📄 softfloat.c
字号:
/*
===============================================================================
This C source file is part of the SoftFloat IEC/IEEE Floating-point
Arithmetic Package, Release 2.
Written by John R. Hauser. This work was made possible in part by the
International Computer Science Institute, located at Suite 600, 1947 Center
Street, Berkeley, California 94704. Funding was partially provided by the
National Science Foundation under grant MIP-9311980. The original version
of this code was written as part of a project to build a fixed-point vector
processor in collaboration with the University of California at Berkeley,
overseen by Profs. Nelson Morgan and John Wawrzynek. More information
is available through the web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
arithmetic/softfloat.html'.
THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort
has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO
PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
Derivative works are acceptable, even for commercial purposes, so long as
(1) they include prominent notice that the work is derivative, and (2) they
include prominent notice akin to these three paragraphs for those parts of
this code that are retained.
===============================================================================
*/
#include "milieu.h"
#include "softfloat.h"
/*
-------------------------------------------------------------------------------
Floating-point rounding mode, extended double-precision rounding precision,
and exception flags.
-------------------------------------------------------------------------------
*/
int8 float_rounding_mode = float_round_nearest_even;
int8 floatx80_rounding_precision = 80;
int8 float_exception_flags;
/*
-------------------------------------------------------------------------------
Primitive arithmetic functions, including multi-word arithmetic, and
division and square root approximations. (Can be specialized to target if
desired.)
-------------------------------------------------------------------------------
*/
#include "softfloat-macros"
/*
-------------------------------------------------------------------------------
Functions and definitions to determine: (1) whether tininess for underflow
is detected before or after rounding by default, (2) what (if anything)
happens when exceptions are raised, (3) how signaling NaNs are distinguished
from quiet NaNs, (4) the default generated quiet NaNs, and (5) how NaNs
are propagated from function inputs to output. These details are target-
specific.
-------------------------------------------------------------------------------
*/
#include "softfloat-specialize"
/*
-------------------------------------------------------------------------------
Takes a 64-bit fixed-point value `absZ' with binary point between bits 6
and 7, and returns the properly rounded 32-bit integer corresponding to the
input. If `zSign' is nonzero, the input is negated before being converted
to an integer. Bit 63 of `absZ' must be zero. Ordinarily, the fixed-point
input is simply rounded to an integer, with the inexact exception raised if
the input cannot be represented exactly as an integer. If the fixed-point
input is too large, however, the invalid exception is raised and the largest
positive or negative integer is returned.
-------------------------------------------------------------------------------
*/
static int32 roundAndPackInt32( flag zSign, bits64 absZ )
{
int8 roundingMode;
flag roundNearestEven;
int8 roundIncrement, roundBits;
int32 z;
roundingMode = float_rounding_mode;
roundNearestEven = ( roundingMode == float_round_nearest_even );
roundIncrement = 0x40;
if ( ! roundNearestEven ) {
if ( roundingMode == float_round_to_zero ) {
roundIncrement = 0;
}
else {
roundIncrement = 0x7F;
if ( zSign ) {
if ( roundingMode == float_round_up ) roundIncrement = 0;
}
else {
if ( roundingMode == float_round_down ) roundIncrement = 0;
}
}
}
roundBits = absZ & 0x7F;
absZ = ( absZ + roundIncrement )>>7;
absZ &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven );
z = absZ;
if ( zSign ) z = - z;
if ( ( absZ>>32 ) || ( z && ( ( z < 0 ) ^ zSign ) ) ) {
float_exception_flags |= float_flag_invalid;
return zSign ? 0x80000000 : 0x7FFFFFFF;
}
if ( roundBits ) float_exception_flags |= float_flag_inexact;
return z;
}
/*
-------------------------------------------------------------------------------
Returns the fraction bits of the single-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE bits32 extractFloat32Frac( float32 a )
{
return a & 0x007FFFFF;
}
/*
-------------------------------------------------------------------------------
Returns the exponent bits of the single-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE int16 extractFloat32Exp( float32 a )
{
return ( a>>23 ) & 0xFF;
}
/*
-------------------------------------------------------------------------------
Returns the sign bit of the single-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE flag extractFloat32Sign( float32 a )
{
return a>>31;
}
/*
-------------------------------------------------------------------------------
Normalizes the subnormal single-precision floating-point value represented
by the denormalized significand `aSig'. The normalized exponent and
significand are stored at the locations pointed to by `zExpPtr' and
`zSigPtr', respectively.
-------------------------------------------------------------------------------
*/
static void
normalizeFloat32Subnormal( bits32 aSig, int16 *zExpPtr, bits32 *zSigPtr )
{
int8 shiftCount;
shiftCount = countLeadingZeros32( aSig ) - 8;
*zSigPtr = aSig<<shiftCount;
*zExpPtr = 1 - shiftCount;
}
/*
-------------------------------------------------------------------------------
Packs the sign `zSign', exponent `zExp', and significand `zSig' into a
single-precision floating-point value, returning the result. After being
shifted into the proper positions, the three fields are simply added
together to form the result. This means that any integer portion of `zSig'
will be added into the exponent. Since a properly normalized significand
will have an integer portion equal to 1, the `zExp' input should be 1 less
than the desired result exponent whenever `zSig' is a complete, normalized
significand.
-------------------------------------------------------------------------------
*/
INLINE float32 packFloat32( flag zSign, int16 zExp, bits32 zSig )
{
#if 0
float32 f;
__asm__("@ packFloat32;
mov %0, %1, asl #31;
orr %0, %2, asl #23;
orr %0, %3"
: /* no outputs */
: "g" (f), "g" (zSign), "g" (zExp), "g" (zSig)
: "cc");
return f;
#else
return ( ( (bits32) zSign )<<31 ) + ( ( (bits32) zExp )<<23 ) + zSig;
#endif
}
/*
-------------------------------------------------------------------------------
Takes an abstract floating-point value having sign `zSign', exponent `zExp',
and significand `zSig', and returns the proper single-precision floating-
point value corresponding to the abstract input. Ordinarily, the abstract
value is simply rounded and packed into the single-precision format, with
the inexact exception raised if the abstract input cannot be represented
exactly. If the abstract value is too large, however, the overflow and
inexact exceptions are raised and an infinity or maximal finite value is
returned. If the abstract value is too small, the input value is rounded to
a subnormal number, and the underflow and inexact exceptions are raised if
the abstract input cannot be represented exactly as a subnormal single-
precision floating-point number.
The input significand `zSig' has its binary point between bits 30
and 29, which is 7 bits to the left of the usual location. This shifted
significand must be normalized or smaller. If `zSig' is not normalized,
`zExp' must be 0; in that case, the result returned is a subnormal number,
and it must not require rounding. In the usual case that `zSig' is
normalized, `zExp' must be 1 less than the ``true'' floating-point exponent.
The handling of underflow and overflow follows the IEC/IEEE Standard for
Binary Floating-point Arithmetic.
-------------------------------------------------------------------------------
*/
static float32 roundAndPackFloat32( flag zSign, int16 zExp, bits32 zSig )
{
int8 roundingMode;
flag roundNearestEven;
int8 roundIncrement, roundBits;
flag isTiny;
roundingMode = float_rounding_mode;
roundNearestEven = ( roundingMode == float_round_nearest_even );
roundIncrement = 0x40;
if ( ! roundNearestEven ) {
if ( roundingMode == float_round_to_zero ) {
roundIncrement = 0;
}
else {
roundIncrement = 0x7F;
if ( zSign ) {
if ( roundingMode == float_round_up ) roundIncrement = 0;
}
else {
if ( roundingMode == float_round_down ) roundIncrement = 0;
}
}
}
roundBits = zSig & 0x7F;
if ( 0xFD <= (bits16) zExp ) {
if ( ( 0xFD < zExp )
|| ( ( zExp == 0xFD )
&& ( (sbits32) ( zSig + roundIncrement ) < 0 ) )
) {
float_raise( float_flag_overflow | float_flag_inexact );
return packFloat32( zSign, 0xFF, 0 ) - ( roundIncrement == 0 );
}
if ( zExp < 0 ) {
isTiny =
( float_detect_tininess == float_tininess_before_rounding )
|| ( zExp < -1 )
|| ( zSig + roundIncrement < 0x80000000 );
shift32RightJamming( zSig, - zExp, &zSig );
zExp = 0;
roundBits = zSig & 0x7F;
if ( isTiny && roundBits ) float_raise( float_flag_underflow );
}
}
if ( roundBits ) float_exception_flags |= float_flag_inexact;
zSig = ( zSig + roundIncrement )>>7;
zSig &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven );
if ( zSig == 0 ) zExp = 0;
return packFloat32( zSign, zExp, zSig );
}
/*
-------------------------------------------------------------------------------
Takes an abstract floating-point value having sign `zSign', exponent `zExp',
and significand `zSig', and returns the proper single-precision floating-
point value corresponding to the abstract input. This routine is just like
`roundAndPackFloat32' except that `zSig' does not have to be normalized in
any way. In all cases, `zExp' must be 1 less than the ``true'' floating-
point exponent.
-------------------------------------------------------------------------------
*/
static float32
normalizeRoundAndPackFloat32( flag zSign, int16 zExp, bits32 zSig )
{
int8 shiftCount;
shiftCount = countLeadingZeros32( zSig ) - 1;
return roundAndPackFloat32( zSign, zExp - shiftCount, zSig<<shiftCount );
}
/*
-------------------------------------------------------------------------------
Returns the fraction bits of the double-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE bits64 extractFloat64Frac( float64 a )
{
return a & LIT64( 0x000FFFFFFFFFFFFF );
}
/*
-------------------------------------------------------------------------------
Returns the exponent bits of the double-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE int16 extractFloat64Exp( float64 a )
{
return ( a>>52 ) & 0x7FF;
}
/*
-------------------------------------------------------------------------------
Returns the sign bit of the double-precision floating-point value `a'.
-------------------------------------------------------------------------------
*/
INLINE flag extractFloat64Sign( float64 a )
{
return a>>63;
}
/*
-------------------------------------------------------------------------------
Normalizes the subnormal double-precision floating-point value represented
by the denormalized significand `aSig'. The normalized exponent and
significand are stored at the locations pointed to by `zExpPtr' and
`zSigPtr', respectively.
-------------------------------------------------------------------------------
*/
static void
normalizeFloat64Subnormal( bits64 aSig, int16 *zExpPtr, bits64 *zSigPtr )
{
int8 shiftCount;
shiftCount = countLeadingZeros64( aSig ) - 11;
*zSigPtr = aSig<<shiftCount;
*zExpPtr = 1 - shiftCount;
}
/*
-------------------------------------------------------------------------------
Packs the sign `zSign', exponent `zExp', and significand `zSig' into a
double-precision floating-point value, returning the result. After being
shifted into the proper positions, the three fields are simply added
together to form the result. This means that any integer portion of `zSig'
will be added into the exponent. Since a properly normalized significand
will have an integer portion equal to 1, the `zExp' input should be 1 less
than the desired result exponent whenever `zSig' is a complete, normalized
significand.
-------------------------------------------------------------------------------
*/
INLINE float64 packFloat64( flag zSign, int16 zExp, bits64 zSig )
{
return ( ( (bits64) zSign )<<63 ) + ( ( (bits64) zExp )<<52 ) + zSig;
}
/*
-------------------------------------------------------------------------------
Takes an abstract floating-point value having sign `zSign', exponent `zExp',
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -