| title | fp_contract | Microsoft Docs | ||
|---|---|---|---|
| ms.custom | |||
| ms.date | 11/04/2016 | ||
| ms.reviewer | |||
| ms.suite | |||
| ms.technology |
|
||
| ms.tgt_pltfrm | |||
| ms.topic | article | ||
| f1_keywords |
|
||
| dev_langs |
|
||
| helpviewer_keywords |
|
||
| ms.assetid | 15b97338-6680-4287-ba2a-2dccc5b2ccf5 | ||
| caps.latest.revision | 12 | ||
| author | corob-msft | ||
| ms.author | corob | ||
| manager | ghogen |
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