Force single precision floating-point in generated c code

I am using the Embedded Coder to generate C code for a 32bit target which supports only single precision floats. In the model I obviously only use "single" signals and the generated code looks fine except three auto-generated files containing "rtGetInf" and "rtGetNaN" functions which still use "real_T". Is there any possibility to force "single" for the autogenerated functions?

[Richard Lobo]

Facing the same issue. Were you able to find a solution?

Answer 1

[Andy Bartlett]

Hi,

There is a Single Precision Converter tool that ships with Fixed-Point Designer that will automatically convert a subsystem or model to use single precision types. Watch this video.

In addition, Model Advisor has a check Identify questionable operations for strict single-precision design that will identify portions of a model or subsystem that are making undesired use of double precision floating point.

If you'd rather manage this manually, then one of the most important things is to set the Embedded Coder parameter

Standard math library      TargetLangStandard

Make sure it is NOT using the library from the super old C90 standard from back in 1989. Back then the standard library functions only supported double precision. In the standard from 1999, C99, single precision floating point versions of library functions were added. Please be sure to set TargetLangStandard to C99 to be able to utilize strict singles C library functions. The Single Precision Converter tool shown in the video will automatically handle this setting.

Regards Andy Bartlett

[Aymen Omrane]

Thanks for your answer. It worked for me

Answer 2

[Naini Dawar]

You can try to change the under-specified data type from double to single. Model Configuration Parameters >> Optimization >> Simulation and Code Generation Default for under-specified data type: Single