massive state space solver improvement: supercomp took ~10s, now ~40ms

achieved by imposing a 15 block limit on each board and changing the
internal representation from std::string to 4x uint64_t

src/main.cpp

@@ -14,11 +14,18 @@
 #include <tracy/Tracy.hpp>
 #endif
 
-// TODO: Manual movement is now completely broken
+// TODO: Add some popups (my split between input.cpp/gui.cpp makes this ugly)
+//       - Clear graph: Notify that this will clear the visited states and move
+//         history
+//       - Reset state: Notify that this will reset the move count
 
-// TODO: Add state space generation time to debug overlay
+// TODO: Reduce memory usage
+//       - The memory model of the puzzle board is terrible (bitboards?)
 
 // TODO: Improve solver
+//       - Move discovery is terrible
+//         - Instead of trying each direction for each block, determine the
+//           possible moves more efficiently (requires a different memory model)
 //       - Implement state discovery/enumeration
 //         - Find all possible initial board states (single one for each
 //           possible statespace). Currently wer're just finding all states
@@ -30,6 +37,8 @@
 
 // TODO: Click states in the graph to display them in the board
 
+// NOTE: Tracy uses a huge amount of memory. For longer testing disable Tracy.
+
 // For profiling explore_state_space
 auto main2(int argc, char* argv[]) -> int
 {
@@ -164,4 +173,4 @@ auto main(int argc, char* argv[]) -> int
     CloseWindow();
 
     return 0;
-}
+}