Optimizations

BenchMarkReport_0.md

@@ -0,0 +1,58 @@
+# Reflection Template Library (RTL) — Benchmark Report
+
+This document presents benchmark results for the **Reflection Template Library (RTL)**.  
+The goal was to measure the runtime overhead of reflective function calls compared to direct calls and `std::function`, under increasing workloads.
+
+---
+
+## Benchmark Setup
+
+We tested:
+
+- **Direct calls** (baseline).
+- **`std::function` calls** and method calls.
+- **RTL reflective calls** (free functions and member methods, with and without return values).
+
+Each benchmark was repeated across workloads of increasing complexity, with times measured in nanoseconds.
+
+---
+
+## Results Summary
+
+| Workload        | Direct Call (ns) | Reflected Call Overhead (ns) | Reflected Method Overhead (ns) | Notes (With Return) |
+|-----------------|------------------|------------------------------|--------------------------------|---------------------|
+| baseline_40ns   | 39.0 / 44.7      | +2.5                         | +6.6                           | +10.6 / +14.3       |
+| workload_80ns   | 82.4 / 82.5      | ~0                           | ~0                             | +12.5 / +15.6       |
+| workload_100ns  | 94.2 / 100.0     | +1.4                         | +8.8                           | +12.0 / +16.0       |
+| workload_150ns* | 139.0 / 158.0    | +2–3                         | +14–17                         | +12–13 / +17–19     |
+
+\*Three independent runs were recorded at ~150 ns workload; numbers are consistent.
+
+---
+
+## Insights
+
+- **Constant Overhead**  
+  Reflection overhead remains almost constant across workloads:
+  - No-return functions: **+2–6 ns**.
+  - Return-value functions: **+10–20 ns**.
+
+- **Percentage Overhead Shrinks**  
+  - At the 40 ns baseline, overhead was ~25%.  
+  - By ~150 ns workloads, overhead dropped below 10%.  
+
+- **No Scaling Penalty**  
+  The overhead does not grow with function complexity.  
+  This indicates that RTL adds only a fixed, predictable cost per call, with no hidden allocations.
+
+- **Performance-Culture Friendly**  
+  This aligns with C++’s ethos: *you only pay a small, predictable cost when you use reflection*.
+
+---
+
+## Conclusion
+
+The Reflection Template Library (RTL) demonstrates:
+
+- **Runtime reflection with constant, minimal overhead**.  
+- **Predictable cost model**: ~10–20 ns for reflective calls with returns.  

CxxRTLTypeRegistration/inc/TestMirrorProvider.h

@@ -22,7 +22,7 @@ namespace test_mirror
 		static std::size_t calender;
 
 		static std::size_t char_t;
-		static std::size_t void_t;
+		static std::size_t int_t;
 		static std::size_t std_string;
 		static std::size_t std_string_view;
 

CxxRTLTypeRegistration/src/TestMirrorProvider.cpp

@@ -37,15 +37,15 @@ namespace test_mirror
             ---------------------------------  */
 
         //  Registering void, valid but not useful at all.
-            rtl::type().record<void>("void").build(),
+            rtl::type().record<int>("int").build(),
 
         //  Registering type 'void' again, ignored & emits-
         //  [WARNING] Multiple registrations of the same type detected.
-            rtl::type().record<void>("void").build(),
+            rtl::type().record<int>("int").build(),
 
         //  Registering type 'void' again, but with different name. ignored & emits-
         //  [WARNING] Multiple registrations of the same type detected.
-            rtl::type().record<void>("ccvoid").build(),
+            rtl::type().record<int>("ccint").build(),
 
         //  Registering pod, reflecting- constructor, copy-constructor & destructor.
             rtl::type().record<char>("char").build(),
@@ -266,7 +266,7 @@ namespace test_mirror
     std::size_t reflected_id::event = rtl::detail::TypeId<nsdate::Event>::get();
     std::size_t reflected_id::calender = rtl::detail::TypeId<nsdate::Calender>::get();
 
-    std::size_t reflected_id::void_t = rtl::detail::TypeId<void>::get();
+    std::size_t reflected_id::int_t = rtl::detail::TypeId<int>::get();
     std::size_t reflected_id::char_t = rtl::detail::TypeId<char>::get();
     std::size_t reflected_id::std_string = rtl::detail::TypeId<std::string>::get();
     std::size_t reflected_id::std_string_view = rtl::detail::TypeId<std::string_view>::get();
@@ -277,7 +277,7 @@ namespace test_mirror
         static std::unordered_map<std::string, std::size_t> nameIdMap(
         {
             { "char", char_t },
-            { "void", void_t },
+            { "int", int_t },
             { "string", std_string },
             { "string_view", std_string_view },