https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60410
--- Comment #20 from M.Schewe <m.schewe at apt dot de> --- Removing -fshort-double completely was not a good idea. It is required for ARM microcontrollers, e.g. ARM Cortex M4F with a FPU which can handly only single precision arithmetic operations. All double precision arithmetics must be emulated by slow integer software. Without this option it is no more possible to compile runtime libraries and application to efficiently use this FPU. Restricing resolution to single precision is acceptable for many embedded realtime applications, often runtime requirements must be met. I agree to you that on architectures with powerful FPUs like x86 and x86_64 this option is not so important and may be unsupported to solve bugs. I suggest to revert the changes from rev. 233218 thus implementing -fshort-double as before. Architectures which have implementation-bugs like x86_64-*-* simply can abort compilation if the option is activated with some error message saying "-fshort-double option is no more supported on this archeticture".
