexpr uint8_t board_width = 4;
-    constexpr uint8_t board_height = 5;
-    constexpr uint8_t goal_x = board_width - 1;
-    constexpr uint8_t goal_y = 2;
+    constexpr u8 board_width = 4;
+    constexpr u8 board_height = 5;
+    constexpr u8 goal_x = board_width - 1;
+    constexpr u8 goal_y = 2;
     constexpr bool restricted = true;
 
-    const boost::unordered_flat_set<puzzle::block, block_hasher2, block_equal2> permitted_blocks = {
-        puzzle::block(0, 0, 2, 1, false, false),
-        puzzle::block(0, 0, 3, 1, false, false),
-        puzzle::block(0, 0, 1, 2, false, false),
-        puzzle::block(0, 0, 1, 3, false, false)
+    const blockset2 permitted_blocks = {
+        block(0, 0, 2, 1, false, false),
+        block(0, 0, 3, 1, false, false),
+        block(0, 0, 1, 2, false, false),
+        block(0, 0, 1, 3, false, false)
     };
-    const puzzle::block target_block = puzzle::block(0, 0, 2, 1, true, false);
-    constexpr std::tuple<uint8_t, uint8_t, uint8_t, uint8_t> target_block_pos_range = {
+    const block target_block = block(0, 0, 2, 1, true, false);
+    constexpr std::tuple<u8, u8, u8, u8> target_block_pos_range = {
         0,
         goal_y,
-        board_width - 1,
+        goal_x,
         goal_y
     };
 
     const puzzle p = puzzle(board_width, board_height, goal_x, goal_y, restricted, true);
 
     for (auto _ : state) {
-        boost::unordered_flat_set<puzzle, puzzle_hasher> result = p.explore_puzzle_space(
+        puzzleset result = p.explore_puzzle_space(
             permitted_blocks,
             target_block,
             target_block_pos_range,
             max_blocks,
+            0,
             std::nullopt);
 
         benchmark::DoNotOptimize(result);
     }
 }
 
+BENCHMARK(unordered_set_uint64<4>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_set_uint64<8>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_set_uint64<16>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_uint64<4>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_uint64<8>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_uint64<16>)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_block_hasher)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_block_hasher2)->Unit(benchmark::kMicrosecond);
+BENCHMARK(unordered_flat_set_puzzle_hasher)->Unit(benchmark::kMicrosecond);
 BENCHMARK(explore_state_space)->DenseRange(0, puzzles.size() - 1)->Unit(benchmark::kMicrosecond);
 BENCHMARK(explore_rush_hour_puzzle_space)->Unit(benchmark::kSecond);
 

flake.lock

@@ -20,11 +20,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1772479524,
-        "narHash": "sha256-u7nCaNiMjqvKpE+uZz9hE7pgXXTmm5yvdtFaqzSzUQI=",
+        "lastModified": 1773201692,
+        "narHash": "sha256-NXrKzNMniu4Oam2kAFvqJ3GB2kAvlAFIriTAheaY8hw=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "4215e62dc2cd3bc705b0a423b9719ff6be378a43",
+        "rev": "b6067cc0127d4db9c26c79e4de0513e58d0c40c9",
         "type": "github"
       },
       "original": {

flake.nix

@@ -111,9 +111,9 @@ rec {
             abbr -a release "${buildRelease} && ./cmake-build-release/masssprings"
 
             abbr -a run "${buildRelease} && ./cmake-build-release/masssprings"
-            abbr -a run-clusters "${buildRelease} && ./cmake-build-release/masssprings --output=clusters.puzzle --space=rh --w=6 --h=6 --gx=4 --gy=2 --blocks=4"
+            abbr -a runclusters "${buildRelease} && ./cmake-build-release/masssprings --output=clusters.puzzle --space=rh --moves=10 --blocks=4"
             abbr -a runtests "${buildDebug} && ./cmake-build-debug/tests"
-            abbr -a runbenchs "${buildRelease} && sudo cpupower frequency-set --governor performance && ./cmake-build-release/benchmarks; sudo cpupower frequency-set --governor powersave"
+            abbr -a runbenchs "mv -f benchs.json benchs.old.json; ${buildRelease} && sudo cpupower frequency-set --governor performance && ./cmake-build-release/benchmarks --benchmark_out=benchs.json --benchmark_out_format=console; sudo cpupower frequency-set --governor powersave"
             abbr -a rungdb "${buildDebug} && gdb --tui ./cmake-build-debug/masssprings"
             abbr -a runvalgrind "${buildDebug} && valgrind --leak-check=full --show-reachable=no --show-leak-kinds=definite,indirect,possible --track-origins=no --suppressions=valgrind.supp --log-file=valgrind.log ./cmake-build-debug/masssprings && cat valgrind.log"
             abbr -a runperf "${buildRelease} && perf record -g ./cmake-build-release/masssprings && hotspot ./perf.data"
@@ -135,7 +135,7 @@ rec {
         # Define custom dependencies
         # ===========================================================================================
 
-        raygui = stdenv.mkDerivation (finalAttrs: {
+        raygui = stdenv.mkDerivation rec {
           pname = "raygui";
           version = "4.0-unstable-2026-02-24";
 
@@ -163,13 +163,13 @@ rec {
             Name: raygui
             Description: Simple and easy-to-use immediate-mode gui library
             URL: https://github.com/raysan5/raygui
-            Version: ${finalAttrs.version}
+            Version: ${version}
             Cflags: -I"{includedir}"
             EOF
 
             runHook postInstall
           '';
-        });
+        };
 
         thread-pool = stdenv.mkDerivation {
           pname = "thread-pool";
@@ -185,8 +185,69 @@ rec {
           # Header-only library
           dontBuild = true;
           installPhase = ''
+            runHook preInstall
+
             mkdir -p $out
-            mv ./include $out/include
+            cp -rv ./include $out/include
+
+            runHook postInstall
+          '';
+        };
+
+        # We can use the pkgs.stdenv for Linux+Windows because it's a header only library.
+        # The build is required to create the pkg-config/cmake configuration files.
+        libmorton = stdenv.mkDerivation {
+          pname = "libmorton";
+          version = "0.2.12-unstable-2023-05-24";
+
+          src = pkgs.fetchFromGitHub {
+            owner = "Forceflow";
+            repo = "libmorton";
+            rev = "7923faa88d7e564020b2d5d408bf8c186ecbe363";
+            hash = "sha256-5LHiWu2GIuDmfM2gXGbRsFasE7AmVCSRphNdFElbbjk=";
+          };
+
+          nativeBuildInputs = with pkgs; [cmake];
+
+          cmakeFlags = [
+            "-DBUILD_TESTING=OFF"
+            "-DCMAKE_INSTALL_INCLUDEDIR=include"
+            "-DCMAKE_INSTALL_DATADIR=share"
+          ];
+        };
+
+        glew-windows = windowsPkgs.stdenv.mkDerivation rec {
+          pname = "glew-windows";
+          version = "2.2.0";
+
+          src = pkgs.fetchurl {
+            url = "https://github.com/nigels-com/glew/releases/download/glew-${version}/glew-${version}.tgz";
+            hash = "sha256-1PyCiTz7ABCVeNChojN/uMozWzzsz5e5flzH8I5DU+E=";
+          };
+
+          nativeBuildInputs = with pkgs; [
+            cmake
+            ninja
+            pkg-config
+          ];
+
+          preConfigure = ''
+            cd build/cmake
+          '';
+
+          cmakeFlags = [
+            "-DBUILD_UTILS=OFF"
+            "-DGLEW_OSMESA=OFF"
+            "-DBUILD_SHARED_LIBS=ON"
+            "-DCMAKE_POLICY_VERSION_MINIMUM=3.5"
+          ];
+
+          installPhase = ''
+            runHook preInstall
+
+            cmake --install . --prefix "$out"
+
+            runHook postInstall
           '';
         };
 
@@ -224,11 +285,13 @@ rec {
           # C/C++:
           raylib
           raygui
+          glew
           thread-pool
+          libmorton
           boost
 
           # Debugging/Testing/Profiling
-          tracy-wayland
+          tracy_0_13
           backward-cpp
           libbfd
           catch2_3
@@ -268,15 +331,29 @@ rec {
           '';
 
           installPhase = ''
+            runHook preInstall
+
             mkdir -p $out/lib
             cp ./${pname} $out/lib/
-            cp $src/default.puzzle $out/lib/
-            cp -r $src/fonts $out/lib/fonts
-            cp -r $src/shader $out/lib/shader
+            cp -rv $src/default.puzzle $out/lib/
+            cp -rv $src/fonts $out/lib/fonts
+            cp -rv $src/shader $out/lib/shader
 
             # The wrapper enters the correct working dir, so fonts/shaders/presets are available
             mkdir -p $out/bin
-            makeWrapper $out/lib/${pname} $out/bin/${pname} --chdir "${placeholder "out"}/lib"
+            makeWrapper $out/lib/${pname} $out/bin/${pname} --chdir "$out/lib"
+
+            # Generate a .desktop file
+            mkdir -p $out/share/applications
+            cat <<INI > $out/share/applications/${pname}.desktop
+            [Desktop Entry]
+            Terminal=true
+            Name=PuzzleSpaces
+            Exec=$out/bin/${pname} %f
+            Type=Application
+            INI
+
+            runHook postInstall
           '';
         };
 
@@ -294,7 +371,9 @@ rec {
           buildInputs = with windowsPkgs; [
             raylib
             raygui
+            glew-windows
             thread-pool
+            libmorton
 
             # Disable stacktrace since that's platform dependant and won't cross compile to windows
             # https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/libraries/boost/generic.nix#L43
@@ -316,11 +395,15 @@ rec {
           ];
 
           installPhase = ''
+            runHook preInstall
+
             mkdir -p $out/bin
-            cp ./${pname}.exe $out/bin/
-            cp $src/default.puzzle $out/bin/
-            cp -r $src/fonts $out/bin/fonts
-            cp -r $src/shader $out/bin/shader
+            cp -rv ./${pname}.exe $out/bin/
+            cp -rv $src/default.puzzle $out/bin/
+            cp -rv $src/fonts $out/bin/fonts
+            cp -rv $src/shader $out/bin/shader
+
+            runHook postInstall
           '';
         };
       in rec {
@@ -364,7 +447,9 @@ rec {
                 # C/C++:
                 raylib
                 raygui
+                glew
                 thread-pool
+                libmorton
                 boost
 
                 # Debugging/Testing/Profiling

include/bits.hpp

@@ -0,0 +1,69 @@
+#ifndef BITS_HPP_
+#define BITS_HPP_
+
+#include "util.hpp"
+
+#include <concepts>
+
+template <class T>
+    requires std::unsigned_integral<T>
+// ReSharper disable once CppRedundantInlineSpecifier
+INLINE inline auto create_mask(const u8 first, const u8 last) -> T
+{
+    // If the mask width is equal the type width return all 1s instead of shifting
+    // as shifting by type-width is undefined behavior.
+    if (static_cast<size_t>(last - first + 1) >= sizeof(T) * 8) {
+        return ~T{0};
+    }
+
+    // Example: first=4, last=7, 7-4+1=4
+    //           1 << 4 = 0b00010000
+    //          32  - 1 = 0b00001111
+    //          31 << 4 = 0b11110000
+    // Subtracting 1 generates a consecutive mask.
+    return ((T{1} << (last - first + 1)) - 1) << first;
+}
+
+template <class T>
+    requires std::unsigned_integral<T>
+// ReSharper disable once CppRedundantInlineSpecifier
+INLINE inline auto clear_bits(T& bits, const u8 first, const u8 last) -> void
+{
+    const T mask = create_mask<T>(first, last);
+
+    bits = bits & ~mask;
+}
+
+template <class T, class U>
+    requires std::unsigned_integral<T> && std::unsigned_integral<U>
+// ReSharper disable once CppRedundantInlineSpecifier
+INLINE inline auto set_bits(T& bits, const u8 first, const u8 last, const U value) -> void
+{
+    const T mask = create_mask<T>(first, last);
+
+    // Example: first=4, last=6, value=0b1110,   bits = 0b 01111110
+    //          mask                                  = 0b 01110000
+    //          bits & ~mask                          = 0b 00001110
+    //          value << 4                            = 0b 11100000
+    //          (value << 4) & mask                   = 0b 01100000
+    //          (bits & ~mask) | (value << 4) & mask  = 0b 01101110
+    //                                 Insert position:     ^^^
+    // First clear the bits, then | with the value positioned at the insertion point.
+    // The value may be larger than [first, last], extra bits are ignored.
+    bits = (bits & ~mask) | ((static_cast<T>(value) << first) & mask);
+}
+
+template <class T>
+    requires std::unsigned_integral<T>
+// ReSharper disable once CppRedundantInlineSpecifier
+INLINE inline auto get_bits(const T bits, const u8 first, const u8 last) -> T
+{
+    const T mask = create_mask<T>(first, last);
+
+    // We can >> without sign extension because T is unsigned_integral
+    return (bits & mask) >> first;
+}
+
+auto print_bitmap(u64 bitmap, u8 w, u8 h, const std::string& title) -> void;
+
+#endif

include/config.hpp

@@ -3,9 +3,9 @@
 
 #include <raylib.h>
 
-
-// TODO: Using the octree from the last frame completely breaks the physics :/
-// #define ASYNC_OCTREE
+// Calculate the octree parallel to the layout calculation.
+// Layout uses the octree from last frame.
+#define ASYNC_OCTREE
 
 // Gets set by CMake
 // #define THREADPOOL // Enable physics threadpool
@@ -23,6 +23,7 @@
 #define BS_THREAD_POOL_NATIVE_EXTENSIONS
 // ReSharper disable once CppUnusedIncludeDirective
 #include <BS_thread_pool.hpp>
+using threadpool = std::optional<BS::thread_pool<>* const>;
 #if defined(_WIN32) // raylib uses these names as function parameters
 #undef near
 #undef far
@@ -46,12 +47,14 @@ constexpr int FONT_SIZE = 26;
 // Camera Controls
 constexpr float CAMERA_FOV = 90.0;
 constexpr float FOV_SPEED = 1.0;
+constexpr float FOV_MULTIPLIER = 4.0;
 constexpr float MIN_FOV = 10.0;
-constexpr float MAX_FOV = 180.0;
+constexpr float MAX_PERSP_FOV = 120.0;
+constexpr float MAX_ORTHO_FOV = 540.0;
 constexpr float CAMERA_DISTANCE = 150.0;
-constexpr float ZOOM_SPEED = 2.5;
 constexpr float MIN_CAMERA_DISTANCE = 2.0;
 constexpr float MAX_CAMERA_DISTANCE = 2000.0;
+constexpr float ZOOM_SPEED = 2.5;
 constexpr float ZOOM_MULTIPLIER = 4.0;
 constexpr float PAN_SPEED = 2.0;
 constexpr float PAN_MULTIPLIER = 10.0;
@@ -63,24 +66,27 @@ constexpr float TARGET_UPS = 90;             // How often to update physics
 constexpr float TIMESTEP = 1.0 / TARGET_UPS; // Update interval in seconds
 constexpr float SIM_SPEED = 4.0;             // How large each update should be
 constexpr float MASS = 1.0;                  // Mass spring system
-constexpr float SPRING_CONSTANT = 5.0;       // Mass spring system
-constexpr float DAMPENING_CONSTANT = 1.0;    // Mass spring system
+constexpr float SPRING_K = 4.0;              // Mass spring system
+constexpr float DAMPENING_K = 1.5;           // Mass spring system
 constexpr float REST_LENGTH = 3.0;           // Mass spring system
-constexpr float VERLET_DAMPENING = 0.05;     // [0, 1]
+constexpr float VERLET_DAMPENING = 0.1;      // [0, 1]
 constexpr float BH_FORCE = 2.5;              // Barnes-Hut [1.0, 3.0]
-constexpr float THETA = 0.8;                 // Barnes-Hut [0.5, 1.0]
-constexpr float SOFTENING = 0.01;            // Barnes-Hut [0.01, 1.0]
+constexpr float THETA = 1.0;                 // Barnes-Hut [0.5, 1.0]
+constexpr float SOFTENING = 0.05;            // Barnes-Hut [0.01, 1.0]
 
 // Graph Drawing
-constexpr Color EDGE_COLOR = DARKBLUE;
-constexpr float VERTEX_SIZE = 0.5;
-static const Color VERTEX_COLOR = Fade(BLUE, 0.5);
-constexpr Color VERTEX_VISITED_COLOR = DARKBLUE;
-constexpr Color VERTEX_PATH_COLOR = GREEN;
-constexpr Color VERTEX_TARGET_COLOR = RED;
+static const Color EDGE_COLOR = Fade(BLUE, 0.3);
+constexpr int DRAW_EDGES_LIMIT = 5'000'000;
+constexpr float VERTEX_SIZE = 0.75;
+constexpr int DRAW_VERTICES_LIMIT = 1'000'000;
+static const Color VERTEX_COLOR = Fade(BLUE, 0.8);
+constexpr Color VERTEX_VISITED_COLOR = ORANGE;
 constexpr Color VERTEX_START_COLOR = ORANGE;
-constexpr Color VERTEX_CURRENT_COLOR = PURPLE;
-constexpr int DRAW_VERTICES_LIMIT = 1000000;
+constexpr Color VERTEX_CURRENT_COLOR = ORANGE;
+constexpr Color VERTEX_PATH_COLOR = GREEN;
+constexpr Color VERTEX_TARGET_COLOR = GREEN;
+static const Color VERTEX_CLOSEST_COLOR = Fade(PINK, 1.0);
+static const Color VERTEX_FARTHEST_COLOR = Fade(DARKBLUE, 0.8);
 
 // Klotski Drawing
 constexpr int BOARD_PADDING = 10;
@@ -91,4 +97,8 @@ constexpr Color BLOCK_COLOR = DARKBLUE;
 constexpr Color TARGET_BLOCK_COLOR = RED;
 constexpr Color WALL_COLOR = BLACK;
 
+// Threadpool
+static constexpr int SMALL_TASK_BLOCK_SIZE = 256; // Weirdly larger blocks decrease performance...
+static constexpr int LARGE_TASK_BLOCK_SIZE = 256;
+
 #endif

include/cpu_layout_engine.hpp

@@ -2,6 +2,8 @@
 #define PHYSICS_HPP_
 
 #include "config.hpp"
+#include "cpu_spring_system.hpp"
+#include "util.hpp"
 
 #include <atomic>
 #include <condition_variable>
@@ -13,7 +15,7 @@
 #include <variant>
 #include <vector>
 
-class threaded_physics
+class cpu_layout_engine
 {
     struct add_mass
     {};
@@ -57,24 +59,21 @@ class threaded_physics
     };
 
 private:
-    std::optional<BS::thread_pool<>* const> thread_pool;
+    threadpool thread_pool;
     std::thread physics;
 
 public:
     physics_state state;
 
 public:
-    explicit threaded_physics(
-        const std::optional<BS::thread_pool<>* const> _thread_pool = std::nullopt)
+    explicit cpu_layout_engine(
+        const threadpool _thread_pool = std::nullopt)
         : thread_pool(_thread_pool), physics(physics_thread, std::ref(state), std::ref(thread_pool))
     {}
 
-    threaded_physics(const threaded_physics& copy) = delete;
-    auto operator=(const threaded_physics& copy) -> threaded_physics& = delete;
-    threaded_physics(threaded_physics&& move) = delete;
-    auto operator=(threaded_physics&& move) -> threaded_physics& = delete;
+    NO_COPY_NO_MOVE(cpu_layout_engine);
 
-    ~threaded_physics()
+    ~cpu_layout_engine()
     {
         state.running = false;
         state.data_ready_cnd.notify_all();
@@ -88,14 +87,14 @@ private:
 #endif
 
     static auto physics_thread(physics_state& state,
-                               std::optional<BS::thread_pool<>* const> thread_pool) -> void;
+                               threadpool thread_pool) -> void;
 
 public:
     auto clear_cmd() -> void;
     auto add_mass_cmd() -> void;
     auto add_spring_cmd(size_t a, size_t b) -> void;
     auto add_mass_springs_cmd(size_t num_masses,
-                              const std::vector<std::pair<size_t, size_t>>& springs) -> void;
+                              const std::vector<spring>& springs) -> void;
 };
 
-#endif
+#endif

include/cpu_spring_system.hpp

@@ -0,0 +1,47 @@
+#ifndef MASS_SPRING_SYSTEM_HPP_
+#define MASS_SPRING_SYSTEM_HPP_
+
+#include "octree.hpp"
+#include "config.hpp"
+
+#include <optional>
+#include <raylib.h>
+
+using spring = std::pair<size_t, size_t>;
+
+class cpu_spring_system
+{
+public:
+    octree tree;
+
+    // This is the main ownership of all the states/masses/springs.
+    std::vector<Vector3> positions;
+    std::vector<Vector3> previous_positions; // for verlet integration
+    std::vector<Vector3> velocities;
+    std::vector<Vector3> forces;
+
+    std::vector<spring> springs;
+
+public:
+    cpu_spring_system() {}
+
+    NO_COPY_NO_MOVE(cpu_spring_system);
+
+public:
+    auto clear() -> void;
+    auto add_mass() -> void;
+    auto add_spring(size_t a, size_t b) -> void;
+
+    auto clear_forces() -> void;
+    auto calculate_spring_force(size_t s) -> void;
+    auto calculate_spring_forces(threadpool thread_pool = std::nullopt) -> void;
+    auto calculate_repulsion_forces(threadpool thread_pool = std::nullopt) -> void;
+    auto integrate_velocity(size_t m, float dt) -> void;
+    auto integrate_position(size_t m, float dt) -> void;
+    auto verlet_update(size_t m, float dt) -> void;
+    auto update(float dt, threadpool thread_pool = std::nullopt) -> void;
+
+    auto center_masses(threadpool thread_pool = std::nullopt) -> void;
+};
+
+#endif

include/graph_distances.hpp

@@ -1,7 +1,8 @@
 #ifndef DISTANCE_HPP_
 #define DISTANCE_HPP_
 
-#include <cstddef>
+#include "cpu_spring_system.hpp"
+
 #include <vector>
 
 class graph_distances
@@ -15,7 +16,8 @@ public:
     auto clear() -> void;
     [[nodiscard]] auto empty() const -> bool;
 
-    auto calculate_distances(size_t node_count, const std::vector<std::pair<size_t, size_t>>& edges,
+    auto calculate_distances(size_t node_count,
+                             const std::vector<spring>& edges,
                              const std::vector<size_t>& targets) -> void;
 
     [[nodiscard]] auto get_shortest_path(size_t source) const -> std::vector<size_t>;

include/input_handler.hpp

@@ -22,8 +22,7 @@ struct show_yes_no_message
     std::function<void()> on_yes;
 };
 
-struct show_save_preset_window
-{};
+struct show_save_preset_window {};
 
 using ui_command = std::variant<show_ok_message, show_yes_no_message, show_save_preset_window>;
 
@@ -75,6 +74,7 @@ public:
     bool mark_path = false;
     bool mark_solutions = false;
     bool connect_solutions = false;
+    bool color_by_distance = false;
 
     // Camera
     bool camera_lock = true;
@@ -86,16 +86,17 @@ public:
     Vector2 mouse = Vector2Zero();
     Vector2 last_mouse = Vector2Zero();
 
+    // State selection from graph
+    size_t collision_mass = -1;
+
 public:
-    input_handler(state_manager& _state, orbit_camera& _camera) : state(_state), camera(_camera)
+    input_handler(state_manager& _state, orbit_camera& _camera)
+        : state(_state), camera(_camera)
     {
         init_handlers();
     }
 
-    input_handler(const input_handler& copy) = delete;
-    auto operator=(const input_handler& copy) -> input_handler& = delete;
-    input_handler(input_handler&& move) = delete;
-    auto operator=(input_handler&& move) -> input_handler& = delete;
+    NO_COPY_NO_MOVE(input_handler);
 
 private:
     auto init_handlers() -> void;
@@ -122,6 +123,7 @@ public:
     auto add_block() -> void;
     auto remove_block() -> void;
     auto place_goal() const -> void;
+    auto select_state() const -> void;
 
     // Key actions
     auto toggle_camera_lock() -> void;
@@ -139,6 +141,7 @@ public:
     auto clear_graph() -> void;
     auto toggle_mark_solutions() -> void;
     auto toggle_connect_solutions() -> void;
+    auto toggle_color_by_distance() -> void;
     auto toggle_mark_path() -> void;
     auto goto_optimal_next_state() const -> void;
     auto goto_most_distant_state() const -> void;
@@ -158,18 +161,14 @@ public:
     // General
     auto register_generic_handler(const std::function<void(input_handler&)>& handler) -> void;
 
-    auto register_mouse_pressed_handler(MouseButton button,
-                                        const std::function<void(input_handler&)>& handler) -> void;
+    auto register_mouse_pressed_handler(MouseButton button, const std::function<void(input_handler&)>& handler) -> void;
 
     auto register_mouse_released_handler(MouseButton button,
-                                         const std::function<void(input_handler&)>& handler)
-        -> void;
+                                         const std::function<void(input_handler&)>& handler) -> void;
 
-    auto register_key_pressed_handler(KeyboardKey key,
-                                      const std::function<void(input_handler&)>& handler) -> void;
+    auto register_key_pressed_handler(KeyboardKey key, const std::function<void(input_handler&)>& handler) -> void;
 
-    auto register_key_released_handler(KeyboardKey key,
-                                       const std::function<void(input_handler&)>& handler) -> void;
+    auto register_key_released_handler(KeyboardKey key, const std::function<void(input_handler&)>& handler) -> void;
 
     auto handle_input() -> void;
 };

include/mass_spring_system.hpp

@@ -1,63 +0,0 @@
-#ifndef MASS_SPRING_SYSTEM_HPP_
-#define MASS_SPRING_SYSTEM_HPP_
-
-#include "octree.hpp"
-#include "config.hpp"
-
-#include <optional>
-#include <raylib.h>
-
-class mass_spring_system
-{
-public:
-    class spring
-    {
-    public:
-        size_t a;
-        size_t b;
-
-    public:
-        spring(const size_t _a, const size_t _b)
-            : a(_a), b(_b) {}
-    };
-
-public:
-    static constexpr int SMALL_TASK_BLOCK_SIZE = 256;
-    static constexpr int LARGE_TASK_BLOCK_SIZE = 256;
-
-    octree tree;
-
-    // This is the main ownership of all the states/masses/springs.
-    std::vector<Vector3> positions;
-    std::vector<Vector3> previous_positions; // for verlet integration
-    std::vector<Vector3> velocities;
-    std::vector<Vector3> forces;
-
-    std::vector<spring> springs;
-
-public:
-    mass_spring_system() {}
-
-    mass_spring_system(const mass_spring_system& copy) = delete;
-    auto operator=(const mass_spring_system& copy) -> mass_spring_system& = delete;
-    mass_spring_system(mass_spring_system& move) = delete;
-    auto operator=(mass_spring_system&& move) -> mass_spring_system& = delete;
-
-public:
-    auto clear() -> void;
-    auto add_mass() -> void;
-    auto add_spring(size_t a, size_t b) -> void;
-
-    auto clear_forces() -> void;
-    auto calculate_spring_force(size_t s) -> void;
-    auto calculate_spring_forces(std::optional<BS::thread_pool<>* const> thread_pool = std::nullopt) -> void;
-    auto calculate_repulsion_forces(std::optional<BS::thread_pool<>* const> thread_pool = std::nullopt) -> void;
-    auto integrate_velocity(size_t m, float dt) -> void;
-    auto integrate_position(size_t m, float dt) -> void;
-    auto verlet_update(size_t m, float dt) -> void;
-    auto update(float dt, std::optional<BS::thread_pool<>* const> thread_pool = std::nullopt) -> void;
-
-    auto center_masses(std::optional<BS::thread_pool<>* const> thread_pool = std::nullopt) -> void;
-};
-
-#endif

include/octree.hpp

@@ -1,11 +1,15 @@
 #ifndef OCTREE_HPP_
 #define OCTREE_HPP_
 
+#include "util.hpp"
+#include "config.hpp"
+
 #include <array>
 #include <vector>
 
 #include <raylib.h>
 #include <raymath.h>
+#include <libmorton/morton.h>
 
 class octree
 {
@@ -14,38 +18,193 @@ class octree
     public:
         Vector3 mass_center = Vector3Zero();
         float mass_total = 0.0;
-        Vector3 box_min = Vector3Zero(); // area start
-        Vector3 box_max = Vector3Zero(); // area end
+        u8 depth = 0;
+        float size = 0.0f; // Because our octree cells are cubic we don't need to store the bounds
         std::array<int, 8> children = {-1, -1, -1, -1, -1, -1, -1, -1};
         int mass_id = -1;
         bool leaf = true;
-
-    public:
-        [[nodiscard]] auto child_count() const -> int;
     };
 
-public:
-    static constexpr int MAX_DEPTH = 20;
+private:
+    // 21 * 3 = 63, fits in u64 for combined x/y/z morton-code
+    static constexpr int MAX_DEPTH = 21;
 
     std::vector<node> nodes;
 
+    // This approach is actually slower than the array of nodes
+    // beacuse we access all the attributes in the same function
+    // std::vector<Vector3> mass_centers;
+    // std::vector<float> mass_totals;
+    // std::vector<Vector3> box_mins;
+    // std::vector<Vector3> box_maxs;
+    // std::vector<std::array<int, 8>> childrens;
+    // std::vector<int> mass_ids;
+    // std::vector<u8> leafs; // bitpacked std::vector<bool> is a lot slower
+
 public:
     octree() = default;
 
-    // octree(const octree& copy) = delete;
-    // auto operator=(const octree& copy) -> octree& = delete;
-    // octree(octree&& move) = delete;
-    // auto operator=(octree&& move) -> octree& = delete;
+    // Required for async octree
+    // NO_COPY_NO_MOVE(octree);
+
+private:
+    [[nodiscard]] INLINE static inline auto get_octant(const Vector3& box_min,
+                                                       const Vector3& box_max,
+                                                       const Vector3& pos) -> int;
+    [[nodiscard]] INLINE static inline auto get_child_bounds(const Vector3& box_min,
+                                                             const Vector3& box_max,
+                                                             int octant) -> std::pair<Vector3, Vector3>;
+
+    // Map a floating point coordinate to a discrete integer (so its morton-code can be computed)
+    // The "bits" parameter determines the discrete axis resolution
+    [[nodiscard]] INLINE static inline auto quantize_axis(float coordinate,
+                                                          float box_min,
+                                                          float box_max,
+                                                          int bits) -> u32;
+
+    [[nodiscard]] INLINE static inline auto pos_to_morton(const Vector3& p,
+                                                          const Vector3& root_min,
+                                                          const Vector3& root_max) -> u64;
+
+    [[nodiscard]] INLINE static inline auto jitter_pos(Vector3 p,
+                                                       u32 seed,
+                                                       const Vector3& root_min,
+                                                       const Vector3& root_max,
+                                                       float root_extent) -> Vector3;
+
+    // Use this to obtain an ancestor node of a leaf node (on any level).
+    // Because the morton codes (interleaved coordinates) encode the octree path, we can take
+    // the morten code of any leaf and only take the 3*n first interleaved bits to find the
+    // leaf ancestor on level n.
+    // Leaf Code: [101 110 100 001] -> Ancestors (from leaf to root):
+    // - [101 110 100]
+    // - [101 110]
+    // - [101] (root)
+    [[nodiscard]] INLINE static inline auto path_to_ancestor(u64 leaf_code, int leaf_depth, int depth) -> u64;
+
+    // Use this to obtain the octant a leaf node is contained in (on any level).
+    // The 3 interleaved bits in the morten code encode the octant [0, 7].
+    // Leaf Code: [101 110 100 001] -> Octants:
+    // - [100] (Level 2)
+    // - [110] (Level 1)
+    // - [101] (Level 0)
+    [[nodiscard]] INLINE static inline auto octant_at_level(u64 leaf_code, int level, int leaf_depth) -> int;
 
 public:
-    [[nodiscard]] auto get_octant(int node_idx, const Vector3& pos) const -> int;
-    [[nodiscard]] auto get_child_bounds(int node_idx, int octant) const
-        -> std::pair<Vector3, Vector3>;
-    auto create_empty_leaf(const Vector3& box_min, const Vector3& box_max) -> int;
-    auto insert(int node_idx, int mass_id, const Vector3& pos, float mass, int depth) -> void;
-    static auto build_octree(octree& t, const std::vector<Vector3>& positions) -> void;
+    auto clear() -> void;
+    auto reserve(size_t count) -> void;
+    [[nodiscard]] auto empty() const -> bool;
+    [[nodiscard]] auto root() const -> const node&;
 
-    [[nodiscard]] auto calculate_force(int node_idx, const Vector3& pos) const -> Vector3;
+    // Morton/linear octree implementation
+    static auto build_octree_morton(octree& t,
+                                    const std::vector<Vector3>& positions,
+                                    const std::optional<BS::thread_pool<>*>& thread_pool) -> void;
+    [[nodiscard]] auto calculate_force_morton(int node_idx, const Vector3& pos, int self_id) const -> Vector3;
 };
 
+INLINE inline auto octree::get_octant(const Vector3& box_min, const Vector3& box_max, const Vector3& pos) -> int
+{
+    auto [cx, cy, cz] = (box_min + box_max) / 2.0f;
+
+    // The octant is encoded as a 3-bit integer "zyx". The node area is split
+    // along all 3 axes, if a position is right of an axis, this bit is set to 1.
+    // If a position is right of the x-axis and y-axis and left of the z-axis, the
+    // encoded octant is "011".
+    return (pos.x >= cx) | ((pos.y >= cy) << 1) | ((pos.z >= cz) << 2);
+}
+
+INLINE inline auto octree::get_child_bounds(const Vector3& box_min,
+                                            const Vector3& box_max,
+                                            const int octant) -> std::pair<Vector3, Vector3>
+{
+    auto [cx, cy, cz] = (box_min + box_max) / 2.0f;
+
+    Vector3 min = Vector3Zero();
+    Vector3 max = Vector3Zero();
+
+    // If (octant & 1), the octant is to the right of the node region's x-axis
+    // (see GetOctant). This means the left bound is the x-axis and the right
+    // bound the node's region max.
+    min.x = octant & 1 ? cx : box_min.x;
+    max.x = octant & 1 ? box_max.x : cx;
+    min.y = octant & 2 ? cy : box_min.y;
+    max.y = octant & 2 ? box_max.y : cy;
+    min.z = octant & 4 ? cz : box_min.z;
+    max.z = octant & 4 ? box_max.z : cz;
+
+    return std::make_pair(min, max);
+}
+
+INLINE inline auto octree::quantize_axis(const float coordinate,
+                                         const float box_min,
+                                         const float box_max,
+                                         const int bits) -> u32
+{
+    const float extent = box_max - box_min;
+    if (extent <= 0.0f) {
+        return 0;
+    }
+
+    float normalized = (coordinate - box_min) / extent;                            // normalize to [0,1]
+    normalized = std::max(0.0f, std::min(normalized, std::nextafter(1.0f, 0.0f))); // avoid exactly 1.0
+
+    // bits up to 21 => (1u << bits) safe in 32-bit
+    const u32 grid_max = (1u << bits) - 1u;
+    return static_cast<u32>(normalized * static_cast<float>(grid_max));
+}
+
+INLINE inline auto octree::pos_to_morton(const Vector3& p, const Vector3& root_min, const Vector3& root_max) -> u64
+{
+    const u32 x = quantize_axis(p.x, root_min.x, root_max.x, MAX_DEPTH);
+    const u32 y = quantize_axis(p.y, root_min.y, root_max.y, MAX_DEPTH);
+    const u32 z = quantize_axis(p.z, root_min.z, root_max.z, MAX_DEPTH);
+    return libmorton::morton3D_64_encode(x, y, z);
+}
+
+INLINE inline auto octree::jitter_pos(Vector3 p,
+                                      const u32 seed,
+                                      const Vector3& root_min,
+                                      const Vector3& root_max,
+                                      const float root_extent) -> Vector3
+{
+    // Use a hash to calculate a deterministic jitter: The same position should always get the same jitter.
+    // We want this to get stable physics, particles at the same position shouldn't get different jitters
+    // across frames...
+    u32 h = (seed ^ 61u) ^ (seed >> 16);
+    h *= 9u;
+    h = h ^ (h >> 4);
+    h *= 0x27d4eb2du;
+    h = h ^ (h >> 15);
+
+    // finest cell size at depth L
+    const float finest_cell = root_extent / static_cast<float>(1u << MAX_DEPTH);
+    const float s = finest_cell * 1e-4f; // small pp
+
+    p.x += (h & 1u) ? +s : -s;
+    p.y += (h & 2u) ? +s : -s;
+    p.z += (h & 4u) ? +s : -s;
+
+    // clamp back into bounds just in case
+    p.x = std::max(root_min.x, std::min(p.x, root_max.x));
+    p.y = std::max(root_min.y, std::min(p.y, root_max.y));
+    p.z = std::max(root_min.z, std::min(p.z, root_max.z));
+
+    return p;
+}
+
+INLINE inline auto octree::path_to_ancestor(const u64 leaf_code, const int leaf_depth, const int depth) -> u64
+{
+    // keep top 3*depth bits; drop the rest
+    const int drop = 3 * (leaf_depth - depth);
+    return (drop > 0) ? (leaf_code >> drop) : leaf_code;
+}
+
+INLINE inline auto octree::octant_at_level(const u64 leaf_code, const int level, const int leaf_depth) -> int
+{
+    // level 1 => child of root => topmost 3 bits
+    const int shift = 3 * (leaf_depth - level);
+    return static_cast<int>((leaf_code >> shift) & 0x7ull);
+}
+
 #endif

include/orbit_camera.hpp

@@ -24,8 +24,7 @@ public:
 
     auto pan(Vector2 last_mouse, Vector2 mouse) -> void;
 
-    auto update(const Vector3& current_target, const Vector3& mass_center, bool lock,
-                bool mass_center_lock) -> void;
+    auto update(const Vector3& current_target, const Vector3& mass_center, bool lock, bool mass_center_lock) -> void;
 };
 
 #endif

include/renderer.hpp

@@ -14,55 +14,73 @@ class renderer
 {
 private:
     const state_manager& state;
-    const input_handler& input;
+    input_handler& input;
     user_interface& gui;
 
     const orbit_camera& camera;
-    RenderTexture graph_target =
-        LoadRenderTexture(GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT);
+    RenderTexture graph_target = LoadRenderTexture(GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT);
 
     // TODO: Those should be moved to the user_interface.h
-    RenderTexture klotski_target =
-        LoadRenderTexture(GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT);
+    RenderTexture klotski_target = LoadRenderTexture(GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT);
     RenderTexture menu_target = LoadRenderTexture(GetScreenWidth(), MENU_HEIGHT);
 
-    // Batching
+    // Edges
+    unsigned int edge_vao_id = 0;
+    unsigned int edge_vbo_id = 0;
+    std::vector<Vector3> edge_vertices;
+    Shader edge_shader = LoadShader("shader/edge_vertex.glsl", "shader/edge_fragment.glsl");
+    int edge_color_loc = -1;
     std::vector<std::pair<Vector3, Vector3>> connections;
 
-    // Instancing
+    // Vertex instancing
     static constexpr int INSTANCE_COLOR_ATTR = 5;
     std::vector<Matrix> transforms;
     std::vector<Color> colors;
     Material vertex_mat = LoadMaterialDefault();
     Mesh cube_instance = GenMeshCube(VERTEX_SIZE, VERTEX_SIZE, VERTEX_SIZE);
-    Shader instancing_shader =
-        LoadShader("shader/instancing_vertex.glsl", "shader/instancing_fragment.glsl");
-
+    Shader instancing_shader = LoadShader("shader/instancing_vertex.glsl", "shader/instancing_fragment.glsl");
     unsigned int color_vbo_id = 0;
 
 public:
-    renderer(const orbit_camera& _camera, const state_manager& _state, const input_handler& _input,
-             user_interface& _gui)
+    // TODO: I am allocating HUGE vertex buffers instead of resizing dynamically...
+    //       Edges: 5'000'000 * 2 * 12 Byte ~= 115 MB
+    //       Verts: 1'000'000 * 16 Byte ~= 15 MB
+    //       This is also allocated on the CPU by the vectors
+    renderer(const orbit_camera& _camera, const state_manager& _state, input_handler& _input, user_interface& _gui)
         : state(_state), input(_input), gui(_gui), camera(_camera)
     {
+        // Edges
+        edge_shader.locs[SHADER_LOC_VERTEX_POSITION] = GetShaderLocationAttrib(edge_shader, "vertexPosition");
+        edge_shader.locs[SHADER_LOC_MATRIX_MVP] = GetShaderLocation(edge_shader, "mvp");
+        edge_shader.locs[SHADER_LOC_COLOR_DIFFUSE] = GetShaderLocation(edge_shader, "colDiffuse");
+        edge_color_loc = GetShaderLocation(edge_shader, "colDiffuse");
+
+        edge_vertices.reserve(DRAW_EDGES_LIMIT * 2);
+
+        edge_vao_id = rlLoadVertexArray();
+        edge_vbo_id = rlLoadVertexBuffer(nullptr, DRAW_EDGES_LIMIT * 2 * sizeof(Vector3), true);
+
+        rlEnableVertexArray(edge_vao_id);
+        rlEnableVertexBuffer(edge_vbo_id);
+
+        rlSetVertexAttribute(0, 3, RL_FLOAT, false, sizeof(Vector3), 0);
+        rlEnableVertexAttribute(0);
+
+        rlDisableVertexBuffer();
+        rlDisableVertexArray();
+
+        // Vertex instancing
         instancing_shader.locs[SHADER_LOC_MATRIX_MVP] = GetShaderLocation(instancing_shader, "mvp");
-        instancing_shader.locs[SHADER_LOC_MATRIX_MODEL] =
-            GetShaderLocationAttrib(instancing_shader, "instanceTransform");
-        instancing_shader.locs[SHADER_LOC_VECTOR_VIEW] =
-            GetShaderLocation(instancing_shader, "viewPos");
+        instancing_shader.locs[SHADER_LOC_MATRIX_MODEL] = GetShaderLocationAttrib(
+            instancing_shader,
+            "instanceTransform");
+        instancing_shader.locs[SHADER_LOC_VECTOR_VIEW] = GetShaderLocation(instancing_shader, "viewPos");
 
-        // infoln("LOC vertexPosition: {}",
-        //        rlGetLocationAttrib(instancing_shader.id, "vertexPosition"));
-        // infoln("LOC instanceTransform: {}",
-        //        rlGetLocationAttrib(instancing_shader.id, "instanceTransform"));
-        // infoln("LOC instanceColor: {}", rlGetLocationAttrib(instancing_shader.id, "instanceColor"));
-
-        // vertex_mat.maps[MATERIAL_MAP_DIFFUSE].color = VERTEX_COLOR;
         vertex_mat.shader = instancing_shader;
 
         transforms.reserve(DRAW_VERTICES_LIMIT);
         colors.reserve(DRAW_VERTICES_LIMIT);
-        color_vbo_id = rlLoadVertexBuffer(colors.data(), DRAW_VERTICES_LIMIT * sizeof(Color), true);
+        color_vbo_id = rlLoadVertexBuffer(nullptr, DRAW_VERTICES_LIMIT * sizeof(Color), true);
 
         rlEnableVertexArray(cube_instance.vaoId);
         rlEnableVertexBuffer(color_vbo_id);
@@ -74,10 +92,7 @@ public:
         rlDisableVertexArray();
     }
 
-    renderer(const renderer& copy) = delete;
-    auto operator=(const renderer& copy) -> renderer& = delete;
-    renderer(renderer&& move) = delete;
-    auto operator=(renderer&& move) -> renderer& = delete;
+    NO_COPY_NO_MOVE(renderer);
 
     ~renderer()
     {
@@ -85,12 +100,19 @@ public:
         UnloadRenderTexture(klotski_target);
         UnloadRenderTexture(menu_target);
 
+        // Edges
+        rlUnloadVertexArray(edge_vao_id);
+        rlUnloadVertexBuffer(edge_vbo_id);
+        UnloadShader(edge_shader);
+
         // Instancing
         UnloadMaterial(vertex_mat);
         UnloadMesh(cube_instance);
 
         // I think the shader already gets unloaded with the material?
         // UnloadShader(instancing_shader);
+
+        rlUnloadVertexBuffer(color_vbo_id);
     }
 
 private:
@@ -102,8 +124,7 @@ private:
     auto draw_textures(int fps, int ups, size_t mass_count, size_t spring_count) const -> void;
 
 public:
-    auto render(const std::vector<Vector3>& masses, int fps, int ups, size_t mass_count,
-                size_t spring_count) -> void;
+    auto render(const std::vector<Vector3>& masses, int fps, int ups, size_t mass_count, size_t spring_count) -> void;
 };
 
 #endif

include/state_manager.hpp

@@ -3,7 +3,7 @@
 
 #include "graph_distances.hpp"
 #include "load_save.hpp"
-#include "threaded_physics.hpp"
+#include "cpu_layout_engine.hpp"
 #include "puzzle.hpp"
 
 #include <boost/unordered/unordered_flat_map.hpp>
@@ -12,7 +12,7 @@
 class state_manager
 {
 private:
-    threaded_physics& physics;
+    cpu_layout_engine& physics;
 
     std::string preset_file;
     size_t current_preset = 0;
@@ -22,9 +22,9 @@ private:
     // State storage (store states twice for bidirectional lookup).
     // Everything else should only store indices to state_pool.
 
-    std::vector<puzzle> state_pool; // Indices are equal to mass_springs mass indices
-    boost::unordered_flat_map<puzzle, size_t, puzzle_hasher> state_indices; // Maps states to indices
-    std::vector<std::pair<size_t, size_t>> links; // Indices are equal to mass_springs springs indices
+    std::vector<puzzle> state_pool;  // Indices are equal to mass_springs mass indices
+    puzzlemap<size_t> state_indices; // Maps states to indices
+    std::vector<spring> links;       // Indices are equal to mass_springs springs indices
 
     graph_distances node_target_distances; // Buffered and reused if the graph doesn't change
     boost::unordered_flat_set<size_t> winning_indices; // Indices of all states where the board is solved
@@ -42,16 +42,13 @@ private:
     bool edited = false;
 
 public:
-    state_manager(threaded_physics& _physics, const std::string& _preset_file)
+    state_manager(cpu_layout_engine& _physics, const std::string& _preset_file)
         : physics(_physics), preset_file(_preset_file)
     {
         reload_preset_file();
     }
 
-    state_manager(const state_manager& copy) = delete;
-    auto operator=(const state_manager& copy) -> state_manager& = delete;
-    state_manager(state_manager&& move) = delete;
-    auto operator=(state_manager&& move) -> state_manager& = delete;
+    NO_COPY_NO_MOVE(state_manager);
 
 private:
     /**
@@ -83,8 +80,7 @@ private:
      * @param states List of states to insert
      * @param _links List of links to insert
      */
-    auto synced_insert_statespace(const std::vector<puzzle>& states,
-                                  const std::vector<std::pair<size_t, size_t>>& _links) -> void;
+    auto synced_insert_statespace(const std::vector<puzzle>& states, const std::vector<spring>& _links) -> void;
 
     /**
      * Clears all states and links (and related) from the state_manager and the physics system.
@@ -130,7 +126,7 @@ public:
     [[nodiscard]] auto get_target_count() const -> size_t;
     [[nodiscard]] auto get_link_count() const -> size_t;
     [[nodiscard]] auto get_path_length() const -> size_t;
-    [[nodiscard]] auto get_links() const -> const std::vector<std::pair<size_t, size_t>>&;
+    [[nodiscard]] auto get_links() const -> const std::vector<spring>&;
     [[nodiscard]] auto get_winning_indices() const -> const boost::unordered_flat_set<size_t>&;
     [[nodiscard]] auto get_visit_counts() const -> const boost::unordered_flat_map<size_t, int>&;
     [[nodiscard]] auto get_winning_path() const -> const std::vector<size_t>&;
@@ -141,6 +137,7 @@ public:
     [[nodiscard]] auto get_current_preset_comment() const -> const std::string&;
     [[nodiscard]] auto has_history() const -> bool;
     [[nodiscard]] auto has_distances() const -> bool;
+    [[nodiscard]] auto get_distances() const -> std::vector<int>;
     [[nodiscard]] auto get_total_moves() const -> size_t;
     [[nodiscard]] auto was_edited() const -> bool;
 };

include/user_interface.hpp

@@ -22,15 +22,17 @@ class user_interface
         const int padding;
 
     public:
-        grid(const int _x, const int _y, const int _width, const int _height, const int _columns,
-             const int _rows, const int _padding)
-            : x(_x), y(_y), width(_width), height(_height), columns(_columns), rows(_rows),
-              padding(_padding)
-        {}
+        grid(const int _x,
+             const int _y,
+             const int _width,
+             const int _height,
+             const int _columns,
+             const int _rows,
+             const int _padding)
+            : x(_x), y(_y), width(_width), height(_height), columns(_columns), rows(_rows), padding(_padding) {}
 
     public:
-        auto update_bounds(int _x, int _y, int _width, int _height, int _columns, int _rows)
-            -> void;
+        auto update_bounds(int _x, int _y, int _width, int _height, int _columns, int _rows) -> void;
         auto update_bounds(int _x, int _y, int _width, int _height) -> void;
         auto update_bounds(int _x, int _y) -> void;
 
@@ -38,8 +40,7 @@ class user_interface
         [[nodiscard]] auto bounds(int _x, int _y, int _width, int _height) const -> Rectangle;
 
         [[nodiscard]] auto square_bounds() const -> Rectangle;
-        [[nodiscard]] auto square_bounds(int _x, int _y, int _width, int _height) const
-            -> Rectangle;
+        [[nodiscard]] auto square_bounds(int _x, int _y, int _width, int _height) const -> Rectangle;
     };
 
     struct style
@@ -87,14 +88,17 @@ private:
 
     grid menu_grid = grid(0, 0, GetScreenWidth(), MENU_HEIGHT, MENU_COLS, MENU_ROWS, MENU_PAD);
 
-    grid board_grid =
-        grid(0, MENU_HEIGHT, GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT,
-             state.get_current_state().get_width(), state.get_current_state().get_height(), BOARD_PADDING);
+    grid board_grid = grid(0,
+                           MENU_HEIGHT,
+                           GetScreenWidth() / 2,
+                           GetScreenHeight() - MENU_HEIGHT,
+                           state.get_current_state().get_width(),
+                           state.get_current_state().get_height(),
+                           BOARD_PADDING);
 
     grid graph_overlay_grid = grid(GetScreenWidth() / 2, MENU_HEIGHT, 200, 100, 1, 4, MENU_PAD);
 
-    grid debug_overlay_grid =
-        grid(GetScreenWidth() / 2, GetScreenHeight() - 75, 200, 75, 1, 3, MENU_PAD);
+    grid debug_overlay_grid = grid(GetScreenWidth() / 2, GetScreenHeight() - 75, 200, 75, 1, 3, MENU_PAD);
 
     // Windows
 
@@ -114,10 +118,7 @@ public:
         init();
     }
 
-    user_interface(const user_interface& copy) = delete;
-    auto operator=(const user_interface& copy) -> user_interface& = delete;
-    user_interface(user_interface&& move) = delete;
-    auto operator=(user_interface&& move) -> user_interface& = delete;
+    NO_COPY_NO_MOVE(user_interface);
 
 private:
     static auto init() -> void;
@@ -133,32 +134,68 @@ private:
 
     [[nodiscard]] static auto popup_bounds() -> Rectangle;
 
-    auto draw_button(Rectangle bounds, const std::string& label, Color color, bool enabled = true,
+    auto draw_button(Rectangle bounds,
+                     const std::string& label,
+                     Color color,
+                     bool enabled = true,
                      int font_size = FONT_SIZE) const -> int;
 
-    auto draw_menu_button(int x, int y, int width, int height, const std::string& label,
-                          Color color, bool enabled = true, int font_size = FONT_SIZE) const -> int;
+    auto draw_menu_button(int x,
+                          int y,
+                          int width,
+                          int height,
+                          const std::string& label,
+                          Color color,
+                          bool enabled = true,
+                          int font_size = FONT_SIZE) const -> int;
 
-    auto draw_toggle_slider(Rectangle bounds, const std::string& off_label,
-                            const std::string& on_label, int* active, Color color,
-                            bool enabled = true, int font_size = FONT_SIZE) const -> int;
+    auto draw_toggle_slider(Rectangle bounds,
+                            const std::string& off_label,
+                            const std::string& on_label,
+                            int* active,
+                            Color color,
+                            bool enabled = true,
+                            int font_size = FONT_SIZE) const -> int;
 
-    auto draw_menu_toggle_slider(int x, int y, int width, int height, const std::string& off_label,
-                                 const std::string& on_label, int* active, Color color,
-                                 bool enabled = true, int font_size = FONT_SIZE) const -> int;
+    auto draw_menu_toggle_slider(int x,
+                                 int y,
+                                 int width,
+                                 int height,
+                                 const std::string& off_label,
+                                 const std::string& on_label,
+                                 int* active,
+                                 Color color,
+                                 bool enabled = true,
+                                 int font_size = FONT_SIZE) const -> int;
 
-    auto draw_spinner(Rectangle bounds, const std::string& label, int* value, int min, int max,
-                      Color color, bool enabled = true, int font_size = FONT_SIZE) const -> int;
+    auto draw_spinner(Rectangle bounds,
+                      const std::string& label,
+                      int* value,
+                      int min,
+                      int max,
+                      Color color,
+                      bool enabled = true,
+                      int font_size = FONT_SIZE) const -> int;
 
-    auto draw_menu_spinner(int x, int y, int width, int height, const std::string& label,
-                           int* value, int min, int max, Color color, bool enabled = true,
+    auto draw_menu_spinner(int x,
+                           int y,
+                           int width,
+                           int height,
+                           const std::string& label,
+                           int* value,
+                           int min,
+                           int max,
+                           Color color,
+                           bool enabled = true,
                            int font_size = FONT_SIZE) const -> int;
 
-    auto draw_label(Rectangle bounds, const std::string& text, Color color, bool enabled = true,
+    auto draw_label(Rectangle bounds,
+                    const std::string& text,
+                    Color color,
+                    bool enabled = true,
                     int font_size = FONT_SIZE) const -> int;
 
-    auto draw_board_block(int x, int y, int width, int height, Color color,
-                          bool enabled = true) const -> bool;
+    auto draw_board_block(int x, int y, int width, int height, Color color, bool enabled = true) const -> bool;
 
     [[nodiscard]] auto window_open() const -> bool;
 

include/util.hpp

@@ -1,13 +1,67 @@
 #ifndef UTIL_HPP_
 #define UTIL_HPP_
 
+#include <vector>
 #include <iostream>
 #include <raylib.h>
 
 #define INLINE __attribute__((always_inline))
 #define PACKED __attribute__((packed))
 
-enum ctrl
+#define STARTTIME const auto start = std::chrono::high_resolution_clock::now()
+#define ENDTIME(msg, cast, unit) const auto end = std::chrono::high_resolution_clock::now(); \
+    infoln("{}. Took {}{}.", msg, std::chrono::duration_cast<cast>(end - start).count(), unit)
+
+#define COMMENT if (false)
+
+#define NO_COPY_NO_MOVE(typename) \
+    typename(const typename& copy) = delete; \
+    auto operator=(const typename& copy) -> typename& = delete; \
+    typename(typename&& move) = delete; \
+    auto operator=(typename&& move) -> typename& = delete;
+
+using u8 = uint8_t;
+using u16 = uint16_t;
+using u32 = uint32_t;
+using u64 = uint64_t;
+
+using i8 = int8_t;
+using i16 = int16_t;
+using i32 = int32_t;
+using i64 = int64_t;
+
+// https://en.cppreference.com/w/cpp/utility/variant/visit
+template <class... Ts>
+struct overloads : Ts...
+{
+    using Ts::operator()...;
+};
+
+inline auto binom(const int n, const int k) -> int
+{
+    std::vector<int> solutions(k);
+    solutions[0] = n - k + 1;
+
+    for (int i = 1; i < k; ++i) {
+        solutions[i] = solutions[i - 1] * (n - k + 1 + i) / (i + 1);
+    }
+
+    return solutions[k - 1];
+}
+
+// Enums
+
+enum dir : u8
+{
+    nor = 1 << 0,
+    eas = 1 << 1,
+    sou = 1 << 2,
+    wes = 1 << 3,
+};
+
+// Ansi
+
+enum class ctrl : u8
 {
     reset = 0,
     bold_bright = 1,
@@ -18,120 +72,40 @@ enum ctrl
     inverse_off = 27
 };
 
-enum fg
+enum class fg : u8
 {
-    fg_black = 30,
-    fg_red = 31,
-    fg_green = 32,
-    fg_yellow = 33,
-    fg_blue = 34,
-    fg_magenta = 35,
-    fg_cyan = 36,
-    fg_white = 37
+    black = 30,
+    red = 31,
+    green = 32,
+    yellow = 33,
+    blue = 34,
+    magenta = 35,
+    cyan = 36,
+    white = 37
 };
 
-enum bg
+enum class bg : u8
 {
-    bg_black = 40,
-    bg_red = 41,
-    bg_green = 42,
-    bg_yellow = 43,
-    bg_blue = 44,
-    bg_magenta = 45,
-    bg_cyan = 46,
-    bg_white = 47
+    black = 40,
+    red = 41,
+    green = 42,
+    yellow = 43,
+    blue = 44,
+    magenta = 45,
+    cyan = 46,
+    white = 47
 };
 
 inline auto ansi_bold_fg(const fg color) -> std::string
 {
-    return std::format("\033[{};{}m", static_cast<int>(bold_bright), static_cast<int>(color));
+    return std::format("\033[{};{}m", static_cast<int>(ctrl::bold_bright), static_cast<int>(color));
 }
 
 inline auto ansi_reset() -> std::string
 {
-    return std::format("\033[{}m", static_cast<int>(reset));
+    return std::format("\033[{}m", static_cast<int>(ctrl::reset));
 }
 
-// Bit shifting + masking
-
-template <class T>
-    requires std::unsigned_integral<T>
-// ReSharper disable once CppRedundantInlineSpecifier
-INLINE inline auto create_mask(const uint8_t first, const uint8_t last) -> T
-{
-    // If the mask width is equal the type width return all 1s instead of shifting
-    // as shifting by type-width is undefined behavior.
-    if (static_cast<size_t>(last - first + 1) >= sizeof(T) * 8) {
-        return ~T{0};
-    }
-
-    // Example: first=4, last=7, 7-4+1=4
-    //           1 << 4 = 0b00010000
-    //          32  - 1 = 0b00001111
-    //          31 << 4 = 0b11110000
-    // Subtracting 1 generates a consecutive mask.
-    return ((T{1} << (last - first + 1)) - 1) << first;
-}
-
-template <class T>
-    requires std::unsigned_integral<T>
-// ReSharper disable once CppRedundantInlineSpecifier
-INLINE inline auto clear_bits(T& bits, const uint8_t first, const uint8_t last) -> void
-{
-    const T mask = create_mask<T>(first, last);
-
-    bits = bits & ~mask;
-}
-
-template <class T, class U>
-    requires std::unsigned_integral<T> && std::unsigned_integral<U>
-// ReSharper disable once CppRedundantInlineSpecifier
-INLINE inline auto set_bits(T& bits, const uint8_t first, const uint8_t last, const U value) -> void
-{
-    const T mask = create_mask<T>(first, last);
-
-    // Example: first=4, last=6, value=0b1110,   bits = 0b 01111110
-    //          mask                                  = 0b 01110000
-    //          bits & ~mask                          = 0b 00001110
-    //          value << 4                            = 0b 11100000
-    //          (value << 4) & mask                   = 0b 01100000
-    //          (bits & ~mask) | (value << 4) & mask  = 0b 01101110
-    //                                 Insert position:     ^^^
-    // First clear the bits, then | with the value positioned at the insertion point.
-    // The value may be larger than [first, last], extra bits are ignored.
-    bits = (bits & ~mask) | ((static_cast<T>(value) << first) & mask);
-}
-
-template <class T>
-    requires std::unsigned_integral<T>
-// ReSharper disable once CppRedundantInlineSpecifier
-INLINE inline auto get_bits(const T bits, const uint8_t first, const uint8_t last) -> T
-{
-    const T mask = create_mask<T>(first, last);
-
-    // We can >> without sign extension because T is unsigned_integral
-    return (bits & mask) >> first;
-}
-
-// std::variant visitor
-
-// https://en.cppreference.com/w/cpp/utility/variant/visit
-template <class... Ts>
-struct overloads : Ts...
-{
-    using Ts::operator()...;
-};
-
-// Enums
-
-enum direction
-{
-    nor = 1 << 0,
-    eas = 1 << 1,
-    sou = 1 << 2,
-    wes = 1 << 3,
-};
-
 // Output
 
 inline auto operator<<(std::ostream& os, const Vector2& v) -> std::ostream&
@@ -150,44 +124,33 @@ inline auto operator<<(std::ostream& os, const Vector3& v) -> std::ostream&
 template <typename... Args>
 auto traceln(std::format_string<Args...> fmt, Args&&... args) -> void
 {
-    std::cout << std::format("[{}TRACE{}]: ", ansi_bold_fg(fg_cyan), ansi_reset()) << std::format(
-        fmt, std::forward<Args>(args)...) << std::endl;
+    std::cout << std::format("[{}TRACE{}]: ", ansi_bold_fg(fg::cyan), ansi_reset()) << std::format(
+        fmt,
+        std::forward<Args>(args)...) << std::endl;
 }
 
 template <typename... Args>
 auto infoln(std::format_string<Args...> fmt, Args&&... args) -> void
 {
-    std::cout << std::format("[{}INFO{}]: ", ansi_bold_fg(fg_blue), ansi_reset()) << std::format(
-        fmt, std::forward<Args>(args)...) << std::endl;
+    std::cout << std::format("[{}INFO{}]: ", ansi_bold_fg(fg::blue), ansi_reset()) << std::format(
+        fmt,
+        std::forward<Args>(args)...) << std::endl;
 }
 
 template <typename... Args>
 auto warnln(std::format_string<Args...> fmt, Args&&... args) -> void
 {
-    std::cout << std::format("[{}WARNING{}]: ", ansi_bold_fg(fg_yellow), ansi_reset()) << std::format(
-        fmt, std::forward<Args>(args)...) << std::endl;
+    std::cout << std::format("[{}WARNING{}]: ", ansi_bold_fg(fg::yellow), ansi_reset()) << std::format(
+        fmt,
+        std::forward<Args>(args)...) << std::endl;
 }
 
 template <typename... Args>
 auto errln(std::format_string<Args...> fmt, Args&&... args) -> void
 {
-    std::cout << std::format("[{}ERROR{}]: ", ansi_bold_fg(fg_red), ansi_reset()) << std::format(
-        fmt, std::forward<Args>(args)...) << std::endl;
-}
-
-inline auto print_bitmap(const uint64_t bitmap, const uint8_t w, const uint8_t h, const std::string& title) -> void
-{
-    traceln("{}:", title);
-    traceln("{}", std::string(2 * w - 1, '='));
-    for (size_t y = 0; y < w; ++y) {
-        std::cout << "         ";
-        for (size_t x = 0; x < h; ++x) {
-            std::cout << static_cast<int>(get_bits(bitmap, y * w + x, y * h + x)) << " ";
-        }
-        std::cout << "\n";
-    }
-    std::cout << std::flush;
-    traceln("{}", std::string(2 * w - 1, '='));
+    std::cout << std::format("[{}ERROR{}]: ", ansi_bold_fg(fg::red), ansi_reset()) << std::format(
+        fmt,
+        std::forward<Args>(args)...) << std::endl;
 }
 
 #endif

shader/edge_fragment.glsl

@@ -0,0 +1,9 @@
+#version 330
+
+uniform vec4 colDiffuse;
+out vec4 finalColor;
+
+void main()
+{
+    finalColor = colDiffuse;
+}

shader/edge_vertex.glsl

@@ -0,0 +1,9 @@
+#version 330
+
+in vec3 vertexPosition;
+uniform mat4 mvp;
+
+void main()
+{
+    gl_Position = mvp * vec4(vertexPosition, 1.0);
+}

src/bits.cpp

@@ -0,0 +1,15 @@
+#include "bits.hpp"
+
+auto print_bitmap(const u64 bitmap, const u8 w, const u8 h, const std::string& title) -> void {
+    traceln("{}:", title);
+    traceln("{}", std::string(2 * w - 1, '='));
+    for (size_t y = 0; y < h; ++y) {
+        std::cout << "         ";
+        for (size_t x = 0; x < w; ++x) {
+            std::cout << static_cast<int>(get_bits(bitmap, y * w + x, y * w + x)) << " ";
+        }
+        std::cout << "\n";
+    }
+    std::cout << std::flush;
+    traceln("{}", std::string(2 * w - 1, '='));
+}

src/cpu_layout_engine.cpp

@@ -1,6 +1,6 @@
-#include "threaded_physics.hpp"
+#include "cpu_layout_engine.hpp"
 #include "config.hpp"
-#include "mass_spring_system.hpp"
+#include "cpu_spring_system.hpp"
 #include "util.hpp"
 
 #include <chrono>
@@ -10,16 +10,16 @@
 #include <vector>
 
 #ifdef ASYNC_OCTREE
-auto threaded_physics::set_octree_pool_thread_name(size_t idx) -> void
+auto cpu_layout_engine::set_octree_pool_thread_name(size_t idx) -> void
 {
     BS::this_thread::set_os_thread_name(std::format("octree-{}", idx));
     // traceln("Using thread \"{}\"", BS::this_thread::get_os_thread_name().value_or("INVALID NAME"));
 }
 #endif
 
-auto threaded_physics::physics_thread(physics_state& state, const std::optional<BS::thread_pool<>* const> thread_pool) -> void
+auto cpu_layout_engine::physics_thread(physics_state& state, const threadpool thread_pool) -> void
 {
-    mass_spring_system mass_springs;
+    cpu_spring_system mass_springs;
 
     #ifdef ASYNC_OCTREE
     BS::this_thread::set_os_thread_name("physics");
@@ -28,7 +28,6 @@ auto threaded_physics::physics_thread(physics_state& state, const std::optional<
     std::future<void> octree_future;
     octree tree_buffer;
     size_t last_mass_count = 0;
-    infoln("Using asynchronous octree builder.");
     #endif
 
     const auto visitor = overloads{
@@ -90,20 +89,20 @@ auto threaded_physics::physics_thread(physics_state& state, const std::optional<
 
             // Start building the octree for the next physics update.
             // Move the snapshot into the closure so it doesn't get captured by reference (don't use [&])
-            octree_future = octree_thread.submit_task([&tree_buffer, positions = std::move(positions)]()
+            octree_future = octree_thread.submit_task([&tree_buffer, &thread_pool, positions = std::move(positions)]()
             {
-                octree::build_octree(tree_buffer, positions);
+                octree::build_octree_morton(tree_buffer, positions, thread_pool);
             });
 
             // Rebuild the tree synchronously if we changed the number of masses to not use
             // an empty tree from the last frame in the frame where the graph was generated
             if (last_mass_count != mass_springs.positions.size()) {
                 traceln("Rebuilding octree synchronously because graph size changed");
-                octree::build_octree(mass_springs.tree, mass_springs.positions);
+                octree::build_octree_morton(mass_springs.tree, mass_springs.positions, thread_pool);
                 last_mass_count = mass_springs.positions.size();
             }
             #else
-            octree::build_octree(mass_springs.tree, mass_springs.positions);
+            octree::build_octree_morton(mass_springs.tree, mass_springs.positions, thread_pool);
             #endif
 
             mass_springs.clear_forces();
@@ -139,10 +138,11 @@ auto threaded_physics::physics_thread(physics_state& state, const std::optional<
             #else
             std::unique_lock<std::mutex> lock(state.data_mtx);
             #endif
-            state.data_consumed_cnd.wait(lock, [&]
-            {
-                return state.data_consumed || !state.running.load();
-            });
+            state.data_consumed_cnd.wait(lock,
+                                         [&]
+                                         {
+                                             return state.data_consumed || !state.running.load();
+                                         });
             if (!state.running.load()) {
                 // Running turned false while we were waiting for the condition
                 break;
@@ -154,10 +154,10 @@ auto threaded_physics::physics_thread(physics_state& state, const std::optional<
                 loop_iterations = 0;
                 ups_accumulator = std::chrono::duration<double>(0);
             }
-            if (mass_springs.tree.nodes.empty()) {
+            if (mass_springs.tree.empty()) {
                 state.mass_center = Vector3Zero();
             } else {
-                state.mass_center = mass_springs.tree.nodes[0].mass_center;
+                state.mass_center = mass_springs.tree.root().mass_center;
             }
 
             state.masses.clear();
@@ -176,12 +176,13 @@ auto threaded_physics::physics_thread(physics_state& state, const std::optional<
         state.data_ready_cnd.notify_all();
 
         #ifdef TRACY
-        FrameMarkEnd("PhysicsThreadProduceLock"); FrameMarkEnd("PhysicsThread");
+        FrameMarkEnd("PhysicsThreadProduceLock");
+        FrameMarkEnd("PhysicsThread");
         #endif
     }
 }
 
-auto threaded_physics::clear_cmd() -> void
+auto cpu_layout_engine::clear_cmd() -> void
 {
     {
         #ifdef TRACY
@@ -193,7 +194,7 @@ auto threaded_physics::clear_cmd() -> void
     }
 }
 
-auto threaded_physics::add_mass_cmd() -> void
+auto cpu_layout_engine::add_mass_cmd() -> void
 {
     {
         #ifdef TRACY
@@ -205,7 +206,7 @@ auto threaded_physics::add_mass_cmd() -> void
     }
 }
 
-auto threaded_physics::add_spring_cmd(const size_t a, const size_t b) -> void
+auto cpu_layout_engine::add_spring_cmd(const size_t a, const size_t b) -> void
 {
     {
         #ifdef TRACY
@@ -217,8 +218,7 @@ auto threaded_physics::add_spring_cmd(const size_t a, const size_t b) -> void
     }
 }
 
-auto threaded_physics::add_mass_springs_cmd(const size_t num_masses,
-                                            const std::vector<std::pair<size_t, size_t>>& springs) -> void
+auto cpu_layout_engine::add_mass_springs_cmd(const size_t num_masses, const std::vector<spring>& springs) -> void
 {
     {
         #ifdef TRACY

src/cpu_spring_system.cpp

@@ -1,69 +1,20 @@
-#include "mass_spring_system.hpp"
+#include "cpu_spring_system.hpp"
 #include "config.hpp"
 
 #include <cfloat>
 #include <cstring>
 
-auto mass_spring_system::calculate_spring_force(const size_t s) -> void
-{
-    const spring _s = springs[s];
-    const Vector3 a_pos = positions[_s.a];
-    const Vector3 b_pos = positions[_s.b];
-    const Vector3 a_vel = velocities[_s.a];
-    const Vector3 b_vel = velocities[_s.b];
-
-    const Vector3 delta_pos = a_pos - b_pos;
-    const Vector3 delta_vel = a_vel - b_vel;
-
-    const float sq_len = Vector3DotProduct(delta_pos, delta_pos);
-    const float inv_len = 1.0f / sqrt(sq_len);
-    const float len = sq_len * inv_len;
-
-    const float hooke = SPRING_CONSTANT * (len - REST_LENGTH);
-    const float dampening = DAMPENING_CONSTANT * Vector3DotProduct(delta_vel, delta_pos) * inv_len;
-
-    const Vector3 a_force = Vector3Scale(delta_pos, -(hooke + dampening) * inv_len);
-    const Vector3 b_force = a_force * -1.0f;
-
-    forces[_s.a] += a_force;
-    forces[_s.b] += b_force;
-}
-
-auto mass_spring_system::integrate_velocity(const size_t m, const float dt) -> void
-{
-    const Vector3 acc = forces[m] / MASS;
-    velocities[m] += acc * dt;
-}
-
-auto mass_spring_system::integrate_position(const size_t m, const float dt) -> void
-{
-    previous_positions[m] = positions[m];
-    positions[m] += velocities[m] * dt;
-}
-
-auto mass_spring_system::verlet_update(const size_t m, const float dt) -> void
-{
-    const Vector3 acc = (forces[m] / MASS) * dt * dt;
-    const Vector3 pos = positions[m];
-
-    Vector3 delta_pos = pos - previous_positions[m];
-    delta_pos *= 1.0 - VERLET_DAMPENING; // Minimal dampening
-
-    positions[m] += delta_pos + acc;
-    previous_positions[m] = pos;
-}
-
-auto mass_spring_system::clear() -> void
+auto cpu_spring_system::clear() -> void
 {
     positions.clear();
     previous_positions.clear();
     velocities.clear();
     forces.clear();
     springs.clear();
-    tree.nodes.clear();
+    tree.clear();
 }
 
-auto mass_spring_system::add_mass() -> void
+auto cpu_spring_system::add_mass() -> void
 {
     // Adding all positions to (0, 0, 0) breaks the octree
 
@@ -80,21 +31,24 @@ auto mass_spring_system::add_mass() -> void
     forces.emplace_back(Vector3Zero());
 }
 
-auto mass_spring_system::add_spring(size_t a, size_t b) -> void
+auto cpu_spring_system::add_spring(size_t a, size_t b) -> void
 {
     // Update masses to be located along a random walk when adding the springs
     const Vector3& mass_a = positions[a];
     const Vector3& mass_b = positions[b];
 
-    Vector3 offset{static_cast<float>(GetRandomValue(-100, 100)),
-                   static_cast<float>(GetRandomValue(-100, 100)),
-                   static_cast<float>(GetRandomValue(-100, 100))};
-    offset = Vector3Normalize(offset) * REST_LENGTH;
+    Vector3 offset{
+        static_cast<float>(GetRandomValue(-100, 100)),
+        static_cast<float>(GetRandomValue(-100, 100)),
+        static_cast<float>(GetRandomValue(-100, 100))
+    };
+
+    // By spawning the masses close together, we "explode" them naturally, so they cluster faster (also looks cool)
+    offset = Vector3Normalize(offset) * REST_LENGTH * 0.1;
 
     // If the offset moves the mass closer to the current center of mass, flip it
-    if (!tree.nodes.empty()) {
-        const Vector3 mass_center_direction =
-            Vector3Subtract(positions[a], tree.nodes[0].mass_center);
+    if (!tree.empty()) {
+        const Vector3 mass_center_direction = Vector3Subtract(positions[a], tree.root().mass_center);
         const float mass_center_distance = Vector3Length(mass_center_direction);
 
         if (mass_center_distance > 0 && Vector3DotProduct(offset, mass_center_direction) < 0.0f) {
@@ -112,28 +66,53 @@ auto mass_spring_system::add_spring(size_t a, size_t b) -> void
     springs.emplace_back(a, b);
 }
 
-auto mass_spring_system::clear_forces() -> void
+auto cpu_spring_system::clear_forces() -> void
 {
-#ifdef TRACY
+    #ifdef TRACY
     ZoneScoped;
-#endif
+    #endif
 
     memset(forces.data(), 0, forces.size() * sizeof(Vector3));
 }
 
-auto mass_spring_system::calculate_spring_forces(
-    const std::optional<BS::thread_pool<>* const> thread_pool) -> void
+auto cpu_spring_system::calculate_spring_force(const size_t s) -> void
 {
-#ifdef TRACY
-    ZoneScoped;
-#endif
+    const spring _s = springs[s];
+    const Vector3 a_pos = positions[_s.first];
+    const Vector3 b_pos = positions[_s.second];
+    const Vector3 a_vel = velocities[_s.first];
+    const Vector3 b_vel = velocities[_s.second];
 
-    const auto solve_spring_force = [&](const int i) { calculate_spring_force(i); };
+    const Vector3 delta_pos = a_pos - b_pos;
+    const Vector3 delta_vel = a_vel - b_vel;
+
+    const float sq_len = Vector3DotProduct(delta_pos, delta_pos);
+    const float inv_len = 1.0f / sqrt(sq_len);
+    const float len = sq_len * inv_len;
+
+    const float hooke = SPRING_K * (len - REST_LENGTH);
+    const float dampening = DAMPENING_K * Vector3DotProduct(delta_vel, delta_pos) * inv_len;
+
+    const Vector3 a_force = Vector3Scale(delta_pos, -(hooke + dampening) * inv_len);
+    const Vector3 b_force = a_force * -1.0f;
+
+    forces[_s.first] += a_force;
+    forces[_s.second] += b_force;
+}
+
+auto cpu_spring_system::calculate_spring_forces(const threadpool thread_pool) -> void
+{
+    #ifdef TRACY
+    ZoneScoped;
+    #endif
+
+    const auto solve_spring_force = [&](const int i)
+    {
+        calculate_spring_force(i);
+    };
 
     if (thread_pool) {
-        (*thread_pool)
-            ->submit_loop(0, springs.size(), solve_spring_force, SMALL_TASK_BLOCK_SIZE)
-            .wait();
+        (*thread_pool)->submit_loop(0, springs.size(), solve_spring_force, SMALL_TASK_BLOCK_SIZE).wait();
     } else {
         for (size_t i = 0; i < springs.size(); ++i) {
             solve_spring_force(i);
@@ -141,24 +120,21 @@ auto mass_spring_system::calculate_spring_forces(
     }
 }
 
-auto mass_spring_system::calculate_repulsion_forces(
-    const std::optional<BS::thread_pool<>* const> thread_pool) -> void
+auto cpu_spring_system::calculate_repulsion_forces(const threadpool thread_pool) -> void
 {
-#ifdef TRACY
+    #ifdef TRACY
     ZoneScoped;
-#endif
+    #endif
 
     const auto solve_octree = [&](const int i)
     {
-        const Vector3 force = tree.calculate_force(0, positions[i]);
+        const Vector3 force = tree.calculate_force_morton(0, positions[i], i);
         forces[i] += force;
     };
 
     // Calculate forces using Barnes-Hut
     if (thread_pool) {
-        (*thread_pool)
-            ->submit_loop(0, positions.size(), solve_octree, LARGE_TASK_BLOCK_SIZE)
-            .wait();
+        (*thread_pool)->submit_loop(0, positions.size(), solve_octree, LARGE_TASK_BLOCK_SIZE).wait();
     } else {
         for (size_t i = 0; i < positions.size(); ++i) {
             solve_octree(i);
@@ -166,14 +142,40 @@ auto mass_spring_system::calculate_repulsion_forces(
     }
 }
 
-auto mass_spring_system::update(const float dt,
-                                const std::optional<BS::thread_pool<>* const> thread_pool) -> void
+auto cpu_spring_system::integrate_velocity(const size_t m, const float dt) -> void
 {
-#ifdef TRACY
-    ZoneScoped;
-#endif
+    const Vector3 acc = forces[m] / MASS;
+    velocities[m] += acc * dt;
+}
 
-    const auto update = [&](const int i) { verlet_update(i, dt); };
+auto cpu_spring_system::integrate_position(const size_t m, const float dt) -> void
+{
+    previous_positions[m] = positions[m];
+    positions[m] += velocities[m] * dt;
+}
+
+auto cpu_spring_system::verlet_update(const size_t m, const float dt) -> void
+{
+    const Vector3 acc = (forces[m] / MASS) * dt * dt;
+    const Vector3 pos = positions[m];
+
+    Vector3 delta_pos = pos - previous_positions[m];
+    delta_pos *= 1.0 - VERLET_DAMPENING; // Minimal dampening
+
+    positions[m] += delta_pos + acc;
+    previous_positions[m] = pos;
+}
+
+auto cpu_spring_system::update(const float dt, const threadpool thread_pool) -> void
+{
+    #ifdef TRACY
+    ZoneScoped;
+    #endif
+
+    const auto update = [&](const int i)
+    {
+        verlet_update(i, dt);
+    };
 
     if (thread_pool) {
         (*thread_pool)->submit_loop(0, positions.size(), update, SMALL_TASK_BLOCK_SIZE).wait();
@@ -184,8 +186,7 @@ auto mass_spring_system::update(const float dt,
     }
 }
 
-auto mass_spring_system::center_masses(const std::optional<BS::thread_pool<>* const> thread_pool)
-    -> void
+auto cpu_spring_system::center_masses(const threadpool thread_pool) -> void
 {
     Vector3 mean = Vector3Zero();
     for (const Vector3& pos : positions) {
@@ -193,7 +194,10 @@ auto mass_spring_system::center_masses(const std::optional<BS::thread_pool<>* co
     }
     mean /= static_cast<float>(positions.size());
 
-    const auto center_mass = [&](const int i) { positions[i] -= mean; };
+    const auto center_mass = [&](const int i)
+    {
+        positions[i] -= mean;
+    };
 
     if (thread_pool) {
         (*thread_pool)->submit_loop(0, positions.size(), center_mass, SMALL_TASK_BLOCK_SIZE).wait();
@@ -202,4 +206,4 @@ auto mass_spring_system::center_masses(const std::optional<BS::thread_pool<>* co
             center_mass(i);
         }
     }
-}
+}

src/graph_distances.cpp

@@ -15,7 +15,7 @@ auto graph_distances::empty() const -> bool
 }
 
 auto graph_distances::calculate_distances(const size_t node_count,
-                                          const std::vector<std::pair<size_t, size_t>>& edges,
+                                          const std::vector<spring>& edges,
                                           const std::vector<size_t>& targets) -> void
 {
     // Build a list of adjacent nodes to speed up BFS

src/input_handler.cpp

@@ -18,6 +18,7 @@ auto input_handler::init_handlers() -> void
     register_mouse_pressed_handler(MOUSE_BUTTON_LEFT, &input_handler::add_block);
     register_mouse_pressed_handler(MOUSE_BUTTON_LEFT, &input_handler::start_add_block);
     register_mouse_pressed_handler(MOUSE_BUTTON_MIDDLE, &input_handler::place_goal);
+    register_mouse_pressed_handler(MOUSE_BUTTON_MIDDLE, &input_handler::select_state);
     register_mouse_pressed_handler(MOUSE_BUTTON_RIGHT, &input_handler::camera_start_rotate);
     register_mouse_pressed_handler(MOUSE_BUTTON_RIGHT, &input_handler::remove_block);
     register_mouse_pressed_handler(MOUSE_BUTTON_RIGHT, &input_handler::clear_add_block);
@@ -42,6 +43,7 @@ auto input_handler::init_handlers() -> void
     register_key_pressed_handler(KEY_C, &input_handler::clear_graph);
     register_key_pressed_handler(KEY_I, &input_handler::toggle_mark_solutions);
     register_key_pressed_handler(KEY_O, &input_handler::toggle_connect_solutions);
+    register_key_pressed_handler(KEY_Z, &input_handler::toggle_color_by_distance);
 
     register_key_pressed_handler(KEY_TAB, &input_handler::toggle_editing);
     register_key_pressed_handler(KEY_F, &input_handler::toggle_restricted_movement);
@@ -142,12 +144,17 @@ auto input_handler::camera_zoom() const -> void
 
 auto input_handler::camera_fov() const -> void
 {
-    if (!mouse_in_graph_pane() || !IsKeyDown(KEY_LEFT_CONTROL) || IsKeyDown(KEY_LEFT_SHIFT)) {
+    if (!mouse_in_graph_pane() || !IsKeyDown(KEY_LEFT_CONTROL)) {
         return;
     }
 
     const float wheel = GetMouseWheelMove();
-    camera.fov -= wheel * FOV_SPEED;
+
+    if (IsKeyDown(KEY_LEFT_SHIFT)) {
+        camera.fov -= wheel * FOV_SPEED * FOV_MULTIPLIER;
+    } else {
+        camera.fov -= wheel * FOV_SPEED;
+    }
 }
 
 auto input_handler::select_block() -> void
@@ -201,7 +208,7 @@ auto input_handler::add_block() -> void
         has_block_add_xy = false;
     } else if (current.covers(block_add_x, block_add_y, block_add_width, block_add_height)) {
         const std::optional<puzzle>& next = current.try_add_block(
-            puzzle::block(block_add_x, block_add_y, block_add_width, block_add_height, false));
+            block(block_add_x, block_add_y, block_add_width, block_add_height, false));
 
         if (next) {
             sel_x = block_add_x;
@@ -217,7 +224,7 @@ auto input_handler::add_block() -> void
 auto input_handler::remove_block() -> void
 {
     const puzzle& current = state.get_current_state();
-    const std::optional<puzzle::block>& b = current.try_get_block(hov_x, hov_y);
+    const std::optional<block>& b = current.try_get_block(hov_x, hov_y);
     if (!editing || has_block_add_xy || !b) {
         return;
     }
@@ -239,7 +246,7 @@ auto input_handler::remove_block() -> void
 auto input_handler::place_goal() const -> void
 {
     const puzzle& current = state.get_current_state();
-    if (!editing || !current.covers(hov_x, hov_y)) {
+    if (!editing || !mouse_in_board_pane() || !current.covers(hov_x, hov_y)) {
         return;
     }
 
@@ -251,6 +258,16 @@ auto input_handler::place_goal() const -> void
     state.edit_starting_state(*next);
 }
 
+auto input_handler::select_state() const -> void
+{
+    if (!mouse_in_graph_pane() || collision_mass == static_cast<size_t>(-1)) {
+        return;
+    }
+
+    const puzzle& selected = state.get_state(collision_mass);
+    state.update_current_state(selected);
+}
+
 auto input_handler::toggle_camera_lock() -> void
 {
     if (!camera_lock) {
@@ -407,6 +424,11 @@ auto input_handler::toggle_connect_solutions() -> void
     connect_solutions = !connect_solutions;
 }
 
+auto input_handler::toggle_color_by_distance() -> void
+{
+    color_by_distance = !color_by_distance;
+}
+
 auto input_handler::toggle_mark_path() -> void
 {
     mark_path = !mark_path;

src/load_save.cpp

@@ -62,7 +62,10 @@ auto append_preset_file(const std::string& preset_file, const std::string& prese
     return true;
 }
 
-auto append_preset_file_quiet(const std::string& preset_file, const std::string& preset_name, const puzzle& p, const bool validate) -> bool
+auto append_preset_file_quiet(const std::string& preset_file,
+                              const std::string& preset_name,
+                              const puzzle& p,
+                              const bool validate) -> bool
 {
     if (validate && p.try_get_invalid_reason()) {
         return false;

src/main.cpp

@@ -1,26 +1,50 @@
-#include <chrono>
-#include <mutex>
-#include <raylib.h>
-
 #include "config.hpp"
 #include "input_handler.hpp"
-#include "threaded_physics.hpp"
+#include "cpu_layout_engine.hpp"
 #include "renderer.hpp"
 #include "state_manager.hpp"
 #include "user_interface.hpp"
 
+#include <chrono>
+#include <mutex>
+#include <GL/glew.h>
+#include <raylib.h>
 #include <filesystem>
 
-#ifndef WIN32
+#if not defined(_WIN32)
 #include <boost/program_options.hpp>
 namespace po = boost::program_options;
 #endif
 
+// Threadpool setup
+#ifdef THREADPOOL
+auto set_pool_thread_name(size_t idx) -> void
+{
+    BS::this_thread::set_os_thread_name(std::format("worker-{}", idx));
+}
+
+BS::thread_pool<> threads(std::thread::hardware_concurrency() - 2, set_pool_thread_name);
+constexpr threadpool thread_pool = &threads;
+#else
+constexpr threadpool thread_pool = std::nullopt;
+#endif
+
+// Argparse defaults
+std::string preset_file = "default.puzzle";
+std::string output_file = "clusters.puzzle";
+int max_blocks = 5;
+int min_moves = 10;
+
+// Puzzle space setup
+int board_width;
+int board_height;
+int goal_x;
+int goal_y;
+bool restricted;
+blockset2 permitted_blocks;
+block target_block;
+std::tuple<u8, u8, u8, u8> target_block_pos_range;
 
-// TODO: Implement state discovery/enumeration
-//       - Find all possible initial board states (single one for each possible statespace).
-//         Currently wer're just finding all states given the initial state
-//       - Would allow to generate random puzzles with a certain move count
 // TODO: Export cluster to graphviz
 // TODO: Fix naming:
 //       - Target: The block that has to leave the board to win
@@ -33,29 +57,6 @@ namespace po = boost::program_options;
 // TODO: Add state space generation time to debug overlay
 // TODO: Move selection accordingly when undoing moves (need to diff two states and get the moved blocks)
 
-// TODO: Click states in the graph to display them in the board
-
-#ifdef THREADPOOL
-auto set_pool_thread_name(size_t idx) -> void
-{
-    BS::this_thread::set_os_thread_name(std::format("worker-{}", idx));
-}
-
-BS::thread_pool<> threads(std::thread::hardware_concurrency() - 2, set_pool_thread_name);
-constexpr std::optional<BS::thread_pool<>* const> thread_pool = &threads;
-#else
-constexpr std::optional<BS::thread_pool<>* const> thread_pool = std::nullopt;
-#endif
-
-std::string preset_file;
-std::string output_file;
-int max_blocks = 5;
-int board_width = 6;
-int board_height = 6;
-int goal_x = 4;
-int goal_y = 2;
-bool restricted = true;
-
 auto ui_mode() -> int
 {
     // RayLib window setup
@@ -66,8 +67,15 @@ auto ui_mode() -> int
     SetConfigFlags(FLAG_WINDOW_ALWAYS_RUN);
     InitWindow(INITIAL_WIDTH * 2, INITIAL_HEIGHT + MENU_HEIGHT, "MassSprings");
 
+    // GLEW setup
+    glewExperimental = GL_TRUE;
+    const GLenum glew_err = glewInit();
+    if (glew_err != GLEW_OK) {
+        TraceLog(LOG_FATAL, "Failed to initialize GLEW: %s", glewGetErrorString(glew_err));
+    }
+
     // Game setup
-    threaded_physics physics(thread_pool);
+    cpu_layout_engine physics(thread_pool);
     state_manager state(physics, preset_file);
     orbit_camera camera;
     input_handler input(state, camera);
@@ -161,51 +169,87 @@ auto ui_mode() -> int
     return 0;
 }
 
-auto rush_hour_puzzle_space() -> int
+auto rush_hour_puzzle_space() -> void
 {
-    const boost::unordered_flat_set<puzzle::block, block_hasher2, block_equal2> permitted_blocks = {
-        puzzle::block(0, 0, 2, 1, false, false),
-        puzzle::block(0, 0, 3, 1, false, false),
-        puzzle::block(0, 0, 1, 2, false, false),
-        puzzle::block(0, 0, 1, 3, false, false)
+    board_width = 6;
+    board_height = 6;
+    goal_x = 4;
+    goal_y = 2;
+    restricted = true;
+    permitted_blocks = {
+        block(0, 0, 2, 1, false, false),
+        block(0, 0, 3, 1, false, false),
+        block(0, 0, 1, 2, false, false),
+        block(0, 0, 1, 3, false, false)
     };
-    const puzzle::block target_block = puzzle::block(0, 0, 2, 1, true, false);
-    const std::tuple<uint8_t, uint8_t, uint8_t, uint8_t> target_block_pos_range = {0, goal_y, board_width - 1, goal_y};
+    target_block = block(0, 0, 2, 1, true, false);
+    target_block_pos_range = {0, goal_y, board_width - target_block.get_width(), goal_y};
+}
 
-    infoln("Exploring Rush-Hour puzzle space:");
+auto klotski_puzzle_space() -> void
+{
+    board_width = 4;
+    board_height = 5;
+    goal_x = 1;
+    goal_y = 3;
+    restricted = false;
+    permitted_blocks = {
+        block(0, 0, 1, 1, false, false),
+        block(0, 0, 1, 2, false, false),
+        block(0, 0, 2, 1, false, false),
+    };
+    target_block = block(0, 0, 2, 2, true, false);
+    target_block_pos_range = {
+        0,
+        0,
+        board_width - target_block.get_width(),
+        board_height - target_block.get_height(),
+    };
+}
+
+auto puzzle_space() -> int
+{
+    // We don't only pick max_blocks out of n (with duplicates), but also 1 out of n, 2, 3, ... max_blocks-1 out of n
+    int upper_set_count = 0;
+    for (int i = 1; i <= max_blocks; ++i) {
+        upper_set_count += binom(permitted_blocks.size() + i - 1, i);
+    }
+
+    infoln("Exploring puzzle space:");
     infoln("- Size:       {}x{}", board_width, board_height);
     infoln("- Goal:       {},{}", goal_x, goal_y);
     infoln("- Restricted: {}", restricted);
     infoln("- Max Blocks: {}", max_blocks);
+    infoln("- Min Moves:  {}", min_moves);
     infoln("- Target:     {}x{}", target_block.get_width(), target_block.get_height());
+    infoln("- Max Sets:   {}", upper_set_count);
     infoln("- Permitted block sizes:");
     std::cout << "         ";
-    for (const puzzle::block b : permitted_blocks) {
+    for (const block b : permitted_blocks) {
         std::cout << std::format(" {}x{},", b.get_width(), b.get_height());
     }
     std::cout << std::endl;
-    const std::chrono::high_resolution_clock::time_point start = std::chrono::high_resolution_clock::now();
 
     const puzzle p = puzzle(board_width, board_height, goal_x, goal_y, restricted, true);
-    const boost::unordered_flat_set<puzzle, puzzle_hasher> result = p.explore_puzzle_space(
+
+    STARTTIME;
+    const puzzleset result = p.explore_puzzle_space(
         permitted_blocks,
         target_block,
         target_block_pos_range,
         max_blocks,
+        min_moves,
         thread_pool);
+    ENDTIME(std::format("Found {} different clusters", result.size()), std::chrono::seconds, "s");
 
-    const std::chrono::high_resolution_clock::time_point end = std::chrono::high_resolution_clock::now();
-    infoln("Found {} different clusters. Took {}s.",
-           result.size(),
-           std::chrono::duration_cast<std::chrono::seconds>(end - start).count());
+    // TODO: The exported clusters are the numerically smallest state of the cluster.
+    //       Not the state with the longest path.
 
     infoln("Sorting clusters...");
     std::vector<puzzle> result_sorted{result.begin(), result.end()};
     std::ranges::sort(result_sorted, std::ranges::greater{});
-    // for (const puzzle& _p : result_sorted) {
-    //     traceln("{}", _p.string_repr());
-    // }
 
+    infoln("Saving clusters...");
     size_t i = 0;
     size_t success = 0;
     std::filesystem::remove(output_file);
@@ -226,26 +270,32 @@ auto rush_hour_puzzle_space() -> int
 
 enum class runmode
 {
-    USER_INTERFACE, RUSH_HOUR_PUZZLE_SPACE, EXIT,
+    USER_INTERFACE,
+    RUSH_HOUR_PUZZLE_SPACE,
+    KLOTSKI_PUZZLE_SPACE,
+    EXIT,
 };
 
 auto argparse(const int argc, char* argv[]) -> runmode
 {
-#ifndef WIN32
+    #if not defined(_WIN32)
     po::options_description desc("Allowed options");
-    desc.add_options()                   //
-        ("help", "produce help message") //
-
-        ("presets", po::value<std::string>()->default_value("default.puzzle"), "load presets from file") //
-
-        ("output", po::value<std::string>()->default_value("clusters.puzzle"), "output file for generated clusters") //
-        ("space", po::value<std::string>()->value_name("rh|klotski"), "generate puzzle space with ruleset")          //
-        ("w", po::value<int>()->default_value(6), "board width")                                                     //
-        ("h", po::value<int>()->default_value(6), "board height")                                                    //
-        ("gx", po::value<int>()->default_value(4), "board goal horizontal position")                                 //
-        ("gy", po::value<int>()->default_value(2), "board goal vertical position")                                   //
-        ("free", "allow free block movement")                                                                        //
-        ("blocks", po::value<int>()->default_value(5), "block limit for puzzle space generation")                    //
+    desc.add_options()                                                                                         //
+        ("help", "produce help message")                                                                       //
+        ("presets", po::value<std::string>()->default_value(preset_file), "load presets from file")            //
+        ("output", po::value<std::string>()->default_value(output_file), "output file for generated clusters") //
+        ("space", po::value<std::string>()->value_name("rh|klotski"), "generate puzzle space with ruleset")    //
+        // ("w", po::value<int>()->default_value(board_width)->value_name("[3, 8]"), "board width")                  //
+        // ("h", po::value<int>()->default_value(board_height)->value_name("[3, 8"), "board height")                 //
+        // ("gx", po::value<int>()->default_value(goal_x)->value_name("[0, w-1]"), "board goal horizontal position") //
+        // ("gy", po::value<int>()->default_value(goal_y)->value_name("[0, h-1]"), "board goal vertical position")   //
+        // ("free", "allow free block movement")                                                                     //
+        ("blocks",
+         po::value<int>()->default_value(max_blocks)->value_name("[1, 15]"),
+         "block limit for puzzle space generation") //
+        ("moves",
+         po::value<int>()->default_value(min_moves),
+         "only save puzzles with at least this many required moves") //
         ;
 
     po::positional_options_description positional;
@@ -259,56 +309,53 @@ auto argparse(const int argc, char* argv[]) -> runmode
         std::cout << desc << std::endl;
         return runmode::EXIT;
     }
-
     if (vm.contains("output")) {
         output_file = vm["output"].as<std::string>();
     }
 
-    if (vm.contains("w")) {
-        board_width = vm["w"].as<int>();
-        board_width = std::max(static_cast<int>(puzzle::MIN_WIDTH),
-                               std::min(board_width, static_cast<int>(puzzle::MAX_WIDTH)));
-    }
-
-    if (vm.contains("h")) {
-        board_height = vm["h"].as<int>();
-        board_height = std::max(static_cast<int>(puzzle::MIN_HEIGHT),
-                                std::min(board_height, static_cast<int>(puzzle::MAX_HEIGHT)));
-    }
-
-    if (vm.contains("gx")) {
-        goal_x = vm["gx"].as<int>();
-        goal_x = std::max(0, std::min(goal_x, static_cast<int>(puzzle::MAX_WIDTH) - 1));
-    }
-
-    if (vm.contains("gy")) {
-        goal_y = vm["gy"].as<int>();
-        goal_y = std::max(0, std::min(goal_y, static_cast<int>(puzzle::MAX_HEIGHT) - 1));
-    }
-
-    if (vm.contains("free")) {
-        restricted = false;
-    }
+    // if (vm.contains("w")) {
+    //     board_width = vm["w"].as<int>();
+    //     board_width = std::max(static_cast<int>(puzzle::MIN_WIDTH),
+    //                            std::min(board_width, static_cast<int>(puzzle::MAX_WIDTH)));
+    // }
+    // if (vm.contains("h")) {
+    //     board_height = vm["h"].as<int>();
+    //     board_height = std::max(static_cast<int>(puzzle::MIN_HEIGHT),
+    //                             std::min(board_height, static_cast<int>(puzzle::MAX_HEIGHT)));
+    // }
+    // if (vm.contains("gx")) {
+    //     goal_x = vm["gx"].as<int>();
+    //     goal_x = std::max(0, std::min(goal_x, static_cast<int>(puzzle::MAX_WIDTH) - 1));
+    // }
+    // if (vm.contains("gy")) {
+    //     goal_y = vm["gy"].as<int>();
+    //     goal_y = std::max(0, std::min(goal_y, static_cast<int>(puzzle::MAX_HEIGHT) - 1));
+    // }
+    // if (vm.contains("free")) {
+    //     restricted = false;
+    // }
 
     if (vm.contains("blocks")) {
         max_blocks = vm["blocks"].as<int>();
         max_blocks = std::max(1, std::min(max_blocks, static_cast<int>(puzzle::MAX_BLOCKS)));
     }
-
+    if (vm.contains("moves")) {
+        min_moves = vm["moves"].as<int>();
+        min_moves = std::max(0, min_moves);
+    }
     if (vm.contains("space")) {
         const std::string ruleset = vm["space"].as<std::string>();
         if (ruleset == "rh") {
             return runmode::RUSH_HOUR_PUZZLE_SPACE;
         }
         if (ruleset == "klotski") {
-            throw std::runtime_error("Not implemented");
+            return runmode::KLOTSKI_PUZZLE_SPACE;
         }
     }
-
     if (vm.contains("presets")) {
         preset_file = vm["presets"].as<std::string>();
     }
-#endif
+    #endif
 
     return runmode::USER_INTERFACE;
 }
@@ -328,7 +375,13 @@ auto main(const int argc, char* argv[]) -> int
     #endif
 
     infoln("Using background thread for physics.");
-    infoln("Using octree-barnes-hut for graph layout.");
+    infoln("Using linear octree + Barnes-Hut for graph layout.");
+
+    #ifdef ASYNC_OCTREE
+    infoln("Using asynchronous octree build.");
+    #else
+    infoln("Using synchronous octree build.");
+    #endif
 
     #ifdef THREADPOOL
     infoln("Additional thread-pool enabled ({} threads).", threads.get_thread_count());
@@ -340,10 +393,14 @@ auto main(const int argc, char* argv[]) -> int
     case runmode::USER_INTERFACE:
         return ui_mode();
     case runmode::RUSH_HOUR_PUZZLE_SPACE:
-        return rush_hour_puzzle_space();
+        rush_hour_puzzle_space();
+        return puzzle_space();
+    case runmode::KLOTSKI_PUZZLE_SPACE:
+        klotski_puzzle_space();
+        return puzzle_space();
     case runmode::EXIT:
         return 0;
     };
 
     return 1;
-}
+}

src/octree.cpp

@@ -4,224 +4,332 @@
 #include <cfloat>
 #include <raymath.h>
 
-auto octree::node::child_count() const -> int
+auto octree::clear() -> void
 {
-    int child_count = 0;
-    for (const int child : children) {
-        if (child != -1) {
-            ++child_count;
-        }
-    }
-    return child_count;
+    nodes.clear();
 }
 
-auto octree::get_octant(const int node_idx, const Vector3& pos) const -> int
+auto octree::reserve(const size_t count) -> void
 {
-    const node& n = nodes[node_idx];
-    auto [cx, cy, cz] = Vector3((n.box_min.x + n.box_max.x) / 2.0f, (n.box_min.y + n.box_max.y) / 2.0f,
-                                (n.box_min.z + n.box_max.z) / 2.0f);
-
-    // The octant is encoded as a 3-bit integer "zyx". The node area is split
-    // along all 3 axes, if a position is right of an axis, this bit is set to 1.
-    // If a position is right of the x-axis and y-axis and left of the z-axis, the
-    // encoded octant is "011".
-    int octant = 0;
-    if (pos.x >= cx) {
-        octant |= 1;
-    }
-    if (pos.y >= cy) {
-        octant |= 2;
-    }
-    if (pos.z >= cz) {
-        octant |= 4;
-    }
-
-    return octant;
+    nodes.reserve(count);
 }
 
-auto octree::get_child_bounds(const int node_idx, const int octant) const -> std::pair<Vector3, Vector3>
+auto octree::empty() const -> bool
 {
-    const node& n = nodes[node_idx];
-    auto [cx, cy, cz] = Vector3((n.box_min.x + n.box_max.x) / 2.0f, (n.box_min.y + n.box_max.y) / 2.0f,
-                                (n.box_min.z + n.box_max.z) / 2.0f);
-
-    Vector3 min = Vector3Zero();
-    Vector3 max = Vector3Zero();
-
-    // If (octant & 1), the octant is to the right of the node region's x-axis
-    // (see GetOctant). This means the left bound is the x-axis and the right
-    // bound the node's region max.
-    min.x = octant & 1 ? cx : n.box_min.x;
-    max.x = octant & 1 ? n.box_max.x : cx;
-    min.y = octant & 2 ? cy : n.box_min.y;
-    max.y = octant & 2 ? n.box_max.y : cy;
-    min.z = octant & 4 ? cz : n.box_min.z;
-    max.z = octant & 4 ? n.box_max.z : cz;
-
-    return std::make_pair(min, max);
+    return nodes.empty();
 }
 
-auto octree::create_empty_leaf(const Vector3& box_min, const Vector3& box_max) -> int
+auto octree::root() const -> const node&
 {
-    node n;
-    n.box_min = box_min;
-    n.box_max = box_max;
-    nodes.emplace_back(n);
-
-    return static_cast<int>(nodes.size() - 1);
+    return nodes[0];
 }
 
-auto octree::insert(const int node_idx, const int mass_id, const Vector3& pos, const float mass,
-                    const int depth) -> void
-{
-    // infoln("Inserting position ({}, {}, {}) into octree at node {} (depth {})", pos.x, pos.y,
-    // pos.z, node_idx, depth);
-    if (depth > MAX_DEPTH) {
-        throw std::runtime_error(std::format("MAX_DEPTH! node={} box_min=({},{},{}) box_max=({},{},{}) pos=({},{},{})",
-                                             node_idx, nodes[node_idx].box_min.x, nodes[node_idx].box_min.y,
-                                             nodes[node_idx].box_min.z, nodes[node_idx].box_max.x,
-                                             nodes[node_idx].box_max.y, nodes[node_idx].box_max.z, pos.x, pos.y,
-                                             pos.z));
-    }
-
-    // NOTE: Do not store a nodes[node_idx] reference as the nodes vector might reallocate during
-    // this function
-
-    // We can place the particle in the empty leaf
-    if (nodes[node_idx].leaf && nodes[node_idx].mass_id == -1) {
-        nodes[node_idx].mass_id = mass_id;
-        nodes[node_idx].mass_center = pos;
-        nodes[node_idx].mass_total = mass;
-        return;
-    }
-
-    // The leaf is occupied, we need to subdivide
-    if (nodes[node_idx].leaf) {
-        const int existing_id = nodes[node_idx].mass_id;
-        const Vector3 existing_pos = nodes[node_idx].mass_center;
-        const float existing_mass = nodes[node_idx].mass_total;
-
-        // If positions are identical we jitter the particles
-        const Vector3 diff = Vector3Subtract(pos, existing_pos);
-        if (diff == Vector3Zero()) {
-            // warnln("Trying to insert an identical partical into octree (jittering position)");
-
-            Vector3 jittered = pos;
-            jittered.x += 0.001;
-            jittered.y += 0.001;
-            insert(node_idx, mass_id, jittered, mass, depth);
-            return;
-
-            // Could also merge them, but that leads to the octree having less leafs than we have
-            // masses nodes[node_idx].mass_total += mass; return;
-        }
-
-        // Convert the leaf to an internal node
-        nodes[node_idx].mass_id = -1;
-        nodes[node_idx].leaf = false;
-        nodes[node_idx].mass_total = 0.0;
-        nodes[node_idx].mass_center = Vector3Zero();
-
-        // Re-insert the existing mass into a new empty leaf (see above)
-        const int oct = get_octant(node_idx, existing_pos);
-        if (nodes[node_idx].children[oct] == -1) {
-            const auto& [min, max] = get_child_bounds(node_idx, oct);
-            const int child_idx = create_empty_leaf(min, max);
-            nodes[node_idx].children[oct] = child_idx;
-        }
-        insert(nodes[node_idx].children[oct], existing_id, existing_pos, existing_mass, depth + 1);
-    }
-
-    // Insert the new mass
-    const int oct = get_octant(node_idx, pos);
-    if (nodes[node_idx].children[oct] == -1) {
-        const auto& [min, max] = get_child_bounds(node_idx, oct);
-        const int child_idx = create_empty_leaf(min, max);
-        nodes[node_idx].children[oct] = child_idx;
-    }
-    insert(nodes[node_idx].children[oct], mass_id, pos, mass, depth + 1);
-
-    // Update the center of mass
-    const float new_mass = nodes[node_idx].mass_total + mass;
-    nodes[node_idx].mass_center.x = (nodes[node_idx].mass_center.x * nodes[node_idx].mass_total + pos.x) / new_mass;
-    nodes[node_idx].mass_center.y = (nodes[node_idx].mass_center.y * nodes[node_idx].mass_total + pos.y) / new_mass;
-    nodes[node_idx].mass_center.z = (nodes[node_idx].mass_center.z * nodes[node_idx].mass_total + pos.z) / new_mass;
-    nodes[node_idx].mass_total = new_mass;
-}
-
-auto octree::build_octree(octree& t, const std::vector<Vector3>& positions) -> void
+// Replaced the 50 line recursive octree insertion with this morton bitch to gain 5 UPS, FML
+auto octree::build_octree_morton(octree& t,
+                                 const std::vector<Vector3>& positions,
+                                 const std::optional<BS::thread_pool<>*>& thread_pool) -> void
 {
     #ifdef TRACY
     ZoneScoped;
     #endif
 
-    t.nodes.clear();
-    t.nodes.reserve(positions.size() * 2);
+    t.clear();
+    if (positions.empty()) {
+        return;
+    }
 
     // Compute bounding box around all masses
-    Vector3 min{FLT_MAX, FLT_MAX, FLT_MAX};
-    Vector3 max{-FLT_MAX, -FLT_MAX, -FLT_MAX};
+    Vector3 root_min{FLT_MAX, FLT_MAX, FLT_MAX};
+    Vector3 root_max{-FLT_MAX, -FLT_MAX, -FLT_MAX};
     for (const auto& [x, y, z] : positions) {
-        min.x = std::min(min.x, x);
-        max.x = std::max(max.x, x);
-        min.y = std::min(min.y, y);
-        max.y = std::max(max.y, y);
-        min.z = std::min(min.z, z);
-        max.z = std::max(max.z, z);
+        root_min.x = std::min(root_min.x, x);
+        root_max.x = std::max(root_max.x, x);
+        root_min.y = std::min(root_min.y, y);
+        root_max.y = std::max(root_max.y, y);
+        root_min.z = std::min(root_min.z, z);
+        root_max.z = std::max(root_max.z, z);
     }
 
-    // Pad the bounding box
-    constexpr float pad = 1.0;
-    min = Vector3Subtract(min, Vector3Scale(Vector3One(), pad));
-    max = Vector3Add(max, Vector3Scale(Vector3One(), pad));
+    constexpr float pad = 1.0f;
+    root_min = Vector3Subtract(root_min, Vector3Scale(Vector3One(), pad));
+    root_max = Vector3Add(root_max, Vector3Scale(Vector3One(), pad));
 
-    // Make it cubic (so subdivisions are balanced)
-    const float max_extent = std::max({max.x - min.x, max.y - min.y, max.z - min.z});
-    max = Vector3Add(min, Vector3Scale(Vector3One(), max_extent));
+    const float max_extent = std::max({root_max.x - root_min.x, root_max.y - root_min.y, root_max.z - root_min.z});
+    root_max = Vector3Add(root_min, Vector3Scale(Vector3One(), max_extent));
 
-    // Root node spans the entire area
-    const int root = t.create_empty_leaf(min, max);
+    const float root_extent = root_max.x - root_min.x; // cubic
 
-    for (size_t i = 0; i < positions.size(); ++i) {
-        t.insert(root, static_cast<int>(i), positions[i], MASS, 0);
+    // Container for building the particle list before sorting by morton code
+    struct sort_node
+    {
+        u64 code;
+        u32 id;
+        Vector3 pos;
+    };
+
+    // Calculate morton code for each node
+    std::vector<sort_node> sort_container;
+    sort_container.resize(positions.size());
+
+    const auto calculate_morton = [&](const u32 i)
+    {
+        sort_container[i] = {pos_to_morton(positions[i], root_min, root_max), i, positions[i]};
+    };
+
+    if (thread_pool) {
+        (*thread_pool)->submit_loop(0, positions.size(), calculate_morton, SMALL_TASK_BLOCK_SIZE).wait();
+    } else {
+        for (u32 i = 0; i < positions.size(); ++i) {
+            calculate_morton(i);
+        }
     }
+
+    // Sort the list by morton codes. Because positions close to each other have similar morten codes,
+    // this provides us with "spatial locality" in the datastructure.
+    auto sort_by_code = [&]()
+    {
+        std::ranges::sort(sort_container,
+                          [](const sort_node& a, const sort_node& b)
+                          {
+                              if (a.code != b.code) {
+                                  return a.code < b.code;
+                              }
+                              return a.id < b.id;
+                          });
+    };
+
+    sort_by_code();
+
+    // Resolve duplicates by jittering the later one deterministically and re-encoding.
+    for (int seed = 0; seed < 8; ++seed) {
+        bool had_dupes = false;
+
+        for (size_t i = 1; i < sort_container.size(); ++i) {
+            // Because elements are spatially ordered after sorting, we can scan for duplicates in O(n)
+            if (sort_container[i].code == sort_container[i - 1].code) {
+                had_dupes = true;
+                sort_container[i].pos = jitter_pos(sort_container[i].pos,
+                                                   sort_container[i].id + seed * 0x9e3779b9u,
+                                                   root_min,
+                                                   root_max,
+                                                   root_extent);
+                sort_container[i].code = pos_to_morton(sort_container[i].pos, root_min, root_max);
+            }
+        }
+
+        if (!had_dupes) {
+            break;
+        }
+        sort_by_code();
+    }
+
+    // Sanity check
+    for (size_t i = 1; i < sort_container.size(); ++i) {
+        if (sort_container[i].code == sort_container[i - 1].code) {
+            throw std::runtime_error("Duplicates remain after jittering");
+        }
+    }
+
+    std::vector<std::vector<node>> tree_levels;
+    tree_levels.assign(MAX_DEPTH + 1, {});
+
+    // Leaves at MAX_DEPTH: 1 particle per leaf in morton order (close particles close together)
+    auto& leafs = tree_levels[MAX_DEPTH];
+    leafs.reserve(sort_container.size());
+    const float leaf_size = root_extent / static_cast<float>(1u << MAX_DEPTH);
+    for (const auto& [code, id, pos] : sort_container) {
+        node leaf;
+        leaf.leaf = true;
+        leaf.mass_id = static_cast<int>(id);
+        leaf.depth = MAX_DEPTH;
+        leaf.size = leaf_size;
+        leaf.mass_total = MASS;
+        leaf.mass_center = pos; // force uses mass_center instead of jittered position
+        leaf.children.fill(-1);
+        leafs.push_back(leaf);
+    }
+
+    // We now have to group the particles (currently we have only sorted the leaves),
+    // but upwards subdivisions have to be created.
+    // For grouping, store a nodes local index in its level.
+    struct leaf
+    {
+        u64 leaf_code;
+        int depth;
+        int level_index;
+    };
+
+    std::vector<leaf> leaves;
+    leaves.reserve(tree_levels[MAX_DEPTH].size());
+    for (int i = 0; i < static_cast<int>(tree_levels[MAX_DEPTH].size()); ++i) {
+        leaves.emplace_back(sort_container[static_cast<size_t>(i)].code, MAX_DEPTH, i);
+    }
+
+    // Build internal levels from MAX_DEPTH-1 to 0
+    for (int current_depth = MAX_DEPTH - 1; current_depth >= 0; --current_depth) {
+        auto& current_level = tree_levels[current_depth];
+        current_level.clear();
+
+        std::vector<leaf> next_refs;
+        next_refs.reserve(leaves.size());
+
+        const float parent_size = root_extent / static_cast<float>(1u << current_depth);
+
+        size_t i = 0;
+        while (i < leaves.size()) {
+            const u64 key = path_to_ancestor(leaves[i].leaf_code, MAX_DEPTH, current_depth);
+
+            size_t j = i + 1;
+            while (j < leaves.size() && path_to_ancestor(leaves[j].leaf_code, MAX_DEPTH, current_depth) == key) {
+                ++j;
+            }
+
+            const size_t group_size = j - i;
+
+            // Unary compression: just carry the child ref upward unchanged.
+            if (group_size == 1) {
+                next_refs.push_back(leaves[i]);
+                i = j;
+                continue;
+            }
+
+            node parent;
+            parent.leaf = false;
+            parent.mass_id = -1;
+            parent.depth = current_depth;
+            parent.size = parent_size;
+            parent.children.fill(-1);
+
+            float mass_total = 0.0f;
+            Vector3 mass_center_acc = Vector3Zero();
+
+            for (size_t k = i; k < j; ++k) {
+                const int child_depth = leaves[k].depth;
+                const int child_local = leaves[k].level_index;
+
+                // Read the child from its actual stored level.
+                const node& child = tree_levels[child_depth][child_local];
+
+                // Which octant of this parent does it belong to?
+                // Octant comes from the NEXT level after current_depth,
+                // but the child might skip levels due to compression.
+                // We must use the child's first level under the parent (current_depth+1).
+                const int oct = octant_at_level(leaves[k].leaf_code, current_depth + 1, MAX_DEPTH);
+
+                // Store global child reference: we only have an int slot, so we need a single index space.
+                parent.children[oct] = (child_depth << 24) | (child_local & 0x00FFFFFF);
+
+                mass_total += child.mass_total;
+                mass_center_acc = Vector3Add(mass_center_acc, Vector3Scale(child.mass_center, child.mass_total));
+            }
+
+            parent.mass_total = mass_total;
+            parent.mass_center = (mass_total > 0.0f) ? Vector3Scale(mass_center_acc, 1.0f / mass_total) : Vector3Zero();
+
+            const int parent_local = static_cast<int>(current_level.size());
+            current_level.push_back(parent);
+
+            next_refs.push_back({leaves[i].leaf_code, current_depth, parent_local});
+            i = j;
+        }
+
+        leaves.swap(next_refs);
+    }
+
+    // Flatten levels and fix child indices from local->global.
+    // All depth 0 nodes come first, then depth 1, last depth MAX_DEPTH.
+    std::vector<int> offsets(tree_levels.size(), 0);
+    int total = 0;
+    for (int d = 0; d <= MAX_DEPTH; ++d) {
+        offsets[d] = total;
+        total += static_cast<int>(tree_levels[d].size());
+    }
+
+    t.nodes.clear();
+    t.nodes.reserve(total);
+    for (int d = 0; d <= MAX_DEPTH; ++d) {
+        t.nodes.insert(t.nodes.end(), tree_levels[d].begin(), tree_levels[d].end());
+    }
+
+    // Fix child indices: convert local index in levels[d+1] to global index in t.nodes
+    for (int d = 0; d <= MAX_DEPTH; ++d) {
+        const int base = offsets[d];
+        for (int i2 = 0; i2 < static_cast<int>(tree_levels[d].size()); ++i2) {
+            node& n = t.nodes[base + i2];
+            if (!n.leaf) {
+                for (int c = 0; c < 8; ++c) {
+                    int packed = n.children[c];
+                    if (packed >= 0) {
+                        const int child_depth = (packed >> 24) & 0xFF;
+                        const int child_local = packed & 0x00FFFFFF;
+                        n.children[c] = offsets[child_depth] + child_local;
+                    }
+                }
+            }
+        }
+    }
+
+    // const size_t _leaves = tree_levels[MAX_DEPTH].size();
+    // const size_t _total = t.nodes.size();
+    // const size_t _internal = _total - _leaves;
+    // traceln("Morton octree nodes: {}, leaves: {}, ratio: {:.3} children per internal node.",
+    //         _total,
+    //         _leaves,
+    //         static_cast<float>(_total - 1) / _internal);
 }
 
-auto octree::calculate_force(const int node_idx, const Vector3& pos) const -> Vector3
+auto octree::calculate_force_morton(const int node_idx, const Vector3& pos, const int self_id) const -> Vector3
 {
     if (node_idx < 0) {
         return Vector3Zero();
     }
 
-    const node& n = nodes[node_idx];
-    if (std::abs(n.mass_total) <= 0.001f) {
-        return Vector3Zero();
-    }
+    // Force accumulator
+    float fx = 0.0f;
+    float fy = 0.0f;
+    float fz = 0.0f;
 
-    const Vector3 diff = Vector3Subtract(pos, n.mass_center);
-    float dist_sq = diff.x * diff.x + diff.y * diff.y + diff.z * diff.z;
+    std::vector<int> stack;
+    stack.reserve(512);
+    stack.push_back(node_idx);
 
-    // Softening
-    dist_sq += SOFTENING;
+    constexpr float theta2 = THETA * THETA;
 
-    // Barnes-Hut
-    const float size = n.box_max.x - n.box_min.x;
-    if (n.leaf || size * size / dist_sq < THETA * THETA) {
-        const float dist = std::sqrt(dist_sq);
-        const float force_mag = BH_FORCE * n.mass_total / dist_sq;
+    while (!stack.empty()) {
+        const int idx = stack.back();
+        stack.pop_back();
 
-        return Vector3Scale(diff, force_mag / dist);
-    }
+        const node& n = nodes[idx];
 
-    // Collect child forces
-    Vector3 force = Vector3Zero();
-    for (const int child : n.children) {
-        if (child >= 0) {
-            const Vector3 child_force = calculate_force(child, pos);
+        // No self-force for single-particle leafs
+        if (n.leaf && n.mass_id == self_id) {
+            continue;
+        }
 
-            force = Vector3Add(force, child_force);
+        const float dx = pos.x - n.mass_center.x;
+        const float dy = pos.y - n.mass_center.y;
+        const float dz = pos.z - n.mass_center.z;
+
+        const float dist_sq = dx * dx + dy * dy + dz * dz + SOFTENING;
+
+        // Barnes–Hut criterion
+        if (n.leaf || ((n.size * n.size) / dist_sq) < theta2) {
+            const float inv_dist = 1.0f / std::sqrt(dist_sq);
+            const float force_mag = (BH_FORCE * n.mass_total) / dist_sq; // ~ 1/r^2
+            const float s = force_mag * inv_dist;                        // scale by 1/r to get vector
+            fx += dx * s;
+            fy += dy * s;
+            fz += dz * s;
+            continue;
+        }
+
+        for (int c = 0; c < 8; ++c) {
+            const int child = n.children[c];
+            if (child >= 0) {
+                stack.push_back(child);
+            }
         }
     }
 
-    return force;
+    return Vector3{fx, fy, fz};
 }

src/orbit_camera.cpp

@@ -31,24 +31,27 @@ auto orbit_camera::pan(const Vector2 last_mouse, const Vector2 mouse) -> void
     const Vector3 right = Vector3Normalize(Vector3CrossProduct(forward, camera.up));
     const Vector3 up = Vector3Normalize(Vector3CrossProduct(right, forward));
 
-    const Vector3 offset =
-        Vector3Add(Vector3Scale(right, -dx * speed), Vector3Scale(up, dy * speed));
+    const Vector3 offset = Vector3Add(Vector3Scale(right, -dx * speed), Vector3Scale(up, dy * speed));
 
     target = Vector3Add(target, offset);
 }
 
-auto orbit_camera::update(const Vector3& current_target, const Vector3& mass_center,
-                          const bool lock, const bool mass_center_lock) -> void
+auto orbit_camera::update(const Vector3& current_target,
+                          const Vector3& mass_center,
+                          const bool lock,
+                          const bool mass_center_lock) -> void
 {
     if (lock) {
         if (mass_center_lock) {
-            target = Vector3MoveTowards(target, mass_center,
-                                        CAMERA_SMOOTH_SPEED * GetFrameTime() *
-                                            Vector3Length(Vector3Subtract(target, mass_center)));
+            target = Vector3MoveTowards(target,
+                                        mass_center,
+                                        CAMERA_SMOOTH_SPEED * GetFrameTime() * Vector3Length(
+                                            Vector3Subtract(target, mass_center)));
         } else {
-            target = Vector3MoveTowards(target, current_target,
-                                        CAMERA_SMOOTH_SPEED * GetFrameTime() *
-                                            Vector3Length(Vector3Subtract(target, current_target)));
+            target = Vector3MoveTowards(target,
+                                        current_target,
+                                        CAMERA_SMOOTH_SPEED * GetFrameTime() * Vector3Length(
+                                            Vector3Subtract(target, current_target)));
         }
     }
 
@@ -63,7 +66,11 @@ auto orbit_camera::update(const Vector3& current_target, const Vector3& mass_cen
     const float y = sin(angle_y) * actual_distance;
     const float z = cos(angle_y) * cos(angle_x) * actual_distance;
 
-    fov = Clamp(fov, MIN_FOV, MAX_FOV);
+    if (projection == CAMERA_ORTHOGRAPHIC) {
+        fov = Clamp(fov, MIN_FOV, MAX_ORTHO_FOV);
+    } else {
+        fov = Clamp(fov, MIN_FOV, MAX_PERSP_FOV);
+    }
 
     camera.position = Vector3Add(target, Vector3(x, y, z));
     camera.target = target;

src/renderer.cpp

@@ -4,6 +4,7 @@
 #include <raylib.h>
 #include <raymath.h>
 #include <rlgl.h>
+#include <GL/glew.h>
 
 auto renderer::update_texture_sizes() -> void
 {
@@ -34,10 +35,24 @@ auto renderer::draw_mass_springs(const std::vector<Vector3>& masses) -> void
         return;
     }
 
-    // Prepare connection batching
+    // Prepare edge buffer
     {
         #ifdef TRACY
-        ZoneNamedN(prepare_connections, "PrepareConnectionsBatching", true);
+        ZoneNamedN(prepare_edge_buffers, "PrepareEdgeBuffers", true);
+        #endif
+
+        edge_vertices.clear();
+        for (const auto& [from, to] : state.get_links()) {
+            edge_vertices.push_back(masses[from]);
+            edge_vertices.push_back(masses[to]);
+        }
+        rlUpdateVertexBuffer(edge_vbo_id, edge_vertices.data(), edge_vertices.size() * sizeof(Vector3), 0);
+    }
+
+    // Prepare connection drawing
+    {
+        #ifdef TRACY
+        ZoneNamedN(prepare_connections, "PrepareConnectionsDrawing", true);
         #endif
 
         connections.clear();
@@ -47,7 +62,6 @@ auto renderer::draw_mass_springs(const std::vector<Vector3>& masses) -> void
                 const Vector3& current_mass = masses[state.get_current_index()];
                 const Vector3& winning_mass = masses[_state];
                 connections.emplace_back(current_mass, winning_mass);
-                DrawLine3D(current_mass, winning_mass, Fade(TARGET_BLOCK_COLOR, 0.5));
             }
         }
     }
@@ -63,6 +77,67 @@ auto renderer::draw_mass_springs(const std::vector<Vector3>& masses) -> void
             transforms.clear();
             colors.clear();
 
+            // Collisions
+            // TODO: This would benefit greatly from a spatial data structure.
+            //       Would it be worth to copy the octree from the physics thread?
+            input.collision_mass = -1;
+            if (input.mouse_in_graph_pane() && IsKeyDown(KEY_Q)) {
+                #ifdef TRACY
+                ZoneNamedN(mass_collisions, "MassCollisions", true);
+                #endif
+                const Ray ray = GetScreenToWorldRayEx(
+                    GetMousePosition() - Vector2(GetScreenWidth() / 2.0f, MENU_HEIGHT),
+                    camera.camera,
+                    graph_target.texture.width,
+                    graph_target.texture.height);
+                // Ray collision hit info
+
+                size_t mass = 0;
+                for (const auto& [x, y, z] : masses) {
+                    const RayCollision collision = GetRayCollisionBox(ray,
+                                                                      BoundingBox{
+                                                                          {
+                                                                              x - VERTEX_SIZE / 2.0f,
+                                                                              y - VERTEX_SIZE / 2.0f,
+                                                                              z - VERTEX_SIZE / 2.0f
+                                                                          },
+                                                                          {
+                                                                              x + VERTEX_SIZE / 2.0f,
+                                                                              y + VERTEX_SIZE / 2.0f,
+                                                                              z + VERTEX_SIZE / 2.0f
+                                                                          }
+                                                                      });
+                    if (collision.hit) {
+                        input.collision_mass = mass;
+                        break;
+                    }
+                    ++mass;
+                }
+            }
+
+            // Find max distance to interpolate colors in the given [0, max] range
+            int max_distance = 0;
+            for (const int distance : state.get_distances()) {
+                if (distance > max_distance) {
+                    max_distance = distance;
+                }
+            }
+
+            const auto lerp_color = [&](const Color from, const Color to, const int distance)
+            {
+                const float weight = 1.0 - static_cast<float>(distance) / max_distance;
+
+                Color result;
+                result.r = static_cast<u8>((1 - weight) * from.r + weight * to.r);
+                result.g = static_cast<u8>((1 - weight) * from.g + weight * to.g);
+                result.b = static_cast<u8>((1 - weight) * from.b + weight * to.b);
+                result.a = static_cast<u8>((1 - weight) * from.a + weight * to.a);
+
+                return result;
+            };
+
+            const std::vector<int>& distances = state.get_distances();
+
             size_t mass = 0;
             for (const auto& [x, y, z] : masses) {
                 transforms.emplace_back(MatrixTranslate(x, y, z));
@@ -81,9 +156,14 @@ auto renderer::draw_mass_springs(const std::vector<Vector3>& masses) -> void
                 } else if (state.get_visit_counts().at(mass) > 0) {
                     // Visited vertex
                     c = VERTEX_VISITED_COLOR;
+                } else if (input.color_by_distance && distances.size() == masses.size()) {
+                    c = lerp_color(VERTEX_FARTHEST_COLOR, VERTEX_CLOSEST_COLOR, static_cast<float>(distances[mass]));
+                }
+                if (mass == input.collision_mass) {
+                    c = RED;
                 }
-                // Current vertex is drawn as individual cube to increase its size
 
+                // Current vertex is drawn as individual cube to increase its size
                 colors.emplace_back(c);
                 ++mass;
             }
@@ -96,23 +176,53 @@ auto renderer::draw_mass_springs(const std::vector<Vector3>& masses) -> void
     ClearBackground(RAYWHITE);
     BeginMode3D(camera.camera);
 
-    // Draw springs (batched)
+    rlDrawRenderBatchActive();
+
+    // Draw edges
     {
         #ifdef TRACY
         ZoneNamedN(draw_springs, "DrawSprings", true);
         #endif
 
-        rlBegin(RL_LINES);
-        for (const auto& [from, to] : state.get_links()) {
-            if (masses.size() > from && masses.size() > to) {
-                const auto& [ax, ay, az] = masses[from];
-                const auto& [bx, by, bz] = masses[to];
-                rlColor4ub(EDGE_COLOR.r, EDGE_COLOR.g, EDGE_COLOR.b, EDGE_COLOR.a);
-                rlVertex3f(ax, ay, az);
-                rlVertex3f(bx, by, bz);
-            }
-        }
-        rlEnd();
+        rlEnableShader(edge_shader.id);
+
+        Matrix modelview = rlGetMatrixModelview();
+        Matrix projection = rlGetMatrixProjection();
+        Matrix mvp = MatrixMultiply(modelview, projection);
+        rlSetUniformMatrix(edge_shader.locs[SHADER_LOC_MATRIX_MVP], mvp);
+
+        const std::array<float, 4> edge_color = {
+            EDGE_COLOR.r / 255.0f,
+            EDGE_COLOR.g / 255.0f,
+            EDGE_COLOR.b / 255.0f,
+            EDGE_COLOR.a / 255.0f
+        };
+        rlSetUniform(edge_color_loc, edge_color.data(), SHADER_UNIFORM_VEC4, 1);
+
+        glBindVertexArray(edge_vao_id);
+        glDrawArrays(GL_LINES, 0, edge_vertices.size());
+        glBindVertexArray(0);
+
+        rlDisableShader();
+
+        // This draws triangles:
+        // rlEnableVertexArray(edge_vao_id);
+        // rlColor4ub(EDGE_COLOR.r, EDGE_COLOR.g, EDGE_COLOR.b, EDGE_COLOR.a);
+        // rlDrawVertexArray(0, edge_vertices.size());
+        // rlDisableVertexArray();
+
+        // This is fucking slow:
+        // rlBegin(RL_LINES);
+        // for (const auto& [from, to] : state.get_links()) {
+        //     if (masses.size() > from && masses.size() > to) {
+        //         const auto& [ax, ay, az] = masses[from];
+        //         const auto& [bx, by, bz] = masses[to];
+        //         rlColor4ub(EDGE_COLOR.r, EDGE_COLOR.g, EDGE_COLOR.b, EDGE_COLOR.a);
+        //         rlVertex3f(ax, ay, az);
+        //         rlVertex3f(bx, by, bz);
+        //     }
+        // }
+        // rlEnd();
     }
 
     // Draw masses (instanced)
@@ -181,32 +291,47 @@ auto renderer::draw_menu() const -> void
     EndTextureMode();
 }
 
-auto renderer::draw_textures(const int fps, const int ups, const size_t mass_count,
+auto renderer::draw_textures(const int fps,
+                             const int ups,
+                             const size_t mass_count,
                              const size_t spring_count) const -> void
 {
     BeginDrawing();
 
-    DrawTextureRec(menu_target.texture, Rectangle(0, 0, menu_target.texture.width, -menu_target.texture.height),
-                   Vector2(0, 0), WHITE);
+    DrawTextureRec(menu_target.texture,
+                   Rectangle(0, 0, menu_target.texture.width, -menu_target.texture.height),
+                   Vector2(0, 0),
+                   WHITE);
     DrawTextureRec(klotski_target.texture,
                    Rectangle(0, 0, klotski_target.texture.width, -klotski_target.texture.height),
-                   Vector2(0, MENU_HEIGHT), WHITE);
-    DrawTextureRec(graph_target.texture, Rectangle(0, 0, graph_target.texture.width, -graph_target.texture.height),
-                   Vector2(GetScreenWidth() / 2.0f, MENU_HEIGHT), WHITE);
+                   Vector2(0, MENU_HEIGHT),
+                   WHITE);
+    DrawTextureRec(graph_target.texture,
+                   Rectangle(0, 0, graph_target.texture.width, -graph_target.texture.height),
+                   Vector2(GetScreenWidth() / 2.0f, MENU_HEIGHT),
+                   WHITE);
 
     // Draw borders
     DrawRectangleLinesEx(Rectangle(0, 0, GetScreenWidth(), MENU_HEIGHT), 1.0f, BLACK);
-    DrawRectangleLinesEx(Rectangle(0, MENU_HEIGHT, GetScreenWidth() / 2.0f, GetScreenHeight() - MENU_HEIGHT), 1.0f,
+    DrawRectangleLinesEx(Rectangle(0, MENU_HEIGHT, GetScreenWidth() / 2.0f, GetScreenHeight() - MENU_HEIGHT),
+                         1.0f,
+                         BLACK);
+    DrawRectangleLinesEx(Rectangle(GetScreenWidth() / 2.0f,
+                                   MENU_HEIGHT,
+                                   GetScreenWidth() / 2.0f,
+                                   GetScreenHeight() - MENU_HEIGHT),
+                         1.0f,
                          BLACK);
-    DrawRectangleLinesEx(Rectangle(GetScreenWidth() / 2.0f, MENU_HEIGHT, GetScreenWidth() / 2.0f,
-                                   GetScreenHeight() - MENU_HEIGHT), 1.0f, BLACK);
 
     gui.draw(fps, ups, mass_count, spring_count);
 
     EndDrawing();
 }
 
-auto renderer::render(const std::vector<Vector3>& masses, const int fps, const int ups, const size_t mass_count,
+auto renderer::render(const std::vector<Vector3>& masses,
+                      const int fps,
+                      const int ups,
+                      const size_t mass_count,
                       const size_t spring_count) -> void
 {
     update_texture_sizes();

src/state_manager.cpp

@@ -28,7 +28,7 @@ auto state_manager::synced_insert_link(size_t first_index, size_t second_index)
 }
 
 auto state_manager::synced_insert_statespace(const std::vector<puzzle>& states,
-                                             const std::vector<std::pair<size_t, size_t>>& _links) -> void
+                                             const std::vector<spring>& _links) -> void
 {
     if (!state_pool.empty() || !state_indices.empty() || !links.empty()) {
         warnln("Inserting statespace but collections haven't been cleared");
@@ -374,7 +374,7 @@ auto state_manager::get_path_length() const -> size_t
     return winning_path.size();
 }
 
-auto state_manager::get_links() const -> const std::vector<std::pair<size_t, size_t>>&
+auto state_manager::get_links() const -> const std::vector<spring>&
 {
     return links;
 }
@@ -429,6 +429,11 @@ auto state_manager::has_distances() const -> bool
     return !node_target_distances.empty();
 }
 
+auto state_manager::get_distances() const -> std::vector<int>
+{
+    return node_target_distances.distances;
+}
+
 auto state_manager::get_total_moves() const -> size_t
 {
     return total_moves;

src/user_interface.cpp

@@ -7,8 +7,12 @@
 #define RAYGUI_IMPLEMENTATION
 #include <raygui.h>
 
-auto user_interface::grid::update_bounds(const int _x, const int _y, const int _width, const int _height,
-                                         const int _columns, const int _rows) -> void
+auto user_interface::grid::update_bounds(const int _x,
+                                         const int _y,
+                                         const int _width,
+                                         const int _height,
+                                         const int _columns,
+                                         const int _rows) -> void
 {
     x = _x;
     y = _y;
@@ -51,7 +55,9 @@ auto user_interface::grid::bounds(const int _x, const int _y, const int _width,
     const int cell_width = (width - padding) / columns;
     const int cell_height = (height - padding) / rows;
 
-    return Rectangle(x + _x * cell_width + padding, y + _y * cell_height + padding, _width * cell_width - padding,
+    return Rectangle(x + _x * cell_width + padding,
+                     y + _y * cell_height + padding,
+                     _width * cell_width - padding,
                      _height * cell_height - padding);
 }
 
@@ -65,7 +71,9 @@ auto user_interface::grid::square_bounds() const -> Rectangle
     return bounds;
 }
 
-auto user_interface::grid::square_bounds(const int _x, const int _y, const int _width,
+auto user_interface::grid::square_bounds(const int _x,
+                                         const int _y,
+                                         const int _width,
                                          const int _height) const -> Rectangle
 {
     // Assumes each cell is square, so either width or height are not completely
@@ -84,8 +92,10 @@ auto user_interface::grid::square_bounds(const int _x, const int _y, const int _
     const int x_offset = (width - grid_width) / 2;
     const int y_offset = (height - grid_height) / 2;
 
-    return Rectangle(x_offset + _x * (cell_size + padding) + padding, y_offset + _y * (cell_size + padding) + padding,
-                     _width * cell_size + padding * (_width - 1), _height * cell_size + padding * (_height - 1));
+    return Rectangle(x_offset + _x * (cell_size + padding) + padding,
+                     y_offset + _y * (cell_size + padding) + padding,
+                     _width * cell_size + padding * (_width - 1),
+                     _height * cell_size + padding * (_height - 1));
 }
 
 auto user_interface::init() -> void
@@ -146,16 +156,26 @@ auto user_interface::get_default_style() -> default_style
     // access...
     return {
         {
-            GuiGetStyle(DEFAULT, BORDER_COLOR_NORMAL), GuiGetStyle(DEFAULT, BASE_COLOR_NORMAL),
-            GuiGetStyle(DEFAULT, TEXT_COLOR_NORMAL), GuiGetStyle(DEFAULT, BORDER_COLOR_FOCUSED),
-            GuiGetStyle(DEFAULT, BASE_COLOR_FOCUSED), GuiGetStyle(DEFAULT, TEXT_COLOR_FOCUSED),
-            GuiGetStyle(DEFAULT, BORDER_COLOR_PRESSED), GuiGetStyle(DEFAULT, BASE_COLOR_PRESSED),
-            GuiGetStyle(DEFAULT, TEXT_COLOR_PRESSED), GuiGetStyle(DEFAULT, BORDER_COLOR_DISABLED),
-            GuiGetStyle(DEFAULT, BASE_COLOR_DISABLED), GuiGetStyle(DEFAULT, TEXT_COLOR_DISABLED)
+            GuiGetStyle(DEFAULT, BORDER_COLOR_NORMAL),
+            GuiGetStyle(DEFAULT, BASE_COLOR_NORMAL),
+            GuiGetStyle(DEFAULT, TEXT_COLOR_NORMAL),
+            GuiGetStyle(DEFAULT, BORDER_COLOR_FOCUSED),
+            GuiGetStyle(DEFAULT, BASE_COLOR_FOCUSED),
+            GuiGetStyle(DEFAULT, TEXT_COLOR_FOCUSED),
+            GuiGetStyle(DEFAULT, BORDER_COLOR_PRESSED),
+            GuiGetStyle(DEFAULT, BASE_COLOR_PRESSED),
+            GuiGetStyle(DEFAULT, TEXT_COLOR_PRESSED),
+            GuiGetStyle(DEFAULT, BORDER_COLOR_DISABLED),
+            GuiGetStyle(DEFAULT, BASE_COLOR_DISABLED),
+            GuiGetStyle(DEFAULT, TEXT_COLOR_DISABLED)
         },
-        GuiGetStyle(DEFAULT, BACKGROUND_COLOR), GuiGetStyle(DEFAULT, LINE_COLOR), GuiGetStyle(DEFAULT, TEXT_SIZE),
-        GuiGetStyle(DEFAULT, TEXT_SPACING), GuiGetStyle(DEFAULT, TEXT_LINE_SPACING),
-        GuiGetStyle(DEFAULT, TEXT_ALIGNMENT_VERTICAL), GuiGetStyle(DEFAULT, TEXT_WRAP_MODE)
+        GuiGetStyle(DEFAULT, BACKGROUND_COLOR),
+        GuiGetStyle(DEFAULT, LINE_COLOR),
+        GuiGetStyle(DEFAULT, TEXT_SIZE),
+        GuiGetStyle(DEFAULT, TEXT_SPACING),
+        GuiGetStyle(DEFAULT, TEXT_LINE_SPACING),
+        GuiGetStyle(DEFAULT, TEXT_ALIGNMENT_VERTICAL),
+        GuiGetStyle(DEFAULT, TEXT_WRAP_MODE)
     };
 }
 
@@ -191,14 +211,21 @@ auto user_interface::get_component_style(const int component) -> component_style
 {
     return {
         {
-            GuiGetStyle(component, BORDER_COLOR_NORMAL), GuiGetStyle(component, BASE_COLOR_NORMAL),
-            GuiGetStyle(component, TEXT_COLOR_NORMAL), GuiGetStyle(component, BORDER_COLOR_FOCUSED),
-            GuiGetStyle(component, BASE_COLOR_FOCUSED), GuiGetStyle(component, TEXT_COLOR_FOCUSED),
-            GuiGetStyle(component, BORDER_COLOR_PRESSED), GuiGetStyle(component, BASE_COLOR_PRESSED),
-            GuiGetStyle(component, TEXT_COLOR_PRESSED), GuiGetStyle(component, BORDER_COLOR_DISABLED),
-            GuiGetStyle(component, BASE_COLOR_DISABLED), GuiGetStyle(component, TEXT_COLOR_DISABLED)
+            GuiGetStyle(component, BORDER_COLOR_NORMAL),
+            GuiGetStyle(component, BASE_COLOR_NORMAL),
+            GuiGetStyle(component, TEXT_COLOR_NORMAL),
+            GuiGetStyle(component, BORDER_COLOR_FOCUSED),
+            GuiGetStyle(component, BASE_COLOR_FOCUSED),
+            GuiGetStyle(component, TEXT_COLOR_FOCUSED),
+            GuiGetStyle(component, BORDER_COLOR_PRESSED),
+            GuiGetStyle(component, BASE_COLOR_PRESSED),
+            GuiGetStyle(component, TEXT_COLOR_PRESSED),
+            GuiGetStyle(component, BORDER_COLOR_DISABLED),
+            GuiGetStyle(component, BASE_COLOR_DISABLED),
+            GuiGetStyle(component, TEXT_COLOR_DISABLED)
         },
-        GuiGetStyle(component, BORDER_WIDTH), GuiGetStyle(component, TEXT_PADDING),
+        GuiGetStyle(component, BORDER_WIDTH),
+        GuiGetStyle(component, TEXT_PADDING),
         GuiGetStyle(component, TEXT_ALIGNMENT)
     };
 }
@@ -229,11 +256,16 @@ auto user_interface::set_component_style(const int component, const component_st
 auto user_interface::popup_bounds() -> Rectangle
 {
     return Rectangle(static_cast<float>(GetScreenWidth()) / 2.0f - POPUP_WIDTH / 2.0f,
-                     static_cast<float>(GetScreenHeight()) / 2.0f - POPUP_HEIGHT / 2.0f, POPUP_WIDTH, POPUP_HEIGHT);
+                     static_cast<float>(GetScreenHeight()) / 2.0f - POPUP_HEIGHT / 2.0f,
+                     POPUP_WIDTH,
+                     POPUP_HEIGHT);
 }
 
-auto user_interface::draw_button(const Rectangle bounds, const std::string& label, const Color color,
-                                 const bool enabled, const int font_size) const -> int
+auto user_interface::draw_button(const Rectangle bounds,
+                                 const std::string& label,
+                                 const Color color,
+                                 const bool enabled,
+                                 const int font_size) const -> int
 {
     // Save original styling
     const default_style original_default = get_default_style();
@@ -263,16 +295,25 @@ auto user_interface::draw_button(const Rectangle bounds, const std::string& labe
     return pressed;
 }
 
-auto user_interface::draw_menu_button(const int x, const int y, const int width, const int height,
-                                      const std::string& label, const Color color, const bool enabled,
+auto user_interface::draw_menu_button(const int x,
+                                      const int y,
+                                      const int width,
+                                      const int height,
+                                      const std::string& label,
+                                      const Color color,
+                                      const bool enabled,
                                       const int font_size) const -> int
 {
     const Rectangle bounds = menu_grid.bounds(x, y, width, height);
     return draw_button(bounds, label, color, enabled, font_size);
 }
 
-auto user_interface::draw_toggle_slider(const Rectangle bounds, const std::string& off_label,
-                                        const std::string& on_label, int* active, Color color, bool enabled,
+auto user_interface::draw_toggle_slider(const Rectangle bounds,
+                                        const std::string& off_label,
+                                        const std::string& on_label,
+                                        int* active,
+                                        Color color,
+                                        bool enabled,
                                         int font_size) const -> int
 {
     // Save original styling
@@ -308,16 +349,29 @@ auto user_interface::draw_toggle_slider(const Rectangle bounds, const std::strin
     return pressed;
 }
 
-auto user_interface::draw_menu_toggle_slider(const int x, const int y, const int width, const int height,
-                                             const std::string& off_label, const std::string& on_label, int* active,
-                                             const Color color, const bool enabled, const int font_size) const -> int
+auto user_interface::draw_menu_toggle_slider(const int x,
+                                             const int y,
+                                             const int width,
+                                             const int height,
+                                             const std::string& off_label,
+                                             const std::string& on_label,
+                                             int* active,
+                                             const Color color,
+                                             const bool enabled,
+                                             const int font_size) const -> int
 {
     const Rectangle bounds = menu_grid.bounds(x, y, width, height);
     return draw_toggle_slider(bounds, off_label, on_label, active, color, enabled, font_size);
 }
 
-auto user_interface::draw_spinner(Rectangle bounds, const std::string& label, int* value, int min, int max, Color color,
-                                  bool enabled, int font_size) const -> int
+auto user_interface::draw_spinner(Rectangle bounds,
+                                  const std::string& label,
+                                  int* value,
+                                  int min,
+                                  int max,
+                                  Color color,
+                                  bool enabled,
+                                  int font_size) const -> int
 {
     // Save original styling
     const default_style original_default = get_default_style();
@@ -352,15 +406,26 @@ auto user_interface::draw_spinner(Rectangle bounds, const std::string& label, in
     return pressed;
 }
 
-auto user_interface::draw_menu_spinner(const int x, const int y, const int width, const int height,
-                                       const std::string& label, int* value, const int min, const int max,
-                                       const Color color, const bool enabled, const int font_size) const -> int
+auto user_interface::draw_menu_spinner(const int x,
+                                       const int y,
+                                       const int width,
+                                       const int height,
+                                       const std::string& label,
+                                       int* value,
+                                       const int min,
+                                       const int max,
+                                       const Color color,
+                                       const bool enabled,
+                                       const int font_size) const -> int
 {
     const Rectangle bounds = menu_grid.bounds(x, y, width, height);
     return draw_spinner(bounds, label, value, min, max, color, enabled, font_size);
 }
 
-auto user_interface::draw_label(const Rectangle bounds, const std::string& text, const Color color, const bool enabled,
+auto user_interface::draw_label(const Rectangle bounds,
+                                const std::string& text,
+                                const Color color,
+                                const bool enabled,
                                 const int font_size) const -> int
 {
     // Save original styling
@@ -391,7 +456,11 @@ auto user_interface::draw_label(const Rectangle bounds, const std::string& text,
     return pressed;
 }
 
-auto user_interface::draw_board_block(const int x, const int y, const int width, const int height, const Color color,
+auto user_interface::draw_board_block(const int x,
+                                      const int y,
+                                      const int width,
+                                      const int height,
+                                      const Color color,
                                       const bool enabled) const -> bool
 {
     component_style s = get_component_style(BUTTON);
@@ -400,7 +469,7 @@ auto user_interface::draw_board_block(const int x, const int y, const int width,
     const Rectangle bounds = board_grid.square_bounds(x, y, width, height);
 
     const bool focused = CheckCollisionPointRec(input.mouse - Vector2(0, MENU_HEIGHT), bounds);
-    const bool pressed = puzzle::block(x, y, width, height, false).covers(input.sel_x, input.sel_y);
+    const bool pressed = block(x, y, width, height, false).covers(input.sel_x, input.sel_y);
 
     // Background to make faded colors work
     DrawRectangleRec(bounds, RAYWHITE);
@@ -438,7 +507,15 @@ auto user_interface::window_open() const -> bool
 auto user_interface::draw_menu_header(const Color color) const -> void
 {
     int preset = static_cast<int>(state.get_current_preset());
-    draw_menu_spinner(0, 0, 1, 1, "Preset: ", &preset, -1, static_cast<int>(state.get_preset_count()), color,
+    draw_menu_spinner(0,
+                      0,
+                      1,
+                      1,
+                      "Preset: ",
+                      &preset,
+                      -1,
+                      static_cast<int>(state.get_preset_count()),
+                      color,
                       !input.editing);
     if (preset > static_cast<int>(state.get_current_preset())) {
         input.load_next_preset();
@@ -446,11 +523,17 @@ auto user_interface::draw_menu_header(const Color color) const -> void
         input.load_previous_preset();
     }
 
-    draw_menu_button(1, 0, 1, 1, std::format("{}: {}/{} Blocks",
-                                             state.was_edited()
-                                                 ? "Modified"
-                                                 : std::format("\"{}\"", state.get_current_preset_comment().substr(2)),
-                                             state.get_current_state().block_count(), puzzle::MAX_BLOCKS), color);
+    draw_menu_button(1,
+                     0,
+                     1,
+                     1,
+                     std::format("{}: {}/{} Blocks",
+                                 state.was_edited()
+                                     ? "Modified"
+                                     : std::format("\"{}\"", state.get_current_preset_comment().substr(2)),
+                                 state.get_current_state().block_count(),
+                                 puzzle::MAX_BLOCKS),
+                     color);
 
     int editing = input.editing;
     draw_menu_toggle_slider(2, 0, 1, 1, "Puzzle Mode (Tab)", "Edit Mode (Tab)", &editing, color);
@@ -461,13 +544,22 @@ auto user_interface::draw_menu_header(const Color color) const -> void
 
 auto user_interface::draw_graph_info(const Color color) const -> void
 {
-    draw_menu_button(0, 1, 1, 1, std::format("Found {} States ({} Winning)", state.get_state_count(),
-                                             state.get_target_count()), color);
+    draw_menu_button(0,
+                     1,
+                     1,
+                     1,
+                     std::format("Found {} States ({} Winning)", state.get_state_count(), state.get_target_count()),
+                     color);
 
     draw_menu_button(1, 1, 1, 1, std::format("Found {} Transitions", state.get_link_count()), color);
 
-    draw_menu_button(2, 1, 1, 1, std::format("{} Moves to Nearest Solution",
-                                             state.get_path_length() > 0 ? state.get_path_length() - 1 : 0), color);
+    draw_menu_button(2,
+                     1,
+                     1,
+                     1,
+                     std::format("{} Moves to Nearest Solution",
+                                 state.get_path_length() > 0 ? state.get_path_length() - 1 : 0),
+                     color);
 }
 
 auto user_interface::draw_graph_controls(const Color color) const -> void
@@ -504,7 +596,14 @@ auto user_interface::draw_camera_controls(const Color color) const -> void
     }
 
     int lock_camera_mass_center = input.camera_mass_center_lock;
-    draw_menu_toggle_slider(1, 3, 1, 1, "Current Block (U)", "Graph Center (U)", &lock_camera_mass_center, color,
+    draw_menu_toggle_slider(1,
+                            3,
+                            1,
+                            1,
+                            "Current Block (U)",
+                            "Graph Center (U)",
+                            &lock_camera_mass_center,
+                            color,
                             input.camera_lock);
     if (lock_camera_mass_center != input.camera_mass_center_lock) {
         input.toggle_camera_mass_center_lock();
@@ -537,8 +636,12 @@ auto user_interface::draw_puzzle_controls(const Color color) const -> void
     };
 
     const int visits = state.get_current_visits();
-    draw_menu_button(0, 4, 1, 1, std::format("{} Moves ({}{} Time at this State)", state.get_total_moves(), visits,
-                                             nth(visits)), color);
+    draw_menu_button(0,
+                     4,
+                     1,
+                     1,
+                     std::format("{} Moves ({}{} Time at this State)", state.get_total_moves(), visits, nth(visits)),
+                     color);
 
     if (draw_menu_button(1, 4, 1, 1, "Make Optimal Move (Space)", color, state.has_distances())) {
         input.goto_optimal_next_state();
@@ -556,7 +659,12 @@ auto user_interface::draw_puzzle_controls(const Color color) const -> void
         input.goto_most_distant_state();
     }
 
-    if (draw_menu_button(2, 5, 1, 1, "Go to Starting State (R)", color,
+    if (draw_menu_button(2,
+                         5,
+                         1,
+                         1,
+                         "Go to Starting State (R)",
+                         color,
                          state.get_current_index() != state.get_starting_index())) {
         input.goto_starting_state();
     }
@@ -634,8 +742,13 @@ auto user_interface::draw_save_preset_popup() -> void
     }
 
     // Returns the pressed button index
-    const int button = GuiTextInputBox(popup_bounds(), "Save as Preset", "Enter Preset Name", "Ok;Cancel",
-                                       preset_comment.data(), 255, nullptr);
+    const int button = GuiTextInputBox(popup_bounds(),
+                                       "Save as Preset",
+                                       "Enter Preset Name",
+                                       "Ok;Cancel",
+                                       preset_comment.data(),
+                                       255,
+                                       nullptr);
     if (button == 1) {
         state.save_current_to_preset_file(preset_comment.data());
     }
@@ -698,7 +811,11 @@ auto user_interface::draw_puzzle_board() -> void
 {
     const puzzle& current = state.get_current_state();
 
-    board_grid.update_bounds(0, MENU_HEIGHT, GetScreenWidth() / 2, GetScreenHeight() - MENU_HEIGHT, current.get_width(),
+    board_grid.update_bounds(0,
+                             MENU_HEIGHT,
+                             GetScreenWidth() / 2,
+                             GetScreenHeight() - MENU_HEIGHT,
+                             current.get_width(),
                              current.get_height());
 
     // Draw outer border
@@ -706,18 +823,22 @@ auto user_interface::draw_puzzle_board() -> void
     DrawRectangleRec(bounds, current.goal_reached() ? BOARD_COLOR_WON : BOARD_COLOR_RESTRICTED);
 
     // Draw inner borders
-    DrawRectangle(bounds.x + BOARD_PADDING, bounds.y + BOARD_PADDING, bounds.width - 2 * BOARD_PADDING,
+    DrawRectangle(bounds.x + BOARD_PADDING,
+                  bounds.y + BOARD_PADDING,
+                  bounds.width - 2 * BOARD_PADDING,
                   bounds.height - 2 * BOARD_PADDING,
                   current.get_restricted() ? BOARD_COLOR_RESTRICTED : BOARD_COLOR_FREE);
 
     // Draw target opening
     // TODO: Only draw single direction (in corner) if restricted (use target block principal
     // direction)
-    const std::optional<puzzle::block> target_block = current.try_get_target_block();
+    const std::optional<block> target_block = current.try_get_target_block();
     const int target_x = current.get_goal_x();
     const int target_y = current.get_goal_y();
     if (current.get_goal() && target_block) {
-        auto [x, y, width, height] = board_grid.square_bounds(target_x, target_y, target_block->get_width(),
+        auto [x, y, width, height] = board_grid.square_bounds(target_x,
+                                                              target_y,
+                                                              target_block->get_width(),
                                                               target_block->get_height());
 
         const Color opening_color = Fade(current.goal_reached() ? BOARD_COLOR_WON : BOARD_COLOR_RESTRICTED, 0.3);
@@ -764,7 +885,7 @@ auto user_interface::draw_puzzle_board() -> void
     }
 
     // Draw blocks
-    for (const puzzle::block b : current.block_view()) {
+    for (const block b : current.block_view()) {
         Color c = BLOCK_COLOR;
         if (b.get_target()) {
             c = TARGET_BLOCK_COLOR;
@@ -782,16 +903,22 @@ auto user_interface::draw_puzzle_board() -> void
         if (current.covers(input.block_add_x, input.block_add_y) && input.hov_x >= input.block_add_x && input.hov_y >=
             input.block_add_y) {
             bool collides = false;
-            for (const puzzle::block b : current.block_view()) {
-                if (b.collides(puzzle::block(input.block_add_x, input.block_add_y, input.hov_x - input.block_add_x + 1,
-                                             input.hov_y - input.block_add_y + 1, false))) {
+            for (const block b : current.block_view()) {
+                if (b.collides(block(input.block_add_x,
+                                     input.block_add_y,
+                                     input.hov_x - input.block_add_x + 1,
+                                     input.hov_y - input.block_add_y + 1,
+                                     false))) {
                     collides = true;
                     break;
                 }
             }
             if (!collides) {
-                draw_board_block(input.block_add_x, input.block_add_y, input.hov_x - input.block_add_x + 1,
-                                 input.hov_y - input.block_add_y + 1, PURPLE);
+                draw_board_block(input.block_add_x,
+                                 input.block_add_y,
+                                 input.hov_x - input.block_add_x + 1,
+                                 input.hov_y - input.block_add_y + 1,
+                                 PURPLE);
             }
         }
     }
@@ -803,7 +930,8 @@ auto user_interface::draw_puzzle_board() -> void
     // Draw goal boundaries when editing
     if (input.editing && current.get_goal() && target_block) {
         DrawRectangleLinesEx(
-            board_grid.square_bounds(target_x, target_y, target_block->get_width(), target_block->get_height()), 2.0,
+            board_grid.square_bounds(target_x, target_y, target_block->get_width(), target_block->get_height()),
+            2.0,
             TARGET_BLOCK_COLOR);
     }
 }

test/bitmap.cpp

@@ -10,7 +10,7 @@ TEST_CASE("bitmap_is_full all bits set", "[puzzle][board]")
     puzzle p2(3, 4);
     puzzle p3(5, 4);
     puzzle p4(3, 7);
-    uint64_t bitmap = -1;
+    u64 bitmap = -1;
 
     REQUIRE(p1.bitmap_is_full(bitmap));
     REQUIRE(p2.bitmap_is_full(bitmap));
@@ -24,7 +24,7 @@ TEST_CASE("bitmap_is_full no bits set", "[puzzle][board]")
     puzzle p2(3, 4);
     puzzle p3(5, 4);
     puzzle p4(3, 7);
-    uint64_t bitmap = 0;
+    u64 bitmap = 0;
 
     REQUIRE_FALSE(p1.bitmap_is_full(bitmap));
     REQUIRE_FALSE(p2.bitmap_is_full(bitmap));
@@ -39,10 +39,10 @@ TEST_CASE("bitmap_is_full necessary bits set", "[puzzle][board]")
     puzzle p3(5, 4);
     puzzle p4(3, 7);
 
-    uint64_t bitmap1 = (1ull << 25) - 1; // 5 * 5
-    uint64_t bitmap2 = (1ull << 12) - 1;  // 3 * 4
-    uint64_t bitmap3 = (1ull << 20) - 1; // 5 * 4
-    uint64_t bitmap4 = (1ull << 21) - 1; // 3 * 7
+    u64 bitmap1 = (1ull << 25) - 1; // 5 * 5
+    u64 bitmap2 = (1ull << 12) - 1;  // 3 * 4
+    u64 bitmap3 = (1ull << 20) - 1; // 5 * 4
+    u64 bitmap4 = (1ull << 21) - 1; // 3 * 7
 
     REQUIRE(p1.bitmap_is_full(bitmap1));
     REQUIRE(p2.bitmap_is_full(bitmap2));
@@ -57,10 +57,10 @@ TEST_CASE("bitmap_is_full necessary bits not set", "[puzzle][board]")
     puzzle p3(5, 4);
     puzzle p4(3, 7);
 
-    uint64_t bitmap1 = (1ull << 25) - 1; // 5 * 5
-    uint64_t bitmap2 = (1ull << 12) - 1;  // 3 * 4
-    uint64_t bitmap3 = (1ull << 20) - 1; // 5 * 4
-    uint64_t bitmap4 = (1ull << 21) - 1; // 3 * 7
+    u64 bitmap1 = (1ull << 25) - 1; // 5 * 5
+    u64 bitmap2 = (1ull << 12) - 1;  // 3 * 4
+    u64 bitmap3 = (1ull << 20) - 1; // 5 * 4
+    u64 bitmap4 = (1ull << 21) - 1; // 3 * 7
 
     bitmap1 &= ~(1ull << 12);
     bitmap2 &= ~(1ull << 6);

test/bitmap_find_first_empty.cpp

@@ -5,7 +5,7 @@
 #include <catch2/catch_test_macros.hpp>
 #include <catch2/generators/catch_generators.hpp>
 
-static auto board_mask(const int w, const int h) -> uint64_t
+static auto board_mask(const int w, const int h) -> u64
 {
     const int cells = w * h;
     if (cells == 64) {
@@ -31,7 +31,7 @@ TEST_CASE("Full board detection respects width*height only", "[puzzle][board]")
 
     puzzle p(w, h);
 
-    uint64_t mask = board_mask(w, h);
+    u64 mask = board_mask(w, h);
 
     int x = -1, y = -1;
 
@@ -51,7 +51,7 @@ TEST_CASE("Single empty cell at various positions", "[puzzle][board]")
 
     auto empty_index = GENERATE_COPY(values<int>({ 0, cells / 2, cells - 1}));
 
-    uint64_t bitmap = board_mask(w, h);
+    u64 bitmap = board_mask(w, h);
     bitmap &= ~(1ULL << empty_index);
 
     int x = -1, y = -1;
@@ -65,10 +65,10 @@ TEST_CASE("Bits outside board are ignored", "[puzzle][board]")
 {
     puzzle p(3, 3); // 9 cells
 
-    uint64_t mask = board_mask(3, 3);
+    u64 mask = board_mask(3, 3);
 
     // Board is full
-    uint64_t bitmap = mask;
+    u64 bitmap = mask;
 
     // Set extra bits outside 9 cells
     bitmap |= (1ULL << 20);
@@ -83,7 +83,7 @@ TEST_CASE("First empty found in forward search branch", "[puzzle][branch]")
     puzzle p(4, 4); // 16 cells
 
     // Only MSB (within board) set
-    uint64_t bitmap = (1ULL << 15);
+    u64 bitmap = (1ULL << 15);
 
     int x = -1, y = -1;
     REQUIRE(p.bitmap_find_first_empty(bitmap, x, y));
@@ -97,7 +97,7 @@ TEST_CASE("Backward search branch finds gap before MSB cluster", "[puzzle][branc
     puzzle p(4, 4); // 16 cells
 
     // Set bits 15,14,13 but leave 12 empty
-    uint64_t bitmap = (1ULL << 15) | (1ULL << 14) | (1ULL << 13);
+    u64 bitmap = (1ULL << 15) | (1ULL << 14) | (1ULL << 13);
 
     int x = -1, y = -1;
     REQUIRE(p.bitmap_find_first_empty(bitmap, x, y));
@@ -112,7 +112,7 @@ TEST_CASE("Rectangular board coordinate mapping", "[puzzle][rect]")
 
     int empty_index = 11;
 
-    uint64_t bitmap = board_mask(5, 3);
+    u64 bitmap = board_mask(5, 3);
     bitmap &= ~(1ULL << empty_index);
 
     int x = -1, y = -1;
@@ -126,7 +126,7 @@ TEST_CASE("Non-64-sized board near limit", "[puzzle][limit]")
 {
     puzzle p(7, 8); // 56 cells
 
-    uint64_t mask = board_mask(7, 8);
+    u64 mask = board_mask(7, 8);
 
     // Full board should return false
     int x = -1, y = -1;
@@ -142,7 +142,7 @@ TEST_CASE("Non-64-sized board near limit", "[puzzle][limit]")
 }
 
 // --- Oracle: find first zero bit inside board ---
-static auto oracle_find_first_empty(uint64_t bitmap, int w, int h, int& x, int& y) -> bool
+static auto oracle_find_first_empty(u64 bitmap, int w, int h, int& x, int& y) -> bool
 {
     int cells = w * h;
 
@@ -164,13 +164,13 @@ TEST_CASE("Oracle validation across board sizes 3x3 to 8x8", "[puzzle][oracle]")
 
     puzzle p(w, h);
 
-    uint64_t mask = board_mask(w, h);
+    u64 mask = board_mask(w, h);
 
     std::mt19937_64 rng(12345);
-    std::uniform_int_distribution<uint64_t> dist(0, UINT64_MAX);
+    std::uniform_int_distribution<u64> dist(0, UINT64_MAX);
 
     for (int iteration = 0; iteration < 200; ++iteration) {
-        uint64_t bitmap = dist(rng);
+        u64 bitmap = dist(rng);
 
         int ox = -1, oy = -1;
         bool oracle_result = oracle_find_first_empty(bitmap, w, h, ox, oy);
@@ -191,7 +191,7 @@ TEST_CASE("Bits set outside board only behaves as empty board", "[puzzle][outsid
 {
     puzzle p(3, 3); // 9 cells
 
-    uint64_t bitmap = (1ULL << 40) | (1ULL << 63);
+    u64 bitmap = (1ULL << 40) | (1ULL << 63);
 
     int x = -1, y = -1;
     REQUIRE(p.bitmap_find_first_empty(bitmap, x, y));
@@ -207,7 +207,7 @@ TEST_CASE("Last valid cell empty stresses upper bound", "[puzzle][boundary]")
     puzzle p(w, h);
 
     int cells = w * h;
-    uint64_t bitmap = board_mask(w, h);
+    u64 bitmap = board_mask(w, h);
 
     // Clear last valid bit
     bitmap &= ~(1ULL << (cells - 1));
@@ -235,7 +235,7 @@ TEST_CASE("Board sizes between 33 and 63 cells", "[puzzle][midrange]")
     for (int index : {31, 32, cells - 2}) {
         if (index >= cells) continue;
 
-        uint64_t bitmap = board_mask(w, h);
+        u64 bitmap = board_mask(w, h);
         bitmap &= ~(1ULL << index);
 
         int x = -1, y = -1;
@@ -250,7 +250,7 @@ TEST_CASE("Multiple holes choose lowest index", "[puzzle][multiple]")
 {
     puzzle p(5, 5);
 
-    uint64_t bitmap = board_mask(5, 5);
+    u64 bitmap = board_mask(5, 5);
 
     // Clear several positions
     bitmap &= ~(1ULL << 3);

test/bits.cpp

@@ -2,7 +2,7 @@
 #include <catch2/catch_template_test_macros.hpp>
 #include <cstdint>
 
-#include "util.hpp"
+#include "bits.hpp"
 
 // =============================================================================
 // Catch2
@@ -31,8 +31,8 @@
 // 4. TEMPLATE_TEST_CASE(name, tags, Type1, Type2, ...)
 //    A parameterised test that is instantiated once for each type listed.
 //    Inside the test body, the alias `TestType` refers to the current type.
-//    This avoids duplicating identical logic for uint8_t, uint16_t, uint32_t,
-//    and uint64_t. Catch2 automatically appends the type name to the test name
+//    This avoids duplicating identical logic for u8, u16, u32,
+//    and u64. Catch2 automatically appends the type name to the test name
 //    in the output so you can see which instantiation failed.
 //
 // 5. Tags (e.g. "[create_mask]", "[round-trip]")
@@ -49,7 +49,7 @@
 // ---------------------------------------------------------------------------
 
 TEMPLATE_TEST_CASE("create_mask produces correct masks", "[create_mask]",
-                   uint8_t, uint16_t, uint32_t, uint64_t)
+                   u8, u16, u32, u64)
 {
     SECTION("single bit mask at bit 0") {
         auto m = create_mask<TestType>(0, 0);
@@ -72,16 +72,16 @@ TEMPLATE_TEST_CASE("create_mask produces correct masks", "[create_mask]",
     }
 
     SECTION("full-width mask returns all ones") {
-        constexpr uint8_t last = sizeof(TestType) * 8 - 1;
+        constexpr u8 last = sizeof(TestType) * 8 - 1;
         auto m = create_mask<TestType>(0, last);
         REQUIRE(m == static_cast<TestType>(~TestType{0}));
     }
 }
 
 TEST_CASE("create_mask 32-bit specific cases", "[create_mask]") {
-    REQUIRE(create_mask<uint32_t>(0, 15)  == 0x0000FFFF);
-    REQUIRE(create_mask<uint32_t>(0, 31)  == 0xFFFFFFFF);
-    REQUIRE(create_mask<uint32_t>(16, 31) == 0xFFFF0000);
+    REQUIRE(create_mask<u32>(0, 15)  == 0x0000FFFF);
+    REQUIRE(create_mask<u32>(0, 31)  == 0xFFFFFFFF);
+    REQUIRE(create_mask<u32>(16, 31) == 0xFFFF0000);
 }
 
 // ---------------------------------------------------------------------------
@@ -89,11 +89,11 @@ TEST_CASE("create_mask 32-bit specific cases", "[create_mask]") {
 // ---------------------------------------------------------------------------
 
 TEMPLATE_TEST_CASE("clear_bits zeroes the specified range", "[clear_bits]",
-                   uint8_t, uint16_t, uint32_t, uint64_t)
+                   u8, u16, u32, u64)
 {
     SECTION("clear all bits") {
         TestType val = static_cast<TestType>(~TestType{0});
-        constexpr uint8_t last = sizeof(TestType) * 8 - 1;
+        constexpr u8 last = sizeof(TestType) * 8 - 1;
         clear_bits(val, 0, last);
         REQUIRE(val == TestType{0});
     }
@@ -128,55 +128,55 @@ TEMPLATE_TEST_CASE("clear_bits zeroes the specified range", "[clear_bits]",
 // ---------------------------------------------------------------------------
 
 TEMPLATE_TEST_CASE("set_bits writes value into the specified range", "[set_bits]",
-                   uint8_t, uint16_t, uint32_t, uint64_t)
+                   u8, u16, u32, u64)
 {
     SECTION("set lower nibble on zero") {
         TestType val = TestType{0};
-        set_bits(val, uint8_t{0}, uint8_t{3}, static_cast<TestType>(0xA));
+        set_bits(val, u8{0}, u8{3}, static_cast<TestType>(0xA));
         REQUIRE(val == static_cast<TestType>(0x0A));
     }
 
     SECTION("set upper nibble on zero") {
         TestType val = TestType{0};
-        set_bits(val, uint8_t{4}, uint8_t{7}, static_cast<TestType>(0xB));
+        set_bits(val, u8{4}, u8{7}, static_cast<TestType>(0xB));
         REQUIRE(val == static_cast<TestType>(0xB0));
     }
 
     SECTION("set_bits replaces existing bits") {
         TestType val = static_cast<TestType>(0xFF);
-        set_bits(val, uint8_t{0}, uint8_t{3}, static_cast<TestType>(0x5));
+        set_bits(val, u8{0}, u8{3}, static_cast<TestType>(0x5));
         REQUIRE(val == static_cast<TestType>(0xF5));
     }
 
     SECTION("set single bit to 1") {
         TestType val = TestType{0};
-        set_bits(val, uint8_t{3}, uint8_t{3}, static_cast<TestType>(1));
+        set_bits(val, u8{3}, u8{3}, static_cast<TestType>(1));
         REQUIRE(val == static_cast<TestType>(0x08));
     }
 
     SECTION("set single bit to 0") {
         TestType val = static_cast<TestType>(0xFF);
-        set_bits(val, uint8_t{3}, uint8_t{3}, static_cast<TestType>(0));
+        set_bits(val, u8{3}, u8{3}, static_cast<TestType>(0));
         REQUIRE(val == static_cast<TestType>(0xF7));
     }
 
     SECTION("setting value 0 clears the range") {
         TestType val = static_cast<TestType>(0xFF);
-        set_bits(val, uint8_t{0}, uint8_t{7}, static_cast<TestType>(0));
+        set_bits(val, u8{0}, u8{7}, static_cast<TestType>(0));
         REQUIRE(val == TestType{0});
     }
 }
 
 TEST_CASE("set_bits with different value type (U != T)", "[set_bits]") {
-    uint32_t val = 0;
-    constexpr uint8_t small_val = 0x3F;
-    set_bits(val, uint8_t{8}, uint8_t{13}, small_val);
-    REQUIRE(val == (uint32_t{0x3F} << 8));
+    u32 val = 0;
+    constexpr u8 small_val = 0x3F;
+    set_bits(val, u8{8}, u8{13}, small_val);
+    REQUIRE(val == (u32{0x3F} << 8));
 }
 
 TEST_CASE("set_bits preserves surrounding bits in 32-bit", "[set_bits]") {
-    uint32_t val = 0xDEADBEEF;
-    set_bits(val, uint8_t{8}, uint8_t{15}, uint32_t{0x42});
+    u32 val = 0xDEADBEEF;
+    set_bits(val, u8{8}, u8{15}, u32{0x42});
     REQUIRE(val == 0xDEAD42EF);
 }
 
@@ -185,53 +185,53 @@ TEST_CASE("set_bits preserves surrounding bits in 32-bit", "[set_bits]") {
 // ---------------------------------------------------------------------------
 
 TEMPLATE_TEST_CASE("get_bits extracts the specified range", "[get_bits]",
-                   uint8_t, uint16_t, uint32_t, uint64_t)
+                   u8, u16, u32, u64)
 {
     SECTION("get lower nibble") {
         TestType val = static_cast<TestType>(0xAB);
-        auto result = get_bits(val, uint8_t{0}, uint8_t{3});
+        auto result = get_bits(val, u8{0}, u8{3});
         REQUIRE(result == TestType{0xB});
     }
 
     SECTION("get upper nibble") {
         TestType val = static_cast<TestType>(0xAB);
-        auto result = get_bits(val, uint8_t{4}, uint8_t{7});
+        auto result = get_bits(val, u8{4}, u8{7});
         REQUIRE(result == TestType{0xA});
     }
 
     SECTION("get single bit that is set") {
         TestType val = static_cast<TestType>(0x08);
-        auto result = get_bits(val, uint8_t{3}, uint8_t{3});
+        auto result = get_bits(val, u8{3}, u8{3});
         REQUIRE(result == TestType{1});
     }
 
     SECTION("get single bit that is clear") {
         TestType val = static_cast<TestType>(0xF7);
-        auto result = get_bits(val, uint8_t{3}, uint8_t{3});
+        auto result = get_bits(val, u8{3}, u8{3});
         REQUIRE(result == TestType{0});
     }
 
     SECTION("get all bits") {
         TestType val = static_cast<TestType>(~TestType{0});
-        constexpr uint8_t last = sizeof(TestType) * 8 - 1;
-        auto result = get_bits(val, uint8_t{0}, last);
+        constexpr u8 last = sizeof(TestType) * 8 - 1;
+        auto result = get_bits(val, u8{0}, last);
         REQUIRE(result == val);
     }
 
     SECTION("get from zero returns zero") {
         TestType val = TestType{0};
-        auto result = get_bits(val, uint8_t{0}, uint8_t{7});
+        auto result = get_bits(val, u8{0}, u8{7});
         REQUIRE(result == TestType{0});
     }
 }
 
 TEST_CASE("get_bits 32-bit specific extractions", "[get_bits]") {
-    constexpr uint32_t val = 0xDEADBEEF;
+    constexpr u32 val = 0xDEADBEEF;
 
-    REQUIRE(get_bits(val, uint8_t{0},  uint8_t{7})  == 0xEF);
-    REQUIRE(get_bits(val, uint8_t{8},  uint8_t{15}) == 0xBE);
-    REQUIRE(get_bits(val, uint8_t{16}, uint8_t{23}) == 0xAD);
-    REQUIRE(get_bits(val, uint8_t{24}, uint8_t{31}) == 0xDE);
+    REQUIRE(get_bits(val, u8{0},  u8{7})  == 0xEF);
+    REQUIRE(get_bits(val, u8{8},  u8{15}) == 0xBE);
+    REQUIRE(get_bits(val, u8{16}, u8{23}) == 0xAD);
+    REQUIRE(get_bits(val, u8{24}, u8{31}) == 0xDE);
 }
 
 // ---------------------------------------------------------------------------
@@ -239,29 +239,29 @@ TEST_CASE("get_bits 32-bit specific extractions", "[get_bits]") {
 // ---------------------------------------------------------------------------
 
 TEST_CASE("set_bits then get_bits round-trips correctly", "[round-trip]") {
-    uint32_t reg = 0;
+    u32 reg = 0;
 
-    set_bits(reg, uint8_t{4}, uint8_t{11}, uint32_t{0xAB});
-    REQUIRE(get_bits(reg, uint8_t{4}, uint8_t{11}) == 0xAB);
+    set_bits(reg, u8{4}, u8{11}, u32{0xAB});
+    REQUIRE(get_bits(reg, u8{4}, u8{11}) == 0xAB);
 
-    REQUIRE(get_bits(reg, uint8_t{0},  uint8_t{3})  == 0x0);
-    REQUIRE(get_bits(reg, uint8_t{12}, uint8_t{31}) == 0x0);
+    REQUIRE(get_bits(reg, u8{0},  u8{3})  == 0x0);
+    REQUIRE(get_bits(reg, u8{12}, u8{31}) == 0x0);
 }
 
 TEST_CASE("multiple set_bits on different ranges", "[round-trip]") {
-    uint32_t reg = 0;
+    u32 reg = 0;
 
-    set_bits(reg, uint8_t{0},  uint8_t{7},  uint32_t{0x01});
-    set_bits(reg, uint8_t{8},  uint8_t{15}, uint32_t{0x02});
-    set_bits(reg, uint8_t{16}, uint8_t{23}, uint32_t{0x03});
-    set_bits(reg, uint8_t{24}, uint8_t{31}, uint32_t{0x04});
+    set_bits(reg, u8{0},  u8{7},  u32{0x01});
+    set_bits(reg, u8{8},  u8{15}, u32{0x02});
+    set_bits(reg, u8{16}, u8{23}, u32{0x03});
+    set_bits(reg, u8{24}, u8{31}, u32{0x04});
 
     REQUIRE(reg == 0x04030201);
 }
 
 TEST_CASE("64-bit round-trip", "[round-trip]") {
-    uint64_t reg = 0;
-    set_bits(reg, uint8_t{32}, uint8_t{63}, uint64_t{0xCAFEBABE});
-    REQUIRE(get_bits(reg, uint8_t{32}, uint8_t{63}) == uint64_t{0xCAFEBABE});
-    REQUIRE(get_bits(reg, uint8_t{0},  uint8_t{31}) == uint64_t{0});
+    u64 reg = 0;
+    set_bits(reg, u8{32}, u8{63}, u64{0xCAFEBABE});
+    REQUIRE(get_bits(reg, u8{32}, u8{63}) == u64{0xCAFEBABE});
+    REQUIRE(get_bits(reg, u8{0},  u8{31}) == u64{0});
 }

test/puzzle.cpp

@@ -12,7 +12,7 @@
 
 TEST_CASE("Block creation and field access", "[block]")
 {
-    puzzle::block b(1, 2, 3, 4, true, false);
+    block b(1, 2, 3, 4, true, false);
 
     CHECK(b.get_x() == 1);
     CHECK(b.get_y() == 2);
@@ -25,16 +25,16 @@ TEST_CASE("Block creation and field access", "[block]")
 
 TEST_CASE("Block invalid by default", "[block]")
 {
-    puzzle::block b;
+    block b;
     CHECK_FALSE(b.valid());
 }
 
 TEST_CASE("Block comparison ordering", "[block]")
 {
     // Row-major: (0,0) < (1,0) < (0,1)
-    puzzle::block a(0, 0, 1, 1);
-    puzzle::block b(1, 0, 1, 1);
-    puzzle::block c(0, 1, 1, 1);
+    block a(0, 0, 1, 1);
+    block b(1, 0, 1, 1);
+    block c(0, 1, 1, 1);
 
     CHECK(a < b);
     CHECK(b < c);
@@ -43,7 +43,7 @@ TEST_CASE("Block comparison ordering", "[block]")
 
 TEST_CASE("Block principal_dirs for horizontal block", "[block]")
 {
-    puzzle::block b(0, 0, 3, 1); // wider than tall
+    block b(0, 0, 3, 1); // wider than tall
     CHECK((b.principal_dirs() & eas));
     CHECK((b.principal_dirs() & wes));
     CHECK_FALSE((b.principal_dirs() & nor));
@@ -52,7 +52,7 @@ TEST_CASE("Block principal_dirs for horizontal block", "[block]")
 
 TEST_CASE("Block principal_dirs for vertical block", "[block]")
 {
-    puzzle::block b(0, 0, 1, 3); // taller than wide
+    block b(0, 0, 1, 3); // taller than wide
     CHECK((b.principal_dirs() & nor));
     CHECK((b.principal_dirs() & sou));
     CHECK_FALSE((b.principal_dirs() & eas));
@@ -61,7 +61,7 @@ TEST_CASE("Block principal_dirs for vertical block", "[block]")
 
 TEST_CASE("Block principal_dirs for square block", "[block]")
 {
-    puzzle::block b(0, 0, 2, 2);
+    block b(0, 0, 2, 2);
     CHECK((b.principal_dirs() & nor));
     CHECK((b.principal_dirs() & sou));
     CHECK((b.principal_dirs() & eas));
@@ -70,7 +70,7 @@ TEST_CASE("Block principal_dirs for square block", "[block]")
 
 TEST_CASE("Block covers", "[block]")
 {
-    puzzle::block b(1, 2, 3, 2);
+    block b(1, 2, 3, 2);
     // Covers (1,2) to (3,3)
     CHECK(b.covers(1, 2));
     CHECK(b.covers(3, 3));
@@ -83,10 +83,10 @@ TEST_CASE("Block covers", "[block]")
 
 TEST_CASE("Block collides", "[block]")
 {
-    puzzle::block a(0, 0, 2, 2);
-    puzzle::block b(1, 1, 2, 2);
-    puzzle::block c(2, 2, 1, 1);
-    puzzle::block d(3, 0, 1, 1);
+    block a(0, 0, 2, 2);
+    block b(1, 1, 2, 2);
+    block c(2, 2, 1, 1);
+    block d(3, 0, 1, 1);
 
     CHECK(a.collides(b));
     CHECK(b.collides(a));
@@ -114,7 +114,7 @@ TEST_CASE("Puzzle creation and meta access", "[puzzle]")
 TEST_CASE("Puzzle add and remove block", "[puzzle]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    puzzle::block b(0, 0, 2, 1);
+    block b(0, 0, 2, 1);
 
     auto p2 = p.try_add_block(b);
     REQUIRE(p2.has_value());
@@ -127,8 +127,8 @@ TEST_CASE("Puzzle add and remove block", "[puzzle]")
 
 TEST_CASE("Puzzle meta roundtrip for all valid sizes", "[puzzle]")
 {
-    for (uint8_t w = 3; w <= 8; ++w) {
-        for (uint8_t h = 3; h <= 8; ++h) {
+    for (u8 w = 3; w <= 8; ++w) {
+        for (u8 h = 3; h <= 8; ++h) {
             puzzle p(w, h, 0, 0, true, false);
             CHECK(p.get_width() == w);
             CHECK(p.get_height() == h);
@@ -151,11 +151,11 @@ TEST_CASE("Puzzle block_count", "[puzzle]")
     puzzle p(4, 4, 0, 0, false, false);
     CHECK(p.block_count() == 0);
 
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 1));
+    auto p2 = p.try_add_block(block(0, 0, 1, 1));
     REQUIRE(p2);
     CHECK(p2->block_count() == 1);
 
-    auto p3 = p2->try_add_block(puzzle::block(2, 2, 1, 1));
+    auto p3 = p2->try_add_block(block(2, 2, 1, 1));
     REQUIRE(p3);
     CHECK(p3->block_count() == 2);
 }
@@ -163,18 +163,18 @@ TEST_CASE("Puzzle block_count", "[puzzle]")
 TEST_CASE("Puzzle cannot add overlapping block", "[puzzle]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 2));
+    auto p2 = p.try_add_block(block(0, 0, 2, 2));
     REQUIRE(p2);
 
     // Overlapping
-    auto p3 = p2->try_add_block(puzzle::block(1, 1, 2, 2));
+    auto p3 = p2->try_add_block(block(1, 1, 2, 2));
     CHECK_FALSE(p3.has_value());
 }
 
 TEST_CASE("Puzzle cannot add block outside board", "[puzzle]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(3, 3, 2, 2));
+    auto p2 = p.try_add_block(block(3, 3, 2, 2));
     CHECK_FALSE(p2.has_value());
 }
 
@@ -184,17 +184,17 @@ TEST_CASE("Puzzle cannot add block outside board", "[puzzle]")
 
 TEST_CASE("bitmap_set_bit and bitmap_get_bit roundtrip", "[bitmap]")
 {
-    for (uint8_t w = 3; w <= 8; ++w) {
-        uint64_t bm = 0;
+    for (u8 w = 3; w <= 8; ++w) {
+        u64 bm = 0;
         // Set every cell on a w x w board
-        for (uint8_t y = 0; y < w; ++y) {
-            for (uint8_t x = 0; x < w; ++x) {
+        for (u8 y = 0; y < w; ++y) {
+            for (u8 x = 0; x < w; ++x) {
                 puzzle::bitmap_set_bit(bm, w, x, y);
             }
         }
         // Verify all set
-        for (uint8_t y = 0; y < w; ++y) {
-            for (uint8_t x = 0; x < w; ++x) {
+        for (u8 y = 0; y < w; ++y) {
+            for (u8 x = 0; x < w; ++x) {
                 CHECK(puzzle::bitmap_get_bit(bm, w, x, y));
             }
         }
@@ -203,12 +203,12 @@ TEST_CASE("bitmap_set_bit and bitmap_get_bit roundtrip", "[bitmap]")
 
 TEST_CASE("bitmap_set_bit sets only the target bit", "[bitmap]")
 {
-    for (uint8_t w = 3; w <= 8; ++w) {
-        for (uint8_t y = 0; y < w; ++y) {
-            for (uint8_t x = 0; x < w; ++x) {
-                uint64_t bm = 0;
+    for (u8 w = 3; w <= 8; ++w) {
+        for (u8 y = 0; y < w; ++y) {
+            for (u8 x = 0; x < w; ++x) {
+                u64 bm = 0;
                 puzzle::bitmap_set_bit(bm, w, x, y);
-                CHECK(bm == (uint64_t(1) << (y * w + x)));
+                CHECK(bm == (u64(1) << (y * w + x)));
             }
         }
     }
@@ -216,11 +216,11 @@ TEST_CASE("bitmap_set_bit sets only the target bit", "[bitmap]")
 
 TEST_CASE("bitmap_clear_bit clears only the target bit", "[bitmap]")
 {
-    uint8_t w = 6;
-    uint64_t bm = 0;
+    u8 w = 6;
+    u64 bm = 0;
     puzzle::bitmap_set_bit(bm, w, 2, 3);
     puzzle::bitmap_set_bit(bm, w, 4, 1);
-    uint64_t before = bm;
+    u64 before = bm;
 
     puzzle::bitmap_clear_bit(bm, w, 2, 3);
     CHECK_FALSE(puzzle::bitmap_get_bit(bm, w, 2, 3));
@@ -230,10 +230,10 @@ TEST_CASE("bitmap_clear_bit clears only the target bit", "[bitmap]")
 TEST_CASE("blocks_bitmap for single block", "[bitmap]")
 {
     puzzle p(6, 6, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 2, 3, 1));
+    auto p2 = p.try_add_block(block(1, 2, 3, 1));
     REQUIRE(p2);
 
-    uint64_t bm = p2->blocks_bitmap();
+    u64 bm = p2->blocks_bitmap();
     // Block at (1,2) width 3 height 1 on a 6-wide board
     // Bits: row 2, cols 1,2,3 -> positions 2*6+1=13, 14, 15
     CHECK(puzzle::bitmap_get_bit(bm, 6, 1, 2));
@@ -248,10 +248,10 @@ TEST_CASE("blocks_bitmap for single block", "[bitmap]")
 TEST_CASE("blocks_bitmap for 2x2 block", "[bitmap]")
 {
     puzzle p(5, 5, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 2, 2));
+    auto p2 = p.try_add_block(block(1, 1, 2, 2));
     REQUIRE(p2);
 
-    uint64_t bm = p2->blocks_bitmap();
+    u64 bm = p2->blocks_bitmap();
     CHECK(puzzle::bitmap_get_bit(bm, 5, 1, 1));
     CHECK(puzzle::bitmap_get_bit(bm, 5, 2, 1));
     CHECK(puzzle::bitmap_get_bit(bm, 5, 1, 2));
@@ -264,13 +264,13 @@ TEST_CASE("blocks_bitmap_h only includes horizontal/square blocks", "[bitmap]")
 {
     puzzle p(6, 6, 0, 0, true, false);
     // Horizontal block (2x1) -> principal_dirs has eas
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 1));
+    auto p2 = p.try_add_block(block(0, 0, 2, 1));
     REQUIRE(p2);
     // Vertical block (1x2) -> principal_dirs has sou, not eas
-    auto p3 = p2->try_add_block(puzzle::block(4, 0, 1, 2));
+    auto p3 = p2->try_add_block(block(4, 0, 1, 2));
     REQUIRE(p3);
 
-    uint64_t bm_h = p3->blocks_bitmap_h();
+    u64 bm_h = p3->blocks_bitmap_h();
     // Horizontal block should be in bitmap_h
     CHECK(puzzle::bitmap_get_bit(bm_h, 6, 0, 0));
     CHECK(puzzle::bitmap_get_bit(bm_h, 6, 1, 0));
@@ -283,13 +283,13 @@ TEST_CASE("blocks_bitmap_v only includes vertical/square blocks", "[bitmap]")
 {
     puzzle p(6, 6, 0, 0, true, false);
     // Vertical block (1x2)
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 2));
+    auto p2 = p.try_add_block(block(0, 0, 1, 2));
     REQUIRE(p2);
     // Horizontal block (2x1)
-    auto p3 = p2->try_add_block(puzzle::block(4, 0, 2, 1));
+    auto p3 = p2->try_add_block(block(4, 0, 2, 1));
     REQUIRE(p3);
 
-    uint64_t bm_v = p3->blocks_bitmap_v();
+    u64 bm_v = p3->blocks_bitmap_v();
     // Vertical block should be in bitmap_v
     CHECK(puzzle::bitmap_get_bit(bm_v, 6, 0, 0));
     CHECK(puzzle::bitmap_get_bit(bm_v, 6, 0, 1));
@@ -301,11 +301,11 @@ TEST_CASE("blocks_bitmap_v only includes vertical/square blocks", "[bitmap]")
 TEST_CASE("blocks_bitmap_h and blocks_bitmap_v both include square blocks", "[bitmap]")
 {
     puzzle p(6, 6, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 2, 2));
+    auto p2 = p.try_add_block(block(1, 1, 2, 2));
     REQUIRE(p2);
 
-    uint64_t bm_h = p2->blocks_bitmap_h();
-    uint64_t bm_v = p2->blocks_bitmap_v();
+    u64 bm_h = p2->blocks_bitmap_h();
+    u64 bm_v = p2->blocks_bitmap_v();
 
     // Square block should appear in both
     CHECK(puzzle::bitmap_get_bit(bm_h, 6, 1, 1));
@@ -320,12 +320,12 @@ TEST_CASE("bitmap_newly_occupied north - single vertical block with space above"
 {
     // 4x4 board, vertical 1x2 block at (1,2)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 2, 1, 2));
+    auto p2 = p.try_add_block(block(1, 2, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, vert, nor);
 
@@ -346,12 +346,12 @@ TEST_CASE("bitmap_newly_occupied south - single vertical block with space below"
 {
     // 4x4 board, vertical 1x2 block at (1,0)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 0, 1, 2));
+    auto p2 = p.try_add_block(block(1, 0, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, vert, sou);
 
@@ -366,12 +366,12 @@ TEST_CASE("bitmap_newly_occupied east - single horizontal block with space right
 {
     // 5x4 board, horizontal 2x1 block at (0,1)
     puzzle p(5, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 1, 2, 1));
+    auto p2 = p.try_add_block(block(0, 1, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, horiz, eas);
 
@@ -386,12 +386,12 @@ TEST_CASE("bitmap_newly_occupied west - single horizontal block with space left"
 {
     // 5x4 board, horizontal 2x1 block at (2,1)
     puzzle p(5, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 1, 2, 1));
+    auto p2 = p.try_add_block(block(2, 1, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, horiz, wes);
 
@@ -406,12 +406,12 @@ TEST_CASE("bitmap_newly_occupied does not wrap east across rows", "[bitmap_move]
 {
     // 4x4 board, horizontal 2x1 block at (2,0) - rightmost position
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 0, 2, 1));
+    auto p2 = p.try_add_block(block(2, 0, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, horiz, eas);
 
@@ -424,12 +424,12 @@ TEST_CASE("bitmap_newly_occupied does not wrap west across rows", "[bitmap_move]
 {
     // 4x4 board, horizontal 2x1 block at (0,1)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 1, 2, 1));
+    auto p2 = p.try_add_block(block(0, 1, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
-    uint64_t bm = combined;
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
+    u64 bm = combined;
 
     p2->bitmap_newly_occupied_after_move(bm, horiz, wes);
 
@@ -441,14 +441,14 @@ TEST_CASE("bitmap_newly_occupied blocked by another block", "[bitmap_move]")
 {
     // 6x4 board, two horizontal blocks side by side
     puzzle p(6, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 1));
+    auto p2 = p.try_add_block(block(0, 0, 2, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(2, 0, 2, 1));
+    auto p3 = p2->try_add_block(block(2, 0, 2, 1));
     REQUIRE(p3);
 
-    uint64_t combined = p3->blocks_bitmap();
-    uint64_t horiz = p3->blocks_bitmap_h();
-    uint64_t bm = combined;
+    u64 combined = p3->blocks_bitmap();
+    u64 horiz = p3->blocks_bitmap_h();
+    u64 bm = combined;
 
     p3->bitmap_newly_occupied_after_move(bm, horiz, eas);
 
@@ -468,11 +468,11 @@ TEST_CASE("restricted_bitmap_get_moves north - returns correct source cell", "[g
 {
     // 4x4 board, vertical 1x2 block at (1,2)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 2, 1, 2));
+    auto p2 = p.try_add_block(block(1, 2, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, vert, nor);
 
@@ -487,11 +487,11 @@ TEST_CASE("restricted_bitmap_get_moves south - returns correct source cell", "[g
 {
     // 4x4 board, vertical 1x2 block at (1,0)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 0, 1, 2));
+    auto p2 = p.try_add_block(block(1, 0, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, vert, sou);
 
@@ -511,11 +511,11 @@ TEST_CASE("restricted_bitmap_get_moves east - returns correct source cell", "[ge
 {
     // 5x4 board, horizontal 2x1 block at (0,1)
     puzzle p(5, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 1, 2, 1));
+    auto p2 = p.try_add_block(block(0, 1, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, horiz, eas);
 
@@ -530,11 +530,11 @@ TEST_CASE("restricted_bitmap_get_moves west - returns correct source cell", "[ge
 {
     // 5x4 board, horizontal 2x1 block at (2,1)
     puzzle p(5, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 1, 2, 1));
+    auto p2 = p.try_add_block(block(2, 1, 2, 1));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t horiz = p2->blocks_bitmap_h();
+    u64 combined = p2->blocks_bitmap();
+    u64 horiz = p2->blocks_bitmap_h();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, horiz, wes);
 
@@ -549,11 +549,11 @@ TEST_CASE("restricted_bitmap_get_moves returns empty when blocked", "[get_moves]
 {
     // 4x4 board, vertical 1x2 block at (1,0) - can't move north (at top edge)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 0, 1, 2));
+    auto p2 = p.try_add_block(block(1, 0, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, vert, nor);
 
@@ -569,11 +569,11 @@ TEST_CASE("restricted_bitmap_get_moves with width 6 board", "[get_moves]")
 {
     // 6x6 board, vertical 1x2 block at (2,3)
     puzzle p(6, 6, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 3, 1, 2));
+    auto p2 = p.try_add_block(block(2, 3, 1, 2));
     REQUIRE(p2);
 
-    uint64_t combined = p2->blocks_bitmap();
-    uint64_t vert = p2->blocks_bitmap_v();
+    u64 combined = p2->blocks_bitmap();
+    u64 vert = p2->blocks_bitmap_v();
 
     auto moves = p2->restricted_bitmap_get_moves(combined, vert, nor);
 
@@ -588,20 +588,20 @@ TEST_CASE("restricted_bitmap_get_moves multiple blocks can move", "[get_moves]")
 {
     // 6x6 board, two vertical blocks with space above
     puzzle p(6, 6, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 2, 1, 2));
+    auto p2 = p.try_add_block(block(1, 2, 1, 2));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(4, 3, 1, 2));
+    auto p3 = p2->try_add_block(block(4, 3, 1, 2));
     REQUIRE(p3);
 
-    uint64_t combined = p3->blocks_bitmap();
-    uint64_t vert = p3->blocks_bitmap_v();
+    u64 combined = p3->blocks_bitmap();
+    u64 vert = p3->blocks_bitmap_v();
 
     auto moves = p3->restricted_bitmap_get_moves(combined, vert, nor);
 
     // Both blocks can move north
     // Block 1: newly occupied (1,1), source (1,2) = 1+2*6 = 13
     // Block 2: newly occupied (4,2), source (4,3) = 4+3*6 = 22
-    std::set<uint8_t> move_set;
+    std::set<u8> move_set;
     for (int i = 0; i < puzzle::MAX_BLOCKS && moves[i] != 0xFF; ++i) {
         move_set.insert(moves[i]);
     }
@@ -617,7 +617,7 @@ TEST_CASE("restricted_bitmap_get_moves multiple blocks can move", "[get_moves]")
 TEST_CASE("try_move_block_at basic moves", "[move]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 1, 1));
+    auto p2 = p.try_add_block(block(1, 1, 1, 1));
     REQUIRE(p2);
 
     auto north = p2->try_move_block_at(1, 1, nor);
@@ -640,7 +640,7 @@ TEST_CASE("try_move_block_at basic moves", "[move]")
 TEST_CASE("try_move_block_at blocked by edge", "[move]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 1));
+    auto p2 = p.try_add_block(block(0, 0, 1, 1));
     REQUIRE(p2);
 
     CHECK_FALSE(p2->try_move_block_at(0, 0, nor).has_value());
@@ -650,9 +650,9 @@ TEST_CASE("try_move_block_at blocked by edge", "[move]")
 TEST_CASE("try_move_block_at blocked by collision", "[move]")
 {
     puzzle p(4, 4, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 1));
+    auto p2 = p.try_add_block(block(0, 0, 1, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(1, 0, 1, 1));
+    auto p3 = p2->try_add_block(block(1, 0, 1, 1));
     REQUIRE(p3);
 
     CHECK_FALSE(p3->try_move_block_at(0, 0, eas).has_value());
@@ -661,14 +661,14 @@ TEST_CASE("try_move_block_at blocked by collision", "[move]")
 TEST_CASE("try_move_block_at_fast matches try_move_block_at", "[move]")
 {
     puzzle p(5, 5, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 2, 1));
+    auto p2 = p.try_add_block(block(1, 1, 2, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(0, 3, 1, 2));
+    auto p3 = p2->try_add_block(block(0, 3, 1, 2));
     REQUIRE(p3);
 
-    uint64_t bm = p3->blocks_bitmap();
+    u64 bm = p3->blocks_bitmap();
 
-    for (const direction d : {nor, eas, sou, wes}) {
+    for (const dir d : {nor, eas, sou, wes}) {
         // Block 0 is the first in sorted order
         auto slow = p3->try_move_block_at(1, 1, d);
         auto fast = p3->try_move_block_at_fast(bm, 0, d);
@@ -686,7 +686,7 @@ TEST_CASE("try_move_block_at restricted mode respects principal dirs", "[move]")
 {
     puzzle p(5, 5, 0, 0, true, false); // restricted
     // Horizontal block 2x1 at (1,2)
-    auto p2 = p.try_add_block(puzzle::block(1, 2, 2, 1));
+    auto p2 = p.try_add_block(block(1, 2, 2, 1));
     REQUIRE(p2);
 
     // In restricted mode, horizontal block can only move east/west
@@ -760,16 +760,16 @@ static auto collect_adjacent_simple(const puzzle& p) -> std::set<size_t>
 // Helper: collect all adjacent states using for_each_adjacent_restricted
 static auto collect_adjacent_restricted(const puzzle& p) -> std::set<size_t>
 {
-    const uint8_t w = p.get_width();
-    std::array<uint8_t, 64> bitmap_block_indices{};
+    const u8 w = p.get_width();
+    std::array<u8, 64> bitmap_block_indices{};
     for (size_t i = 0; i < puzzle::MAX_BLOCKS; ++i) {
-        const puzzle::block b(p.repr_view().data()[i]);
+        const block b(p.repr_view().data()[i]);
         if (i >= static_cast<size_t>(p.block_count())) {
             break;
         }
         const auto [bx, by, bw, bh, bt, bi] = b.unpack_repr();
-        for (uint8_t x = bx; x < bx + bw; ++x) {
-            for (uint8_t y = by; y < by + bh; ++y) {
+        for (u8 x = bx; x < bx + bw; ++x) {
+            for (u8 y = by; y < by + bh; ++y) {
                 bitmap_block_indices[y * w + x] = i;
             }
         }
@@ -786,7 +786,7 @@ static auto collect_adjacent_restricted(const puzzle& p) -> std::set<size_t>
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - single block", "[cross_validate]")
 {
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 1, 1));
+    auto p2 = p.try_add_block(block(1, 1, 1, 1));
     REQUIRE(p2);
 
     auto simple = collect_adjacent_simple(*p2);
@@ -798,7 +798,7 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - single block
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - horizontal block", "[cross_validate]")
 {
     puzzle p(5, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 2, 1));
+    auto p2 = p.try_add_block(block(1, 1, 2, 1));
     REQUIRE(p2);
 
     auto simple = collect_adjacent_simple(*p2);
@@ -810,7 +810,7 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - horizontal b
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - vertical block", "[cross_validate]")
 {
     puzzle p(4, 5, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 1, 2));
+    auto p2 = p.try_add_block(block(1, 1, 1, 2));
     REQUIRE(p2);
 
     auto simple = collect_adjacent_simple(*p2);
@@ -822,9 +822,9 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - vertical blo
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - two blocks", "[cross_validate]")
 {
     puzzle p(5, 5, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 1)); // horizontal
+    auto p2 = p.try_add_block(block(0, 0, 2, 1)); // horizontal
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(4, 1, 1, 2)); // vertical
+    auto p3 = p2->try_add_block(block(4, 1, 1, 2)); // vertical
     REQUIRE(p3);
 
     auto simple = collect_adjacent_simple(*p3);
@@ -836,13 +836,13 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - two blocks",
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - crowded board", "[cross_validate]")
 {
     puzzle p(5, 5, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 1));
+    auto p2 = p.try_add_block(block(0, 0, 2, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(3, 0, 1, 2));
+    auto p3 = p2->try_add_block(block(3, 0, 1, 2));
     REQUIRE(p3);
-    auto p4 = p3->try_add_block(puzzle::block(0, 2, 1, 3));
+    auto p4 = p3->try_add_block(block(0, 2, 1, 3));
     REQUIRE(p4);
-    auto p5 = p4->try_add_block(puzzle::block(2, 3, 2, 1));
+    auto p5 = p4->try_add_block(block(2, 3, 2, 1));
     REQUIRE(p5);
 
     auto simple = collect_adjacent_simple(*p5);
@@ -854,13 +854,13 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - crowded boar
 TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - 6x6 board", "[cross_validate]")
 {
     puzzle p(6, 6, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 3, 1)); // horizontal
+    auto p2 = p.try_add_block(block(0, 0, 3, 1)); // horizontal
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(5, 0, 1, 2)); // vertical
+    auto p3 = p2->try_add_block(block(5, 0, 1, 2)); // vertical
     REQUIRE(p3);
-    auto p4 = p3->try_add_block(puzzle::block(2, 2, 1, 2)); // vertical
+    auto p4 = p3->try_add_block(block(2, 2, 1, 2)); // vertical
     REQUIRE(p4);
-    auto p5 = p4->try_add_block(puzzle::block(3, 2, 2, 1)); // horizontal
+    auto p5 = p4->try_add_block(block(3, 2, 2, 1)); // horizontal
     REQUIRE(p5);
 
     auto simple = collect_adjacent_simple(*p5);
@@ -889,7 +889,7 @@ TEST_CASE("for_each_adjacent_restricted matches for_each_adjacent - the main puz
 TEST_CASE("explore_state_space simple case - single 1x1 block", "[explore]")
 {
     puzzle p(3, 3, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 1, 1, 1));
+    auto p2 = p.try_add_block(block(1, 1, 1, 1));
     REQUIRE(p2);
 
     auto [states, transitions] = p2->explore_state_space();
@@ -901,7 +901,7 @@ TEST_CASE("explore_state_space simple case - single 1x1 block", "[explore]")
 TEST_CASE("explore_state_space restricted - vertical block", "[explore]")
 {
     puzzle p(3, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(1, 0, 1, 2));
+    auto p2 = p.try_add_block(block(1, 0, 1, 2));
     REQUIRE(p2);
 
     auto [states, transitions] = p2->explore_state_space();
@@ -914,7 +914,7 @@ TEST_CASE("explore_state_space restricted - vertical block", "[explore]")
 TEST_CASE("explore_state_space restricted - horizontal block", "[explore]")
 {
     puzzle p(4, 3, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 1, 2, 1));
+    auto p2 = p.try_add_block(block(0, 1, 2, 1));
     REQUIRE(p2);
 
     auto [states, transitions] = p2->explore_state_space();
@@ -927,7 +927,7 @@ TEST_CASE("explore_state_space restricted - horizontal block", "[explore]")
 TEST_CASE("explore_state_space restricted - square block", "[explore]")
 {
     puzzle p(3, 3, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 1));
+    auto p2 = p.try_add_block(block(0, 0, 1, 1));
     REQUIRE(p2);
 
     auto [states, transitions] = p2->explore_state_space();
@@ -939,7 +939,7 @@ TEST_CASE("explore_state_space restricted - square block", "[explore]")
 TEST_CASE("explore_state_space preserves board metadata", "[explore]")
 {
     puzzle p(5, 4, 2, 1, true, true);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 1, 1, true));
+    auto p2 = p.try_add_block(block(0, 0, 1, 1, true));
     REQUIRE(p2);
 
     auto [states, transitions] = p2->explore_state_space();
@@ -959,9 +959,9 @@ TEST_CASE("explore_state_space two blocks restricted", "[explore]")
     // 4x4 board, restricted
     // Horizontal 2x1 at (0,0) and vertical 1x2 at (3,0)
     puzzle p(4, 4, 0, 0, true, false);
-    auto p2 = p.try_add_block(puzzle::block(0, 0, 2, 1));
+    auto p2 = p.try_add_block(block(0, 0, 2, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(3, 0, 1, 2));
+    auto p3 = p2->try_add_block(block(3, 0, 1, 2));
     REQUIRE(p3);
 
     auto [states, transitions] = p3->explore_state_space();
@@ -988,15 +988,15 @@ TEST_CASE("explore_state_space two blocks restricted", "[explore]")
 
 TEST_CASE("East move does not wrap for various board widths", "[bitmap_move]")
 {
-    for (uint8_t w = 3; w <= 8; ++w) {
+    for (u8 w = 3; w <= 8; ++w) {
         puzzle p(w, 3, 0, 0, true, false);
         // Place a 1x1 block at rightmost column, row 0
-        auto p2 = p.try_add_block(puzzle::block(w - 1, 0, 1, 1));
+        auto p2 = p.try_add_block(block(w - 1, 0, 1, 1));
         REQUIRE(p2);
 
-        uint64_t combined = p2->blocks_bitmap();
-        uint64_t horiz = p2->blocks_bitmap_h(); // 1x1 is square, so in both h and v
-        uint64_t bm = combined;
+        u64 combined = p2->blocks_bitmap();
+        u64 horiz = p2->blocks_bitmap_h(); // 1x1 is square, so in both h and v
+        u64 bm = combined;
 
         p2->bitmap_newly_occupied_after_move(bm, horiz, eas);
 
@@ -1009,15 +1009,15 @@ TEST_CASE("East move does not wrap for various board widths", "[bitmap_move]")
 
 TEST_CASE("West move does not wrap for various board widths", "[bitmap_move]")
 {
-    for (uint8_t w = 3; w <= 8; ++w) {
+    for (u8 w = 3; w <= 8; ++w) {
         puzzle p(w, 3, 0, 0, true, false);
         // Place a 1x1 block at leftmost column, row 1
-        auto p2 = p.try_add_block(puzzle::block(0, 1, 1, 1));
+        auto p2 = p.try_add_block(block(0, 1, 1, 1));
         REQUIRE(p2);
 
-        uint64_t combined = p2->blocks_bitmap();
-        uint64_t horiz = p2->blocks_bitmap_h();
-        uint64_t bm = combined;
+        u64 combined = p2->blocks_bitmap();
+        u64 horiz = p2->blocks_bitmap_h();
+        u64 bm = combined;
 
         p2->bitmap_newly_occupied_after_move(bm, horiz, wes);
 
@@ -1034,34 +1034,34 @@ TEST_CASE("West move does not wrap for various board widths", "[bitmap_move]")
 TEST_CASE("bitmap_check_collision detects collision", "[collision]")
 {
     puzzle p(5, 5, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 2, 1, 1));
+    auto p2 = p.try_add_block(block(2, 2, 1, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(0, 0, 2, 1));
+    auto p3 = p2->try_add_block(block(0, 0, 2, 1));
     REQUIRE(p3);
 
-    uint64_t bm = p3->blocks_bitmap();
+    u64 bm = p3->blocks_bitmap();
 
     // Check if a hypothetical block at (1,2) 2x1 collides
-    puzzle::block hyp(1, 2, 2, 1);
+    block hyp(1, 2, 2, 1);
     CHECK(p3->bitmap_check_collision(bm, hyp)); // overlaps with (2,2)
 
     // Check if a hypothetical block at (3,3) 1x1 collides
-    puzzle::block hyp2(3, 3, 1, 1);
+    block hyp2(3, 3, 1, 1);
     CHECK_FALSE(p3->bitmap_check_collision(bm, hyp2));
 }
 
 TEST_CASE("bitmap_check_collision directional", "[collision]")
 {
     puzzle p(5, 5, 0, 0, false, false);
-    auto p2 = p.try_add_block(puzzle::block(2, 0, 1, 1));
+    auto p2 = p.try_add_block(block(2, 0, 1, 1));
     REQUIRE(p2);
-    auto p3 = p2->try_add_block(puzzle::block(2, 1, 1, 1));
+    auto p3 = p2->try_add_block(block(2, 1, 1, 1));
     REQUIRE(p3);
 
-    uint64_t bm = p3->blocks_bitmap();
+    u64 bm = p3->blocks_bitmap();
 
     // Clear block at (2,1) from bitmap to check if (2,1) can move north
-    puzzle::block b(2, 1, 1, 1);
+    block b(2, 1, 1, 1);
     p3->bitmap_clear_block(bm, b);
 
     // (2,1) moving north -> check (2,0) which has a block