Hi everyone, with the increasingly common of floating-point hardware in recent NXP Kinetis parts including the T3.5, 3.6, 4X, I wonder whether anyone here has begun using binary data transfer of single-precision floating point values between hosts and Teensies, and if so, has it caused unexpected or subtle problems?

My project at work would benefit from transferring a variety of conversion constants between the embedded side and a host system. I would like to avoid the development, documentation and potential reliability costs of manually scaling these constants, instead counting on using of single-precision values. Our project does not depend on various corner cases like NaNs, denormals, etc.

My initial experiments show that it works, with the host-resident code using the python "struct" library module as well as C language code interpreting the same values in a C structure.

I am old enough to remember the challenges of interoperability with every manufacturer supporting a different floating-point format.