mirror of
git://sourceware.org/git/newlib-cygwin.git
synced 2025-01-15 11:00:04 +08:00
c156098271
Based on code from https://github.com/ARM-software/optimized-routines/ This patch adds a highly optimized generic implementation of expf, exp2f, logf, log2f and powf. The new functions are not only faster (6x for powf!), but are also smaller and more accurate. In order to achieve this, the algorithm uses double precision arithmetic for accuracy, avoids divisions and uses small table lookups to minimize the polynomials. Special cases are handled inline to avoid the unnecessary overhead of wrapper functions and set errno to POSIX requirements. The new functions are added under newlib/libm/common, but the old implementations are kept (in newlib/libm/math) for non-IEEE or pre-C99 systems. Targets can enable the new math code by defining __OBSOLETE_MATH_DEFAULT to 0 in newlib/libc/include/machine/ieeefp.h, users can override the default by defining __OBSOLETE_MATH. Currently the new code is enabled for AArch64 and AArch32 with VFP. Targets with a single precision FPU may still prefer the old implementation. libm.a size changes: arm: -1692 arm/thumb/v7-a/nofp: -878 arm/thumb/v7-a+fp/hard: -864 arm/thumb/v7-a+fp/softfp: -908 aarch64: -1476
102 lines
2.8 KiB
C
102 lines
2.8 KiB
C
/* ef_exp.c -- float version of e_exp.c.
|
|
* Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
|
|
*/
|
|
|
|
/*
|
|
* ====================================================
|
|
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
|
|
*
|
|
* Developed at SunPro, a Sun Microsystems, Inc. business.
|
|
* Permission to use, copy, modify, and distribute this
|
|
* software is freely granted, provided that this notice
|
|
* is preserved.
|
|
* ====================================================
|
|
*/
|
|
|
|
#include "fdlibm.h"
|
|
|
|
#if __OBSOLETE_MATH
|
|
#ifdef __v810__
|
|
#define const
|
|
#endif
|
|
|
|
#ifdef __STDC__
|
|
static const float
|
|
#else
|
|
static float
|
|
#endif
|
|
one = 1.0,
|
|
halF[2] = {0.5,-0.5,},
|
|
huge = 1.0e+30,
|
|
twom100 = 7.8886090522e-31, /* 2**-100=0x0d800000 */
|
|
ln2HI[2] ={ 6.9313812256e-01, /* 0x3f317180 */
|
|
-6.9313812256e-01,}, /* 0xbf317180 */
|
|
ln2LO[2] ={ 9.0580006145e-06, /* 0x3717f7d1 */
|
|
-9.0580006145e-06,}, /* 0xb717f7d1 */
|
|
invln2 = 1.4426950216e+00, /* 0x3fb8aa3b */
|
|
P1 = 1.6666667163e-01, /* 0x3e2aaaab */
|
|
P2 = -2.7777778450e-03, /* 0xbb360b61 */
|
|
P3 = 6.6137559770e-05, /* 0x388ab355 */
|
|
P4 = -1.6533901999e-06, /* 0xb5ddea0e */
|
|
P5 = 4.1381369442e-08; /* 0x3331bb4c */
|
|
|
|
#ifdef __STDC__
|
|
float __ieee754_expf(float x) /* default IEEE double exp */
|
|
#else
|
|
float __ieee754_expf(x) /* default IEEE double exp */
|
|
float x;
|
|
#endif
|
|
{
|
|
float y,hi,lo,c,t;
|
|
__int32_t k = 0,xsb,sx;
|
|
__uint32_t hx;
|
|
|
|
GET_FLOAT_WORD(sx,x);
|
|
xsb = (sx>>31)&1; /* sign bit of x */
|
|
hx = sx & 0x7fffffff; /* high word of |x| */
|
|
|
|
/* filter out non-finite argument */
|
|
if(FLT_UWORD_IS_NAN(hx))
|
|
return x+x; /* NaN */
|
|
if(FLT_UWORD_IS_INFINITE(hx))
|
|
return (xsb==0)? x:0.0; /* exp(+-inf)={inf,0} */
|
|
if(sx > FLT_UWORD_LOG_MAX)
|
|
return huge*huge; /* overflow */
|
|
if(sx < 0 && hx > FLT_UWORD_LOG_MIN)
|
|
return twom100*twom100; /* underflow */
|
|
|
|
/* argument reduction */
|
|
if(hx > 0x3eb17218) { /* if |x| > 0.5 ln2 */
|
|
if(hx < 0x3F851592) { /* and |x| < 1.5 ln2 */
|
|
hi = x-ln2HI[xsb]; lo=ln2LO[xsb]; k = 1-xsb-xsb;
|
|
} else {
|
|
k = invln2*x+halF[xsb];
|
|
t = k;
|
|
hi = x - t*ln2HI[0]; /* t*ln2HI is exact here */
|
|
lo = t*ln2LO[0];
|
|
}
|
|
x = hi - lo;
|
|
}
|
|
else if(hx < 0x31800000) { /* when |x|<2**-28 */
|
|
if(huge+x>one) return one+x;/* trigger inexact */
|
|
}
|
|
|
|
/* x is now in primary range */
|
|
t = x*x;
|
|
c = x - t*(P1+t*(P2+t*(P3+t*(P4+t*P5))));
|
|
if(k==0) return one-((x*c)/(c-(float)2.0)-x);
|
|
else y = one-((lo-(x*c)/((float)2.0-c))-hi);
|
|
if(k >= -125) {
|
|
__uint32_t hy;
|
|
GET_FLOAT_WORD(hy,y);
|
|
SET_FLOAT_WORD(y,hy+(k<<23)); /* add k to y's exponent */
|
|
return y;
|
|
} else {
|
|
__uint32_t hy;
|
|
GET_FLOAT_WORD(hy,y);
|
|
SET_FLOAT_WORD(y,hy+((k+100)<<23)); /* add k to y's exponent */
|
|
return y*twom100;
|
|
}
|
|
}
|
|
#endif /* __OBSOLETE_MATH */
|