The hard float instructions are being emulated by software in the kernel. Usually generating hard float instructions in this way is a bad idea, because every float instruction results in a trap to the kernel. This makes floating point very slow. Usually platforms that lack an FPU provide an OS and toolchain that support soft float, and this is what CDC supports on ARM platforms that lack an FPU.
[Message sent by forum member 'cjplummer' (cjplummer)]