The -fshort-double option controls the size of the
double type.
When building to the C99 standard, all floating-point types must be the
IEEE754 32-bit format. If you are building for C90 standard, you may use this option to
explicitly request the 24-bit form of this format for double objects. When
using the -fno-short-double form of the option, the
double type can be changed to the full 32-bit IEEE754 format. The
selection of this option must be consistent across all modules of the program.