A Case Study of Performance Degradation Attributable to Run-Time Bounds Checks on C++ Vector Access

Area of Science:

• Computer Science
• Software Engineering
• Performance Optimization

Background:

• Programmers often disable run-time safety checks, such as array bounds checking, to avoid perceived performance impacts.
• This omission can lead to security vulnerabilities and software quality issues.
• Array index out-of-bounds errors are a common programming mistake.

Purpose of the Study:

• To quantify the performance overhead of implementing array bounds-checking in C and C++.
• To evaluate the performance penalty of two distinct bounds-checking implementations.
• To assess the feasibility of incorporating safety checks without compromising application speed.

Main Methods:

• A benchmark was designed featuring a loop that sequentially writes to array elements.
• Performance was measured on a modern desktop PC with a superscalar CPU.
• Two different bounds-checking implementations in C and C++ were tested and compared.

Main Results:

• A standard bounds-checking access method in C++ showed a mean performance degradation of only (0.881 ± 0.009)% compared to the fastest unchecked access.
• The performance penalty was minimal, suggesting that safety checks do not necessarily lead to significant slowdowns.
• The study highlights the low cost of implementing basic safety checks.

Conclusions:

• The performance cost of array bounds-checking is negligible in many common scenarios.
• This case study indicates that safety checks can be integrated into C++ applications without substantial performance loss.
• Further research is recommended across diverse use cases, languages, and environments to fully understand the trade-offs between performance, security, and software quality.