Differences in the default floating point precision between Windows and Linux

I haven't written anything for a long time.
Floating point numbers have precision control on FPU processing. They can be single precision (24 bits in 32 bits), double precision (64 bits in length), and extended precision (80 bits in length and 64 bits in tail ). This is controlled by a single segment of the floating point control word.
In a Windows/VC environment (unconfirmed, maybe related to the compiler, such as Vc), the default precision is double precision.
In Linux/GCC, the default value is extended precision.
This can cause you to include floating point computing ProgramThere are some minor differences during transplantation.
Windows/VC can use the _ controlfp function to change the default settings, while Linux/GCC uses the _ fpu_setcw macro.
In addition, when dx is initialized, the floating point precision can be set to single precision. The interface should be implemented by changing the control word of FPU.