Floating-point to byte conversions are unchanged

[joelawryshort]

Type of issue

Missing information

Description

This change was made to standardize all floating point-to-integer conversions

It looks like floating point to byte/sbyte conversions are unaffected in .NET 9 (see fiddle). Should this be mentioned explicitly?

Page URL

https://learn.microsoft.com/en-us/dotnet/core/compatibility/jit/9.0/fp-to-integer

Content source URL

https://github.com/dotnet/docs/blob/main/docs/core/compatibility/jit/9.0/fp-to-integer.md

Document Version Independent Id

94cb081e-0c73-76b9-c756-ea5f779c5c28

Article author

@CamSoper

Metadata

• ID: 49ffed11-e4a2-7096-8a2f-e5490dda89dd
• Service: dotnet-fundamentals

Related Issues

[joelawryshort]

I was provided with more information about this here.

Conversion behaviour from floating-point types to "small" integral types (e.g. byte, short) is undefined, and is therefore not guaranteed to use the new saturating behaviour. Depending on the platform, floating-point numbers are often first converted to 32-bit integers before being converted to the target "small" integral type.

On the platforms I have tested, the first step now uses saturating arithmetic, but the second part remains truncating, resulting in mixed behaviour above and below the boundary of the max value of a 32-bit integer. For example the double value of 256001.0 truncates to a byte value of 1, but any value greater than or equal to 2147483647.0 saturates to 255.

Edit: added detail and precision about the behaviour being undefined.