fp_contract

 

Determines whether floating-point contraction will take place.

#pragma fp_contract [ON | OFF]  

By default, fp_contract is ON.

For more information on floating-point behavior, see /fp (Specify Floating-Point Behavior).

Other floating-point pragmas include:

The code generated from this sample does not use the Fused Multiply Add (fma) instruction on Itanium processors. If you comment out #pragma fp_contract (off), the generated code will use the fma instruction.

// pragma_directive_fp_contract.cpp  
// compile with: /O2  
#include <stdio.h>  
#include <float.h>  
  
#pragma fp_contract (off)   
  
int main() {  
   double z, b, t;  
  
   for (int i = 0; i < 10; i++) {  
      b = i * 5.5;  
      t = i * 56.025;  
      _set_controlfp(_PC_24, _MCW_PC);  
  
      z = t * i + b;  
      printf_s ("out=%.15e\n", z);  
   }  
}  

out=0.000000000000000e+000  
out=6.152500152587891e+001  
out=2.351000061035156e+002  
out=5.207249755859375e+002  
out=9.184000244140625e+002  
out=1.428125000000000e+003  
out=2.049899902343750e+003  
out=2.783724853515625e+003  
out=3.629600097656250e+003  
out=4.587524902343750e+003  

Pragma Directives and the __Pragma Keyword

Show: