EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Jul 30·5 min read
C++ float-to-int conversion can be undefined behavior
TL;DR
C++ float-to-int conversion can lead to undefined behavior if the float value exceeds the range of the destination integer type. Many developers inadvertently write code that triggers this issue without any compiler warnings.
✦ Why It Matters
Review your code for float-to-int conversions to prevent undefined behavior in your applications.
Key Takeaways
- Always check bounds before converting float to int.
- Use gsl::narrow cautiously; it doesn't handle float-to-int safely.
- Test your code with Undefined Behavior Sanitizer to catch issues.
- Understand that undefined behavior can lead to inconsistent results across platforms.
How It Works
C++ allows conversion from floating-point types to integer types, truncating the fractional part. However, if the float value exceeds the integer's range, the behavior is undefined, which can lead to unpredictable results.