performance: avoid memory allocations with cast_to in binary operatio…

…ns (#180)

also preallocate 64M objects

source/gc.hpp

@@ -1,28 +1,41 @@
 #pragma once
 
 #include <concepts>
+#include <cstddef>
 #include <cstdlib>
 #include <utility>
 #include <vector>
 
 template<typename T>
 class gc
 {
+    using store = std::vector<T*>;
+
   public:
-    static void track(T* obj) { get_allocation_list().push_back(obj); }
+    static void track(T* obj) { get_store().push_back(obj); }
 
   private:
     static void cleanup()
     {
-        for (T* obj : get_allocation_list()) {
+        for (T* obj : get_store()) {
             delete obj;
         }
     }
 
-    static auto get_allocation_list() -> std::vector<T*>&
+    static auto get_store() -> store&
     {
-        static std::vector<T*> allocations;
-        static const bool registered = []() { return std::atexit(cleanup); }();
+        static store allocations;
+        static const bool registered = []()
+        {
+            if constexpr (std::is_same_v<T, struct object>) {
+                constexpr auto object_reserve = static_cast<const size_t>(64 * 1024 * 1024);
+                allocations.reserve(object_reserve);
+            } else {
+                constexpr auto reserve = static_cast<const size_t>(128);
+                allocations.reserve(reserve);
+            }
+            return std::atexit(cleanup);
+        }();
         (void)registered;
         return allocations;
     }