Perf/both bench opt

CMakeLists.txt

@@ -62,3 +62,18 @@ if (ENABLE_TESTING)
 endif()
 
 target_link_libraries(main PRIVATE ${CPP_ORDERBOOK} Boost::intrusive decimal pool)
+
+# --- Link-time optimization (LTO/IPO) for optimized builds only. ------------
+# Enabled only for Release-type configs and only when the toolchain supports
+# it. Debug builds and the sanitizer builds (ASan/UBSan/TSan) are left
+# untouched on purpose: LTO is slow under sanitizers and can misbehave. Note
+# the sanitizer CI builds the engine via bench/build.sh, not this CMake, so it
+# never reaches here; the gate below additionally keeps a Debug configure clean.
+include(CheckIPOSupported)
+check_ipo_supported(RESULT _ipo_ok OUTPUT _ipo_msg LANGUAGES CXX)
+if (_ipo_ok AND CMAKE_BUILD_TYPE MATCHES "^(Release|RelWithDebInfo|MinSizeRel)$")
+    set_property(TARGET ${CPP_ORDERBOOK} main PROPERTY INTERPROCEDURAL_OPTIMIZATION TRUE)
+    message(STATUS "LTO/IPO enabled for ${CMAKE_BUILD_TYPE} build")
+elseif (NOT _ipo_ok)
+    message(STATUS "LTO/IPO not supported by toolchain, continuing without it: ${_ipo_msg}")
+endif()

bench/build.sh

@@ -98,9 +98,19 @@ for p in "$DECIMAL_INC" "$POOL_INC"; do
     fi
 done
 
+# --- Link-time optimization. This adapter is a perf/Release artifact, so it
+# gets -flto for cross-TU inlining of the matching chain. The single g++ call
+# below both compiles and links, so one -flto covers both phases. We DROP -flto
+# when ME_EXTRA_CXXFLAGS asks for a sanitizer build (ASan/UBSan/TSan), where LTO
+# is slow and can misbehave with the instrumentation.
+LTO_FLAG="-flto"
+case "${ME_EXTRA_CXXFLAGS:-}" in
+    *sanitize*) LTO_FLAG="" ;;
+esac
+
 # --- Compile the adapter + engine TUs into the .so. -------------------------
 OUT="$DIR/cpp_orderbook_adapter.so"
-g++ -std=c++20 -O3 -march=native -fPIC -shared \
+g++ -std=c++20 -O3 -march=native $LTO_FLAG -fPIC -shared \
     -I"$BENCH/api" \
     -I"$ENGINE/include" \
     -I"$ENGINE/src" \

include/function_ref.hpp

@@ -0,0 +1,31 @@
+#pragma once
+
+#include <type_traits>
+#include <utility>
+
+namespace orderbook {
+
+// Non-owning view over any callable, modeled after std::function_ref (P0792).
+// Stores a type-erased object pointer plus a thunk; trivially copyable, two
+// pointers wide, performs no allocation. The referenced callable must outlive
+// every invocation through the function_ref.
+template <class Signature>
+class function_ref;
+
+template <class R, class... Args>
+class function_ref<R(Args...)> {
+    void* obj_ = nullptr;
+    R (*thunk_)(void*, Args...) = nullptr;
+
+   public:
+    template <class F, class = std::enable_if_t<!std::is_same_v<std::decay_t<F>, function_ref>>>
+    function_ref(F&& f) noexcept
+        : obj_(const_cast<void*>(static_cast<const void*>(std::addressof(f)))),
+          thunk_([](void* obj, Args... args) -> R {
+              return (*static_cast<std::remove_reference_t<F>*>(obj))(std::forward<Args>(args)...);
+          }) {}
+
+    R operator()(Args... args) const { return thunk_(obj_, std::forward<Args>(args)...); }
+};
+
+}  // namespace orderbook

include/orderqueue.hpp

@@ -1,17 +1,17 @@
 #pragma once
 
 #include <cstdint>
-#include <functional>
 
 #include "boost/intrusive/list.hpp"
 #include "boost/intrusive/set_hook.hpp"
+#include "function_ref.hpp"
 #include "order.hpp"
 
 namespace orderbook {
 
 using TradeNotification =
-    std::function<void(OrderID mOrderID, OrderID tOrderID, OrderStatus mOrderStatus, OrderStatus tOrderStatus, Decimal qty, Decimal price)>;
-using PostOrderFill = std::function<void(OrderID canceledOrderID)>;
+    function_ref<void(OrderID mOrderID, OrderID tOrderID, OrderStatus mOrderStatus, OrderStatus tOrderStatus, Decimal qty, Decimal price)>;
+using PostOrderFill = function_ref<void(OrderID canceledOrderID)>;
 
 using OrderList = boost::intrusive::list<Order>;
 

src/pricelevel.cpp

@@ -52,7 +52,7 @@ void PriceLevel<P>::remove(Order* order) {
     }
 
     if (q->len() == 0) {
-        price_tree_.erase(q->price());
+        price_tree_.erase(price_tree_.iterator_to(*q));
         --depth_;
         queue_pool_.release(&*q);
     }

test/orderqueue_test.cpp

@@ -54,9 +54,9 @@ TEST_F(OrderQueueTest, TestOrderQueue_ProcessUpdatesTotalQtyOnFullFill) {
     oq->append(&o1);
     oq->append(&o2);
 
-    const TradeNotification tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
+    const auto tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
                                     Decimal matchedPrice) {};
-    const PostOrderFill pf = [&oq, &o1, &o2](OrderID id) {
+    const auto pf = [&oq, &o1, &o2](OrderID id) {
         if (id == 1) {
             oq->remove(&o1);
         } else if (id == 2) {
@@ -84,9 +84,9 @@ TEST_F(OrderQueueTest, TestOrderQueue_ProcessUpdatesTotalQtyOnExactFill) {
     oq->append(&o1);
     oq->append(&o2);
 
-    const TradeNotification tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
+    const auto tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
                                     Decimal matchedPrice) {};
-    const PostOrderFill pf = [&oq, &o1, &o2](OrderID id) {
+    const auto pf = [&oq, &o1, &o2](OrderID id) {
         if (id == 1) {
             oq->remove(&o1);
         } else if (id == 2) {
@@ -112,9 +112,9 @@ TEST_F(OrderQueueTest, TestOrderQueue_ProcessZeroesFilledOrderBeforePostFill) {
     oq->append(&o2);
 
     bool sawZeroQtyOnPostFill = false;
-    const TradeNotification tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
+    const auto tn = [](OrderID makerOrderID, OrderID takerOrderID, OrderStatus makerOrderStatus, OrderStatus takerOrderStatus, Decimal matchedQty,
                                     Decimal matchedPrice) {};
-    const PostOrderFill pf = [&oq, &o1, &o2, &sawZeroQtyOnPostFill](OrderID id) {
+    const auto pf = [&oq, &o1, &o2, &sawZeroQtyOnPostFill](OrderID id) {
         if (id == 1) {
             sawZeroQtyOnPostFill = (o1.qty == Decimal(0, 0));
             oq->remove(&o1);
@@ -146,12 +146,12 @@ TEST_F(OrderQueueTest, TestOrderQueue_ProcessUsesSnapshotForTradeNotificationAft
 
     OrderID makerOrderID = 0;
     Decimal matchedPrice(0, 0);
-    const TradeNotification tn = [&makerOrderID, &matchedPrice](OrderID makerID, OrderID takerOrderID, OrderStatus makerOrderStatus,
+    const auto tn = [&makerOrderID, &matchedPrice](OrderID makerID, OrderID takerOrderID, OrderStatus makerOrderStatus,
                                                                 OrderStatus takerOrderStatus, Decimal matchedQty, Decimal priceValue) {
         makerOrderID = makerID;
         matchedPrice = priceValue;
     };
-    const PostOrderFill pf = [&oq, &o1](OrderID id) {
+    const auto pf = [&oq, &o1](OrderID id) {
         if (id == 1) {
             oq->remove(&o1);
             o1.id = 999;