std::vector vs. Plain Arrays: Performance Evaluation

While it's commonly believed that std::vector operates similarly to arrays, recent tests have challenged this notion. In this article, we'll examine the performance differences between std::vector and plain arrays and shed light on the underlying reasons.

To conduct the tests, a benchmark was implemented that involved repeatedly creating and modifying large arrays of pixel objects. The tests were performed using two implementations: std::vector and raw pointers simulating plain arrays.

The initial results revealed a significant performance gap, with std::vector being around 3-4 times slower than plain arrays.

UseArray completed in 2.619 seconds
UseVector completed in 9.284 seconds
UseVectorPushBack completed in 14.669 seconds

This surprising finding prompted a closer examination of the code. It was discovered that the apparent performance difference stemmed from the fact that modifying an element in the vector required accessing it twice: once to resize the vector and again to initialize the element.

By optimizing the vector implementation to initialize each element only once, we obtained dramatically improved performance:

UseVector completed in 2.216 seconds

This revised performance is now comparable to that of plain arrays. It's important to note that the slight difference observed could be attributed to factors unrelated to the underlying data structure.

In conclusion, while std::vector is a convenient and versatile data structure, its performance can be influenced by factors such as the manner in which elements are initialized and accessed. When performance is critical, it's wise to carefully consider the specific requirements and explore alternative options such as plain arrays or other data structures.