diff --git a/optuna/samplers/nsgaii/_constraints_evaluation.py b/optuna/samplers/nsgaii/_constraints_evaluation.py
index f7e14638c9..33f3b998e5 100644
--- a/optuna/samplers/nsgaii/_constraints_evaluation.py
+++ b/optuna/samplers/nsgaii/_constraints_evaluation.py
@@ -92,13 +92,15 @@ def _evaluate_penalty(population: Sequence[FrozenTrial]) -> np.ndarray:
         feasible/infeasible and None means that the trial does not have constraint values.
     """
 
-    penalty: list[float] = []
-    for trial in population:
-        constraints = trial.system_attrs.get(_CONSTRAINTS_KEY)
-        if constraints is None:
-            penalty.append(np.nan)
-        else:
-            penalty.append(sum(v for v in constraints if v > 0))
+    # Use list comprehension for better performance
+    penalty = [
+        (
+            np.nan
+            if (constraints := trial.system_attrs.get(_CONSTRAINTS_KEY)) is None
+            else sum(v for v in constraints if v > 0)
+        )
+        for trial in population
+    ]
     return np.array(penalty)
 
 
diff --git a/optuna/samplers/nsgaii/_elite_population_selection_strategy.py b/optuna/samplers/nsgaii/_elite_population_selection_strategy.py
index bf416267ae..5920b95715 100644
--- a/optuna/samplers/nsgaii/_elite_population_selection_strategy.py
+++ b/optuna/samplers/nsgaii/_elite_population_selection_strategy.py
@@ -130,7 +130,9 @@ def _rank_population(
     penalty = _evaluate_penalty(population) if is_constrained else None
 
     domination_ranks = _fast_non_domination_rank(objective_values, penalty=penalty)
-    population_per_rank: list[list[FrozenTrial]] = [[] for _ in range(max(domination_ranks) + 1)]
+    num_ranks = domination_ranks.max() + 1
+    population_per_rank: list[list[FrozenTrial]] = [[] for _ in range(num_ranks)]
+    # Use more efficient grouping via preallocated lists and bulk push
     for trial, rank in zip(population, domination_ranks):
         if rank == -1:
             continue
diff --git a/optuna/study/_multi_objective.py b/optuna/study/_multi_objective.py
index 71eb465288..976e7517ec 100644
--- a/optuna/study/_multi_objective.py
+++ b/optuna/study/_multi_objective.py
@@ -97,25 +97,35 @@ def _fast_non_domination_rank(
 
     ranks = np.full(len(loss_values), -1, dtype=int)
     is_penalty_nan = np.isnan(penalty)
-    is_feasible = np.logical_and(~is_penalty_nan, penalty <= 0)
-    is_infeasible = np.logical_and(~is_penalty_nan, penalty > 0)
+    is_feasible = (~is_penalty_nan) & (penalty <= 0)
+    is_infeasible = (~is_penalty_nan) & (penalty > 0)
 
     # First, we calculate the domination rank for feasible trials.
-    ranks[is_feasible] = _calculate_nondomination_rank(loss_values[is_feasible], n_below=n_below)
-    n_below -= int(np.count_nonzero(is_feasible))
+    feasible_count = np.count_nonzero(is_feasible)
+    if feasible_count:
+        feasible_ranks = _calculate_nondomination_rank(loss_values[is_feasible], n_below=n_below)
+        ranks[is_feasible] = feasible_ranks
+        n_below -= feasible_count
 
     # Second, we calculate the domination rank for infeasible trials.
-    top_rank_infeasible = np.max(ranks[is_feasible], initial=-1) + 1
-    ranks[is_infeasible] = top_rank_infeasible + _calculate_nondomination_rank(
-        penalty[is_infeasible][:, np.newaxis], n_below=n_below
-    )
-    n_below -= int(np.count_nonzero(is_infeasible))
+    infeasible_count = np.count_nonzero(is_infeasible)
+    if infeasible_count:
+        top_rank_infeasible = np.max(ranks[is_feasible], initial=-1) + 1
+        infeasible_ranks = _calculate_nondomination_rank(
+            penalty[is_infeasible][:, np.newaxis], n_below=n_below
+        )
+        ranks[is_infeasible] = top_rank_infeasible + infeasible_ranks
+        n_below -= infeasible_count
 
     # Third, we calculate the domination rank for trials with no penalty information.
-    top_rank_penalty_nan = np.max(ranks[~is_penalty_nan], initial=-1) + 1
-    ranks[is_penalty_nan] = top_rank_penalty_nan + _calculate_nondomination_rank(
-        loss_values[is_penalty_nan], n_below=n_below
-    )
+    penalty_nan_count = np.count_nonzero(is_penalty_nan)
+    if penalty_nan_count:
+        top_rank_penalty_nan = np.max(ranks[~is_penalty_nan], initial=-1) + 1
+        penalty_nan_ranks = _calculate_nondomination_rank(
+            loss_values[is_penalty_nan], n_below=n_below
+        )
+        ranks[is_penalty_nan] = top_rank_penalty_nan + penalty_nan_ranks
+
     assert np.all(ranks != -1), "All the rank must be updated."
     return ranks