solver: prune unplaceable large squares

A skyline may retain enough total empty area while no longer containing a box for its largest remaining square. Scan for the required consecutive low columns and reject such monotonic dead states.

Keep each pruning combination independently measurable and record the small public-path gain, the direct-order-9 regression, and the rejected periodic schedule.

Tests: Debug CTest (13 passed)

Tests: ASan+UBSan CTest (13 passed)

Refs: #15

BENCHMARKING.md

@@ -21,8 +21,9 @@ and `--candidate-order`.  The choices are `ascending`, `descending`, and
 The production default also removes equivalent D4 board orientations by
 constraining the unique unit square.  Pass `--no-symmetry` to obtain an
 otherwise identical unconstrained baseline.
-The production default also applies the valley-capacity rule described below.
-Pass `--no-pruning` to obtain an otherwise identical unpruned search.
+The production default also applies the pruning rules described below.
+Use `--pruning valley-capacity`, `--pruning large-square`, or
+`--no-pruning` to measure each rule alone or obtain an unpruned search.
 Order 9 uses the constructive odd-order path, searching order 8 and then tiling
 the enlarged border, so it is suitable for normal local benchmarking:
 
@@ -107,6 +108,55 @@ The rule reduced nodes by 7.8% for order 7 and 7.1% for order 8.  Its low
 per-node cost also reduced median counted solve time by 10.3% and 9.9%,
 respectively, so it remains enabled by default.
 
+## Remaining-large-square pruning
+
+For a remaining side `k`, a skyline can contain an empty `k`-by-`k` box only
+if it has `k` consecutive columns whose filled heights are no greater than the
+board height minus `k`.  A single linear scan tracks qualifying consecutive
+columns.  If no such run exists, the square cannot be placed and the state is
+infeasible even when its total empty area is sufficient.
+
+Empty-box feasibility is monotonic in the side length: a box which fits the
+largest remaining square also fits every smaller remaining size.  The solver
+therefore invokes only one `O(board width)` scan per surviving node, after the
+cheaper valley-capacity check.  It does not attempt an unproven multiplicity
+bound.  `--pruning large-square` measures the rule independently;
+`--pruning valley-capacity` provides the production baseline without it.
+Instrumented output records `large_square_checks` and
+`large_square_prunes`.
+
+Measurements used the issue #15 working tree based on commit `e27427d`, Apple
+Clang 21.0.0, `-O3 -DNDEBUG`, macOS arm64, one worker, one warm-up, and seven
+measured repetitions.  Counter-free and counted runs were interleaved; the
+table reports the counter-free production instantiation.  The baseline keeps
+valley-capacity pruning enabled, so it isolates the new rule.  All results
+passed independent validation and counters were stable:
+
+| Search | Large-square check | Median solve | Nodes | Checks | Prunes |
+| --- | --- | ---: | ---: | ---: | ---: |
+| Order 7 exhaustive | enabled | 0.946 s | 13,221,239 | 6,163,390 | 227,321 |
+| Order 7 exhaustive | disabled | 0.947 s | 13,833,048 | 0 | 0 |
+| Order 8 first solution | enabled | 0.195 s | 2,597,678 | 1,063,472 | 38,177 |
+| Order 8 first solution | disabled | 0.197 s | 2,724,096 | 0 | 0 |
+| Order 9 direct | enabled | 1.225 s | 14,840,146 | 5,592,843 | 54,900 |
+| Order 9 direct | disabled | 1.208 s | 14,995,127 | 0 | 0 |
+
+The rule reduced nodes by 4.4% for order 7, 4.6% for order 8, and 1.0% for
+direct order 9.  Counter-free median time improved by 0.12% and 0.89% on the
+two distinct searches in the public benchmark set.  Public order 9 constructs
+from the improving order-8 search.  Direct order 9, which bypasses that public
+route, regressed by 1.39%; it remains documented as a caution for future
+policy tuning.
+
+Checking only every fourth placement depth was also measured.  It retained
+fewer prunes and was slower than checking every surviving node by 2.2% for
+order 7, 2.3% for order 8, and 0.6% for direct order 9, so the periodic
+schedule was rejected.  Incremental maintenance would need additional
+per-size window state and undo logic for an `O(W)` scan whose public net cost
+is already recovered; it was not added without evidence that the complexity
+would improve elapsed time.  The simple every-node scan remains enabled by
+default for the measured public benchmark benefit.
+
 ## D4 board symmetry
 
 Every solution contains exactly one 1-by-1 square.  Rotations and reflections

CMakeLists.txt

@@ -27,6 +27,7 @@ if(BUILD_TESTING)
   add_test(NAME skyline-search COMMAND partridge_tests skyline-search)
   add_test(NAME d4-symmetry COMMAND partridge_tests d4-symmetry)
   add_test(NAME valley-capacity COMMAND partridge_tests valley-capacity)
+  add_test(NAME large-square COMMAND partridge_tests large-square)
   find_package(Python3 COMPONENTS Interpreter)
   if(Python3_Interpreter_FOUND)
     add_test(NAME benchmark-format

benchmarks/benchmark.cc

@@ -92,13 +92,24 @@ namespace {
   }
 
   auto boolean(bool value) -> char const * { return value ? "true" : "false"; }
+
+  auto pruning_name(Pruning const pruning) -> char const * {
+    switch (pruning) {
+      case Pruning::all: return "all";
+      case Pruning::valley_capacity: return "valley-capacity";
+      case Pruning::large_square: return "large-square";
+      case Pruning::disabled: return "none";
+    }
+    assert(false);
+    return "none";
+  }
 }
 
 int main(int argc, char **argv) {
   if (argc < 3 || argc > 7) {
     std::cerr << "usage: partridge_benchmark ORDER counters|plain "
                  "[ascending|descending|best-fit] [public|direct] "
-                 "[d4|none] [valley-capacity|none]\n";
+                 "[d4|none] [all|valley-capacity|large-square|none]\n";
     return 2;
   }
   auto const order = static_cast<std::uint64_t>(std::strtoull(argv[1], nullptr, 10));
@@ -136,12 +147,19 @@ int main(int argc, char **argv) {
     return 2;
   }
   auto const pruning =
-      argc < 7 || std::string_view(argv[6]) == "valley-capacity"
+      argc < 7 || std::string_view(argv[6]) == "all"
+          ? Pruning::all
+          : std::string_view(argv[6]) == "valley-capacity"
                 ? Pruning::valley_capacity
+                : std::string_view(argv[6]) == "large-square"
+                      ? Pruning::large_square
                       : Pruning::disabled;
-  if (argc == 7 && std::string_view(argv[6]) != "valley-capacity" &&
+  if (argc == 7 && std::string_view(argv[6]) != "all" &&
+      std::string_view(argv[6]) != "valley-capacity" &&
+      std::string_view(argv[6]) != "large-square" &&
       std::string_view(argv[6]) != "none") {
-    std::cerr << "pruning mode must be valley-capacity or none\n";
+    std::cerr << "pruning mode must be all, valley-capacity, "
+                 "large-square, or none\n";
     return 2;
   }
 
@@ -177,8 +195,7 @@ int main(int argc, char **argv) {
       << (symmetry == SymmetryBreaking::d4_unit_square ? "d4" : "none")
       << "\""
       << ",\"pruning\":\""
-      << (pruning == Pruning::valley_capacity ? "valley-capacity" : "none")
-      << "\""
+      << pruning_name(pruning) << "\""
       << ",\"solved\":" << boolean(!timed.result.squares().empty())
       << ",\"valid\":" << boolean(validation_ok)
       << ",\"timing_seconds\":{\"solve\":" << timed.search_seconds
@@ -195,6 +212,10 @@ int main(int argc, char **argv) {
       << counters.valley_capacity_checks
       << ",\"valley_capacity_prunes\":"
       << counters.valley_capacity_prunes
+      << ",\"large_square_checks\":"
+      << counters.large_square_checks
+      << ",\"large_square_prunes\":"
+      << counters.large_square_prunes
       << ",\"generated_tasks\":" << counters.generated_tasks
       << ",\"completed_tasks\":" << counters.completed_tasks << "}}\n";
   return validation_ok ? 0 : 1;

benchmarks/run.py

@@ -238,10 +238,16 @@ def main():
         action="store_true",
         help="disable unique-unit-square D4 symmetry breaking",
     )
+    parser.add_argument(
+        "--pruning",
+        choices=("all", "valley-capacity", "large-square", "none"),
+        default="all",
+        help="select independently measurable pruning rules",
+    )
     parser.add_argument(
         "--no-pruning",
         action="store_true",
-        help="disable valley-capacity pruning",
+        help="disable all pruning (compatibility alias for --pruning none)",
     )
     parser.add_argument(
         "--measure-overhead",
@@ -275,7 +281,7 @@ def main():
                     args.search_policy,
                     "direct" if args.direct_search else "public",
                     "none" if args.no_symmetry else "d4",
-                    "none" if args.no_pruning else "valley-capacity",
+                    "none" if args.no_pruning else args.pruning,
                     args.timeout,
                 )
         for trial in range(args.repetitions):
@@ -288,7 +294,7 @@ def main():
                         args.search_policy,
                         "direct" if args.direct_search else "public",
                         "none" if args.no_symmetry else "d4",
-                        "none" if args.no_pruning else "valley-capacity",
+                        "none" if args.no_pruning else args.pruning,
                         args.timeout,
                     )
                 )
@@ -318,7 +324,7 @@ def main():
             "candidate_order": args.search_policy,
             "search_route": "direct" if args.direct_search else "public",
             "symmetry_breaking": "none" if args.no_symmetry else "d4",
-            "pruning": "none" if args.no_pruning else "valley-capacity",
+            "pruning": "none" if args.no_pruning else args.pruning,
             "stdout": "captured; rendered grid suppressed by probe",
         },
         "cases": cases,

main.cc

@@ -152,6 +152,8 @@ namespace {
     size_t prune_hits = 0;
     size_t valley_capacity_checks = 0;
     size_t valley_capacity_prunes = 0;
+    size_t large_square_checks = 0;
+    size_t large_square_prunes = 0;
     size_t generated_tasks = 0;
     size_t completed_tasks = 0;
   };
@@ -172,8 +174,10 @@ namespace {
 
   /** Cheap necessary conditions applied before branching at a search node. */
   enum class Pruning {
-    disabled,
-    valley_capacity,
+    disabled = 0,
+    valley_capacity = 1,
+    large_square = 2,
+    all = 3,
   };
 
   /** Return whether a cell is the canonical representative of its D4 orbit.
@@ -259,7 +263,46 @@ namespace {
     return available_area >= required_area;
   }
 
-  template<bool Instrument, bool BreakD4Symmetry, bool PruneValleyCapacity>
+  /** Return whether the skyline contains an empty box for a square.
+   *
+   * A side-k square fits exactly when k consecutive columns have heights no
+   * greater than board height minus k.  Tracking the current qualifying run
+   * checks this in one pass without allocations.
+   */
+  [[nodiscard]] auto skyline_has_empty_square(
+      std::vector<size_t> const &skyline, size_t const side) noexcept -> bool {
+    assert(side > 0);
+    assert(side <= skyline.size());
+    auto const maximum_height = skyline.size() - side;
+    size_t run = 0;
+    for (auto const height: skyline) {
+      run = height <= maximum_height ? run + 1 : 0;
+      if (run == side) {
+        return true;
+      }
+    }
+    return false;
+  }
+
+  /** Return whether the largest remaining square has any feasible position.
+   *
+   * Feasible empty boxes are monotonic in the side length: a box which fits
+   * the largest remaining square also fits every smaller one.  It is therefore
+   * sufficient to test only the largest size with non-zero multiplicity.
+   */
+  [[nodiscard]] auto remaining_large_square_fits(
+      std::vector<size_t> const &skyline, Avail const &available) noexcept
+      -> bool {
+    for (auto side = available.size() - 1; side != 0; --side) {
+      if (available[side] != 0) {
+        return skyline_has_empty_square(skyline, side);
+      }
+    }
+    return true;
+  }
+
+  template<bool Instrument, bool BreakD4Symmetry, bool PruneValleyCapacity,
+           bool PruneLargeSquare>
   auto search_skyline(size_t const n, size_t const length,
                       SearchPolicy const policy,
                       std::vector<size_t> &skyline, Avail &available,
@@ -288,6 +331,19 @@ namespace {
         return false;
       }
     }
+    if constexpr (PruneLargeSquare) {
+      if constexpr (Instrument) {
+        ++counters->prune_checks;
+        ++counters->large_square_checks;
+      }
+      if (!remaining_large_square_fits(skyline, available)) {
+        if constexpr (Instrument) {
+          ++counters->prune_hits;
+          ++counters->large_square_prunes;
+        }
+        return false;
+      }
+    }
     auto const largest =
         std::min({n, valley.width, length - valley.height});
     auto try_side = [&](size_t const side) {
@@ -320,7 +376,8 @@ namespace {
                   side, valley.height + side);
       squares.emplace_back(valley.x + valley.height * length, side);
 
-      if (search_skyline<Instrument, BreakD4Symmetry, PruneValleyCapacity>(
+      if (search_skyline<Instrument, BreakD4Symmetry, PruneValleyCapacity,
+                         PruneLargeSquare>(
               n, length, policy, skyline, available, squares, counters)) {
         return true;
       }
@@ -367,7 +424,8 @@ namespace {
    * candidates and updating up to n columns for each costs O(n^2), for
    * O(board width + n^2) local work per node and the same total state.
    */
-  template<bool Instrument, bool BreakD4Symmetry, bool PruneValleyCapacity>
+  template<bool Instrument, bool BreakD4Symmetry, bool PruneValleyCapacity,
+           bool PruneLargeSquare>
   auto search_solution_impl(size_t const n, SearchPolicy const policy,
                             SearchCounters *const counters) noexcept
       -> Results {
@@ -380,29 +438,50 @@ namespace {
     std::vector<Square> squares;
     squares.reserve(length);
     static_cast<void>(
-        search_skyline<Instrument, BreakD4Symmetry, PruneValleyCapacity>(
+        search_skyline<Instrument, BreakD4Symmetry, PruneValleyCapacity,
+                       PruneLargeSquare>(
             n, length, policy, skyline, available, squares, counters));
 
     return {length, std::move(squares)};
   }
 
+  template<bool Instrument, bool BreakD4Symmetry>
+  auto search_solution_dispatch(size_t const n, SearchPolicy const policy,
+                                Pruning const pruning,
+                                SearchCounters *const counters) noexcept
+      -> Results {
+    switch (pruning) {
+      case Pruning::all:
+        return search_solution_impl<Instrument, BreakD4Symmetry, true, true>(
+            n, policy, counters);
+      case Pruning::valley_capacity:
+        return search_solution_impl<Instrument, BreakD4Symmetry, true, false>(
+            n, policy, counters);
+      case Pruning::large_square:
+        return search_solution_impl<Instrument, BreakD4Symmetry, false, true>(
+            n, policy, counters);
+      case Pruning::disabled:
+        return search_solution_impl<Instrument, BreakD4Symmetry, false, false>(
+            n, policy, counters);
+    }
+    assert(false);
+    return search_solution_impl<Instrument, BreakD4Symmetry, false, false>(
+        n, policy, counters);
+  }
+
   auto search_solution(
       size_t const n,
       SearchPolicy const policy = SearchPolicy::ascending,
       SymmetryBreaking const symmetry =
           SymmetryBreaking::d4_unit_square,
-      Pruning const pruning = Pruning::valley_capacity) noexcept
+      Pruning const pruning = Pruning::all) noexcept
       -> Results {
     if (symmetry == SymmetryBreaking::d4_unit_square) {
-      if (pruning == Pruning::valley_capacity) {
-        return search_solution_impl<false, true, true>(n, policy, nullptr);
+      return search_solution_dispatch<false, true>(
+          n, policy, pruning, nullptr);
     }
-      return search_solution_impl<false, true, false>(n, policy, nullptr);
-    }
-    if (pruning == Pruning::valley_capacity) {
-      return search_solution_impl<false, false, true>(n, policy, nullptr);
-    }
-    return search_solution_impl<false, false, false>(n, policy, nullptr);
+    return search_solution_dispatch<false, false>(
+        n, policy, pruning, nullptr);
   }
 
   auto search_solution_instrumented(size_t const n,
@@ -412,21 +491,15 @@ namespace {
                                     SymmetryBreaking const symmetry =
                                         SymmetryBreaking::d4_unit_square,
                                     Pruning const pruning =
-                                        Pruning::valley_capacity) noexcept
+                                        Pruning::all) noexcept
       -> Results {
     counters = {};
     if (symmetry == SymmetryBreaking::d4_unit_square) {
-      if (pruning == Pruning::valley_capacity) {
-        return search_solution_impl<true, true, true>(
-            n, policy, &counters);
+      return search_solution_dispatch<true, true>(
+          n, policy, pruning, &counters);
     }
-      return search_solution_impl<true, true, false>(n, policy, &counters);
-    }
-    if (pruning == Pruning::valley_capacity) {
-      return search_solution_impl<true, false, true>(
-          n, policy, &counters);
-    }
-    return search_solution_impl<true, false, false>(n, policy, &counters);
+    return search_solution_dispatch<true, false>(
+        n, policy, pruning, &counters);
   }
 
   /** Construct an odd-order solution from its even-order predecessor. */
@@ -464,7 +537,7 @@ namespace {
       SearchPolicy const policy = SearchPolicy::ascending,
       SymmetryBreaking const symmetry =
           SymmetryBreaking::d4_unit_square,
-      Pruning const pruning = Pruning::valley_capacity) noexcept -> Results {
+      Pruning const pruning = Pruning::all) noexcept -> Results {
     if (uses_odd_construction(n)) {
       return construct_odd_solution(
           n, search_solution(n - 1, policy, symmetry, pruning));
@@ -479,7 +552,7 @@ namespace {
                                   SymmetryBreaking const symmetry =
                                       SymmetryBreaking::d4_unit_square,
                                   Pruning const pruning =
-                                      Pruning::valley_capacity) noexcept
+                                      Pruning::all) noexcept
       -> Results {
     if (uses_odd_construction(n)) {
       return construct_odd_solution(

tests/tests.cc

@@ -369,6 +369,8 @@ namespace {
     failures += expect(counters.prune_checks >= counters.prune_hits &&
                            counters.valley_capacity_checks >=
                                counters.valley_capacity_prunes &&
+                           counters.large_square_checks >=
+                               counters.large_square_prunes &&
                            counters.generated_tasks == 0 &&
                            counters.completed_tasks == 0,
                        "search counters are inconsistent");
@@ -582,6 +584,71 @@ namespace {
     return failures;
   }
 
+  auto test_large_square_pruning() -> int {
+    int failures = 0;
+    auto const fragmented =
+        std::vector<std::uint64_t>{0, 5, 0, 5, 0, 5};
+    failures += expect(
+        !skyline_has_empty_square(fragmented, 2),
+        "large-square check accepted a fragmented state without a 2x2 box");
+    failures += expect(
+        skyline_has_empty_square(
+            std::vector<std::uint64_t>{0, 0, 5, 5, 5, 5}, 2),
+        "large-square check rejected an available 2x2 box");
+
+    Avail available(4);
+    available[1] = 1;
+    available[2] = 1;
+    failures += expect(
+        !remaining_large_square_fits(fragmented, available),
+        "remaining-square check ignored the impossible largest square");
+    available[2] = 0;
+    failures += expect(
+        remaining_large_square_fits(fragmented, available),
+        "remaining-square check did not use geometric size monotonicity");
+
+    // The fragmented profile has 21 empty cells, more than the area of the
+    // remaining 2x2 square, but no two adjacent columns have two free rows.
+    std::uint64_t filled_area = 0;
+    for (auto const height: fragmented) {
+      filled_area += height;
+    }
+    failures += expect(
+        filled_area <= 36 - 4,
+        "fragmented test state does not have sufficient total empty area");
+
+    for (auto const order: std::array<std::uint64_t, 5>{1, 2, 3, 4, 5}) {
+      SearchCounters pruned_counters;
+      SearchCounters baseline_counters;
+      auto const pruned = search_solution_instrumented(
+          order, pruned_counters, SearchPolicy::ascending,
+          SymmetryBreaking::disabled, Pruning::all);
+      auto const baseline = search_solution_instrumented(
+          order, baseline_counters, SearchPolicy::ascending,
+          SymmetryBreaking::disabled, Pruning::valley_capacity);
+      failures += expect(
+          pruned.squares().empty() == baseline.squares().empty(),
+          "large-square pruning changed exhaustive order-" +
+              std::to_string(order) + " feasibility");
+      failures += expect(
+          pruned_counters.search_nodes <= baseline_counters.search_nodes,
+          "large-square pruning enlarged the exhaustive order-" +
+              std::to_string(order) + " search");
+    }
+
+    SearchCounters counters;
+    static_cast<void>(search_solution_instrumented(
+        5, counters, SearchPolicy::ascending, SymmetryBreaking::disabled,
+        Pruning::large_square));
+    failures += expect(
+        counters.large_square_checks > 0 &&
+            counters.large_square_prunes > 0 &&
+            counters.prune_checks == counters.large_square_checks &&
+            counters.prune_hits == counters.large_square_prunes,
+        "large-square instrumentation did not count checks and prunes");
+    return failures;
+  }
+
   auto test_solver_completion() -> int {
     int failures = 0;
     for (auto const order: std::array<std::uint64_t, 2>{1, 8}) {
@@ -633,6 +700,9 @@ int main(int argc, char **argv) {
   if (test == "valley-capacity") {
     return test_valley_capacity_pruning();
   }
+  if (test == "large-square") {
+    return test_large_square_pruning();
+  }
   std::cerr << "unknown test: " << test << '\n';
   return 2;
 }