Performance testing in C++

By , last updated July 4, 2017

While developing a message-driven system we’ve come to a problem of what type of object creation/destruction uses less system resources: stack or heap. And although you maybe know the answer it’s always interesting to measure to be sure. We’ve done it with the help of unit-testing.

First a test object have been created:

class TestObject
{
	public:
		explicit TestObject(vector a, vector b) : mv1(a), mv2(b)
		{

		}

		vector mv1;
		vector mv2;
};

Another version of test object is using const reference for parameters:

class TestObject
{
	public:
		explicit TestObject(const vector &a, const vector&b) : mv1(a), mv2(b)
		{

		}

		vector mv1;
		vector mv2;
};

Now we test memory allocation for 1 million test objects and measure time:

void testAllocHeap()
{
	vector a(0,0,0);
	vector b(1,1,1);

	time_type startTime = local_time();
	for(int i=0; i<1000000; i++)
	{
		TestObject *pointerToTestObject = new TestObject(a, b);
		delete pointerToTestObject;
	}

	time_type stopTime = local_time();
	std::cout << "Time heap: " << stopTime-startTime <<"n";
}

void testAllocStack()
{
	vector a(0,0,0);
	vector b(1,1,1);

	time_type startTime = local_time();
	for(int i=0; i<1000000; i++)
	{
		TestObject pointerToTestObject(a, b);
	}
	time_type stopTime = local_time();
	std::cout << "Time stack: " << stopTime-startTime << "n";
}

Our measurements show that object creation as a const reference on stack is approximately 20 times more efficient then object allocation on heap.

Senior Software Engineer developing all kinds of stuff.

Comments

Be the first to comment.

Leave a Reply


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

*