How Can I Portably Adjust Floating-Point Rounding Behavior in C and Assembly?

Portably Adjusting Floating-Point Rounding Behavior

The need to modify the IEEE 754 rounding mode arises in various scenarios. This article elucidates efficient methods to achieve this, both in portable C and using x86 assembly language.

C Solution

C99 provides a comprehensive solution for manipulating rounding modes:

#include <fenv.h>
#pragma STDC FENV_ACCESS ON

int main() {
  // Preserve the initial rounding mode
  const int roundingMode = fegetround();

  // Set the desired rounding mode (towards zero)
  fesetround(FE_TOWARDZERO);

  // Perform floating-point operations...

  // Revert to the original rounding mode
  fesetround(roundingMode);

  return 0;
}</fenv.h>

x86 Assembly Solution

On platforms not supporting C99, x86 assembly offers an alternative:

x87 Floating-Point Unit: Use fldcw to adjust the rounding mode.

SSE Unit: Use ldmxcsr to configure the SIMD rounding mode.

MSVC-Specific Solution

Microsoft Visual C provides the non-standard _controlfp() function:

unsigned int roundingMode = _controlfp(0, 0);
_controlfp(_RC_CHOP, _MCW_RC);

// ...

_controlfp(roundingMode, _MCW_RC);

Rounding Mode Constants

The following table provides a reference for rounding mode constants:

Rounding Mode	C Name	MSVC Name
Nearest	FE_TONEAREST	_RC_NEAR
Towards Zero	FE_TOWARDZERO	_RC_CHOP
Positive Infinity	FE_UPWARD	_RC_UP
Negative Infinity	FE_DOWNWARD	_RC_DOWN

The above is the detailed content of How Can I Portably Adjust Floating-Point Rounding Behavior in C and Assembly?. For more information, please follow other related articles on the PHP Chinese website!