Optimizations for trivial types

Relying on various compiler primitives we can reduce the work done
in our memory allocators and CowData. For types with trivial ctors or
dtors we can skip looping over all elements when creating, resizing,
and destroying lists of objects.

These primitives are supported by clang, msvc, and GCC. However, once
we've moved to C++11 we can rely on several std:: primitives that do
the same thing and are standardized.

In my testing the extra conditionals introduced here get removed from
the generated program entirely as the results for these primitives is
known at compile time.

2 files changed, 44 insertions, 25 deletions

diff --git a/core/cowdata.h b/core/cowdata.h
index d5d49215c2..319e61d261 100644
--- a/core/cowdata.h
+++ b/core/cowdata.h
@@ -31,6 +31,8 @@
 #ifndef COWDATA_H_
 #define COWDATA_H_
 
+#include <string.h>
+
 #include "core/os/memory.h"
 #include "core/safe_refcount.h"
 
@@ -194,12 +196,14 @@ void CowData<T>::_unref(void *p_data) {
 		return; // still in use
 	// clean up
 
-	uint32_t *count = _get_size();
-	T *data = (T *)(count + 1);
+	if (!__has_trivial_destructor(T)) {
+		uint32_t *count = _get_size();
+		T *data = (T *)(count + 1);
 
-	for (uint32_t i = 0; i < *count; ++i) {
-		// call destructors
-		data[i].~T();
+		for (uint32_t i = 0; i < *count; ++i) {
+			// call destructors
+			data[i].~T();
+		}
 	}
 
 	// free mem
@@ -226,9 +230,13 @@ void CowData<T>::_copy_on_write() {
 		T *_data = (T *)(mem_new);
 
 		// initialize new elements
-		for (uint32_t i = 0; i < current_size; i++) {
+		if (__has_trivial_copy(T)) {
+			memcpy(mem_new, _ptr, current_size * sizeof(T));
 
-			memnew_placement(&_data[i], T(_get_data()[i]));
+		} else {
+			for (uint32_t i = 0; i < current_size; i++) {
+				memnew_placement(&_data[i], T(_get_data()[i]));
+			}
 		}
 
 		_unref(_ptr);
@@ -275,22 +283,25 @@ Error CowData<T>::resize(int p_size) {
 		}
 
 		// construct the newly created elements
-		T *elems = _get_data();
 
-		for (int i = *_get_size(); i < p_size; i++) {
+		if (!__has_trivial_constructor(T)) {
+			T *elems = _get_data();
 
-			memnew_placement(&elems[i], T);
+			for (int i = *_get_size(); i < p_size; i++) {
+				memnew_placement(&elems[i], T);
+			}
 		}
 
 		*_get_size() = p_size;
 
 	} else if (p_size < size()) {
 
-		// deinitialize no longer needed elements
-		for (uint32_t i = p_size; i < *_get_size(); i++) {
-
-			T *t = &_get_data()[i];
-			t->~T();
+		if (!__has_trivial_destructor(T)) {
+			// deinitialize no longer needed elements
+			for (uint32_t i = p_size; i < *_get_size(); i++) {
+				T *t = &_get_data()[i];
+				t->~T();
+			}
 		}
 
 		void *_ptrnew = (T *)Memory::realloc_static(_ptr, alloc_size, true);
diff --git a/core/os/memory.h b/core/os/memory.h
index 558dc5ddf3..f3ca9fc614 100644
--- a/core/os/memory.h
+++ b/core/os/memory.h
@@ -116,7 +116,9 @@ void memdelete(T *p_class) {
 
 	if (!predelete_handler(p_class))
 		return; // doesn't want to be deleted
-	p_class->~T();
+	if (!__has_trivial_destructor(T))
+		p_class->~T();
+
 	Memory::free_static(p_class, false);
 }
 
@@ -125,7 +127,9 @@ void memdelete_allocator(T *p_class) {
 
 	if (!predelete_handler(p_class))
 		return; // doesn't want to be deleted
-	p_class->~T();
+	if (!__has_trivial_destructor(T))
+		p_class->~T();
+
 	A::free(p_class);
 }
 
@@ -150,11 +154,13 @@ T *memnew_arr_template(size_t p_elements, const char *p_descr = "") {
 	ERR_FAIL_COND_V(!mem, failptr);
 	*(mem - 1) = p_elements;
 
-	T *elems = (T *)mem;
+	if (!__has_trivial_constructor(T)) {
+		T *elems = (T *)mem;
 
-	/* call operator new */
-	for (size_t i = 0; i < p_elements; i++) {
-		new (&elems[i], sizeof(T), p_descr) T;
+		/* call operator new */
+		for (size_t i = 0; i < p_elements; i++) {
+			new (&elems[i], sizeof(T), p_descr) T;
+		}
 	}
 
 	return (T *)mem;
@@ -177,12 +183,14 @@ void memdelete_arr(T *p_class) {
 
 	uint64_t *ptr = (uint64_t *)p_class;
 
-	uint64_t elem_count = *(ptr - 1);
+	if (!__has_trivial_destructor(T)) {
+		uint64_t elem_count = *(ptr - 1);
 
-	for (uint64_t i = 0; i < elem_count; i++) {
+		for (uint64_t i = 0; i < elem_count; i++) {
+			p_class[i].~T();
+		}
+	}
 
-		p_class[i].~T();
-	};
 	Memory::free_static(ptr, true);
 }