In our embedded project, targeted for the ARM Cortex M* cores, we are using the ARM clang-based toolchain (v6). We use float type for our floating-point math. As we try to minimize footprint, we found out that some space is taken by unnecessary double<->float conversions.
The source of some of those conversions was the use of literals not including the f specifier (e.g., use 12.34 instead of 12.34f). To eliminate that, I am looking for a compiler option that uses float type by default. Looking at the compiler documentation and at the general clang help, I could not find any.
Is there any option (command option, pragma, attribute, etc.) that can achieve that goal?