Are Virtual Functions in C a Performance Bottleneck?

Performance Implications of Virtual Functions in C

You expressed concerns about the performance impact of virtual functions in your class design. Is this a valid concern, or can it be considered premature optimization?

To address your question, we conducted performance tests on a 3GHz PowerPC CPU. We created a simple 4D vector class with get/set functions. We ran tests with the functions defined as inline, virtual, and regular function calls.

Here are the results:

• Inline: 8ms (0.65ns per call)
• Direct: 68ms (5.53ns per call)
• Virtual: 160ms (13ns per call)

In this case, with the data fitting in L1 cache, virtual function calls were approximately 20 times slower than inline calls. However, it's crucial to consider the context of this performance difference.

Each iteration of the test loop involved 12,288 function calls. Therefore, the virtual loop took 92ms longer than the direct loop, resulting in an additional overhead of just 7 nanoseconds per function call.

Based on these results, we conclude that:

• Virtual functions are indeed slower than direct functions.
• The performance penalty is negligible unless you intend to call them extremely frequently (on the order of millions of times per second).

In most situations, unless optimizing for extreme performance, the trade-offs associated with using virtual functions for encapsulation and flexibility outweigh any potential performance concerns.

The above is the detailed content of Are Virtual Functions in C a Performance Bottleneck?. For more information, please follow other related articles on the PHP Chinese website!