diff --git a/system/lib/libc/README.md b/system/lib/libc/README.md index f26571c6b44e7..c0c53c0152d60 100644 --- a/system/lib/libc/README.md +++ b/system/lib/libc/README.md @@ -13,7 +13,6 @@ Some changes have been made to the version that was taken from upstream, includi * Simplify stdout stream handling: do not support seeking, terminal handling, etc., as it just increases code size and Emscripten doesn't have those features anyhow. * Setting `_POSIX_REALTIME_SIGNALS` and `_POSIX_SPAWN` macros to -1, to exclude unsupported functions. -Backported src/stdio/vswprintf.c from 1.1.23 to fix #9305. -Backported src/string/{memccpy,memchr,memmove,stpcpy,stpncpy,strchrnul,strlcpy,strlen}.c from 1.2.0 to fix #7279. -Backported src/internal/floatscan.c to latest, see #11445. -Backported src/linux/gettid.c +Copy log.c and log2.c from ealier version of musl which result in smaller +binary size since they do not rely data tables in log_data.c and log2_data.c. +See /~https://github.com/emscripten-core/emscripten/issues/15483. diff --git a/system/lib/libc/musl/src/math/log2_small.c b/system/lib/libc/musl/src/math/log2_small.c new file mode 100644 index 0000000000000..0aafad4b86c1c --- /dev/null +++ b/system/lib/libc/musl/src/math/log2_small.c @@ -0,0 +1,122 @@ +/* origin: FreeBSD /usr/src/lib/msun/src/e_log2.c */ +/* + * ==================================================== + * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. + * + * Developed at SunSoft, a Sun Microsystems, Inc. business. + * Permission to use, copy, modify, and distribute this + * software is freely granted, provided that this notice + * is preserved. + * ==================================================== + */ +/* + * Return the base 2 logarithm of x. See log.c for most comments. + * + * Reduce x to 2^k (1+f) and calculate r = log(1+f) - f + f*f/2 + * as in log.c, then combine and scale in extra precision: + * log2(x) = (f - f*f/2 + r)/log(2) + k + */ + +#include +#include + +static const double +ivln2hi = 1.44269504072144627571e+00, /* 0x3ff71547, 0x65200000 */ +ivln2lo = 1.67517131648865118353e-10, /* 0x3de705fc, 0x2eefa200 */ +Lg1 = 6.666666666666735130e-01, /* 3FE55555 55555593 */ +Lg2 = 3.999999999940941908e-01, /* 3FD99999 9997FA04 */ +Lg3 = 2.857142874366239149e-01, /* 3FD24924 94229359 */ +Lg4 = 2.222219843214978396e-01, /* 3FCC71C5 1D8E78AF */ +Lg5 = 1.818357216161805012e-01, /* 3FC74664 96CB03DE */ +Lg6 = 1.531383769920937332e-01, /* 3FC39A09 D078C69F */ +Lg7 = 1.479819860511658591e-01; /* 3FC2F112 DF3E5244 */ + +double log2(double x) +{ + union {double f; uint64_t i;} u = {x}; + double_t hfsq,f,s,z,R,w,t1,t2,y,hi,lo,val_hi,val_lo; + uint32_t hx; + int k; + + hx = u.i>>32; + k = 0; + if (hx < 0x00100000 || hx>>31) { + if (u.i<<1 == 0) + return -1/(x*x); /* log(+-0)=-inf */ + if (hx>>31) + return (x-x)/0.0; /* log(-#) = NaN */ + /* subnormal number, scale x up */ + k -= 54; + x *= 0x1p54; + u.f = x; + hx = u.i>>32; + } else if (hx >= 0x7ff00000) { + return x; + } else if (hx == 0x3ff00000 && u.i<<32 == 0) + return 0; + + /* reduce x into [sqrt(2)/2, sqrt(2)] */ + hx += 0x3ff00000 - 0x3fe6a09e; + k += (int)(hx>>20) - 0x3ff; + hx = (hx&0x000fffff) + 0x3fe6a09e; + u.i = (uint64_t)hx<<32 | (u.i&0xffffffff); + x = u.f; + + f = x - 1.0; + hfsq = 0.5*f*f; + s = f/(2.0+f); + z = s*s; + w = z*z; + t1 = w*(Lg2+w*(Lg4+w*Lg6)); + t2 = z*(Lg1+w*(Lg3+w*(Lg5+w*Lg7))); + R = t2 + t1; + + /* + * f-hfsq must (for args near 1) be evaluated in extra precision + * to avoid a large cancellation when x is near sqrt(2) or 1/sqrt(2). + * This is fairly efficient since f-hfsq only depends on f, so can + * be evaluated in parallel with R. Not combining hfsq with R also + * keeps R small (though not as small as a true `lo' term would be), + * so that extra precision is not needed for terms involving R. + * + * Compiler bugs involving extra precision used to break Dekker's + * theorem for spitting f-hfsq as hi+lo, unless double_t was used + * or the multi-precision calculations were avoided when double_t + * has extra precision. These problems are now automatically + * avoided as a side effect of the optimization of combining the + * Dekker splitting step with the clear-low-bits step. + * + * y must (for args near sqrt(2) and 1/sqrt(2)) be added in extra + * precision to avoid a very large cancellation when x is very near + * these values. Unlike the above cancellations, this problem is + * specific to base 2. It is strange that adding +-1 is so much + * harder than adding +-ln2 or +-log10_2. + * + * This uses Dekker's theorem to normalize y+val_hi, so the + * compiler bugs are back in some configurations, sigh. And I + * don't want to used double_t to avoid them, since that gives a + * pessimization and the support for avoiding the pessimization + * is not yet available. + * + * The multi-precision calculations for the multiplications are + * routine. + */ + + /* hi+lo = f - hfsq + s*(hfsq+R) ~ log(1+f) */ + hi = f - hfsq; + u.f = hi; + u.i &= (uint64_t)-1<<32; + hi = u.f; + lo = f - hi - hfsq + s*(hfsq+R); + + val_hi = hi*ivln2hi; + val_lo = (lo+hi)*ivln2lo + lo*ivln2hi; + + /* spadd(val_hi, val_lo, y), except for not using double_t: */ + y = k; + w = y + val_hi; + val_lo += (y - w) + val_hi; + val_hi = w; + + return val_lo + val_hi; +} diff --git a/system/lib/libc/musl/src/math/log_small.c b/system/lib/libc/musl/src/math/log_small.c new file mode 100644 index 0000000000000..e61e113d41af9 --- /dev/null +++ b/system/lib/libc/musl/src/math/log_small.c @@ -0,0 +1,118 @@ +/* origin: FreeBSD /usr/src/lib/msun/src/e_log.c */ +/* + * ==================================================== + * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. + * + * Developed at SunSoft, a Sun Microsystems, Inc. business. + * Permission to use, copy, modify, and distribute this + * software is freely granted, provided that this notice + * is preserved. + * ==================================================== + */ +/* log(x) + * Return the logarithm of x + * + * Method : + * 1. Argument Reduction: find k and f such that + * x = 2^k * (1+f), + * where sqrt(2)/2 < 1+f < sqrt(2) . + * + * 2. Approximation of log(1+f). + * Let s = f/(2+f) ; based on log(1+f) = log(1+s) - log(1-s) + * = 2s + 2/3 s**3 + 2/5 s**5 + ....., + * = 2s + s*R + * We use a special Remez algorithm on [0,0.1716] to generate + * a polynomial of degree 14 to approximate R The maximum error + * of this polynomial approximation is bounded by 2**-58.45. In + * other words, + * 2 4 6 8 10 12 14 + * R(z) ~ Lg1*s +Lg2*s +Lg3*s +Lg4*s +Lg5*s +Lg6*s +Lg7*s + * (the values of Lg1 to Lg7 are listed in the program) + * and + * | 2 14 | -58.45 + * | Lg1*s +...+Lg7*s - R(z) | <= 2 + * | | + * Note that 2s = f - s*f = f - hfsq + s*hfsq, where hfsq = f*f/2. + * In order to guarantee error in log below 1ulp, we compute log + * by + * log(1+f) = f - s*(f - R) (if f is not too large) + * log(1+f) = f - (hfsq - s*(hfsq+R)). (better accuracy) + * + * 3. Finally, log(x) = k*ln2 + log(1+f). + * = k*ln2_hi+(f-(hfsq-(s*(hfsq+R)+k*ln2_lo))) + * Here ln2 is split into two floating point number: + * ln2_hi + ln2_lo, + * where n*ln2_hi is always exact for |n| < 2000. + * + * Special cases: + * log(x) is NaN with signal if x < 0 (including -INF) ; + * log(+INF) is +INF; log(0) is -INF with signal; + * log(NaN) is that NaN with no signal. + * + * Accuracy: + * according to an error analysis, the error is always less than + * 1 ulp (unit in the last place). + * + * Constants: + * The hexadecimal values are the intended ones for the following + * constants. The decimal values may be used, provided that the + * compiler will convert from decimal to binary accurately enough + * to produce the hexadecimal values shown. + */ + +#include +#include + +static const double +ln2_hi = 6.93147180369123816490e-01, /* 3fe62e42 fee00000 */ +ln2_lo = 1.90821492927058770002e-10, /* 3dea39ef 35793c76 */ +Lg1 = 6.666666666666735130e-01, /* 3FE55555 55555593 */ +Lg2 = 3.999999999940941908e-01, /* 3FD99999 9997FA04 */ +Lg3 = 2.857142874366239149e-01, /* 3FD24924 94229359 */ +Lg4 = 2.222219843214978396e-01, /* 3FCC71C5 1D8E78AF */ +Lg5 = 1.818357216161805012e-01, /* 3FC74664 96CB03DE */ +Lg6 = 1.531383769920937332e-01, /* 3FC39A09 D078C69F */ +Lg7 = 1.479819860511658591e-01; /* 3FC2F112 DF3E5244 */ + +double log(double x) +{ + union {double f; uint64_t i;} u = {x}; + double_t hfsq,f,s,z,R,w,t1,t2,dk; + uint32_t hx; + int k; + + hx = u.i>>32; + k = 0; + if (hx < 0x00100000 || hx>>31) { + if (u.i<<1 == 0) + return -1/(x*x); /* log(+-0)=-inf */ + if (hx>>31) + return (x-x)/0.0; /* log(-#) = NaN */ + /* subnormal number, scale x up */ + k -= 54; + x *= 0x1p54; + u.f = x; + hx = u.i>>32; + } else if (hx >= 0x7ff00000) { + return x; + } else if (hx == 0x3ff00000 && u.i<<32 == 0) + return 0; + + /* reduce x into [sqrt(2)/2, sqrt(2)] */ + hx += 0x3ff00000 - 0x3fe6a09e; + k += (int)(hx>>20) - 0x3ff; + hx = (hx&0x000fffff) + 0x3fe6a09e; + u.i = (uint64_t)hx<<32 | (u.i&0xffffffff); + x = u.f; + + f = x - 1.0; + hfsq = 0.5*f*f; + s = f/(2.0+f); + z = s*s; + w = z*z; + t1 = w*(Lg2+w*(Lg4+w*Lg6)); + t2 = z*(Lg1+w*(Lg3+w*(Lg5+w*Lg7))); + R = t2 + t1; + dk = k; + return s*(hfsq+R) + dk*ln2_lo - hfsq + f + dk*ln2_hi; +} diff --git a/system/lib/libc/musl/src/math/pow_small.c b/system/lib/libc/musl/src/math/pow_small.c new file mode 100644 index 0000000000000..b66f632d8eea9 --- /dev/null +++ b/system/lib/libc/musl/src/math/pow_small.c @@ -0,0 +1,328 @@ +/* origin: FreeBSD /usr/src/lib/msun/src/e_pow.c */ +/* + * ==================================================== + * Copyright (C) 2004 by Sun Microsystems, Inc. All rights reserved. + * + * Permission to use, copy, modify, and distribute this + * software is freely granted, provided that this notice + * is preserved. + * ==================================================== + */ +/* pow(x,y) return x**y + * + * n + * Method: Let x = 2 * (1+f) + * 1. Compute and return log2(x) in two pieces: + * log2(x) = w1 + w2, + * where w1 has 53-24 = 29 bit trailing zeros. + * 2. Perform y*log2(x) = n+y' by simulating muti-precision + * arithmetic, where |y'|<=0.5. + * 3. Return x**y = 2**n*exp(y'*log2) + * + * Special cases: + * 1. (anything) ** 0 is 1 + * 2. 1 ** (anything) is 1 + * 3. (anything except 1) ** NAN is NAN + * 4. NAN ** (anything except 0) is NAN + * 5. +-(|x| > 1) ** +INF is +INF + * 6. +-(|x| > 1) ** -INF is +0 + * 7. +-(|x| < 1) ** +INF is +0 + * 8. +-(|x| < 1) ** -INF is +INF + * 9. -1 ** +-INF is 1 + * 10. +0 ** (+anything except 0, NAN) is +0 + * 11. -0 ** (+anything except 0, NAN, odd integer) is +0 + * 12. +0 ** (-anything except 0, NAN) is +INF, raise divbyzero + * 13. -0 ** (-anything except 0, NAN, odd integer) is +INF, raise divbyzero + * 14. -0 ** (+odd integer) is -0 + * 15. -0 ** (-odd integer) is -INF, raise divbyzero + * 16. +INF ** (+anything except 0,NAN) is +INF + * 17. +INF ** (-anything except 0,NAN) is +0 + * 18. -INF ** (+odd integer) is -INF + * 19. -INF ** (anything) = -0 ** (-anything), (anything except odd integer) + * 20. (anything) ** 1 is (anything) + * 21. (anything) ** -1 is 1/(anything) + * 22. (-anything) ** (integer) is (-1)**(integer)*(+anything**integer) + * 23. (-anything except 0 and inf) ** (non-integer) is NAN + * + * Accuracy: + * pow(x,y) returns x**y nearly rounded. In particular + * pow(integer,integer) + * always returns the correct integer provided it is + * representable. + * + * Constants : + * The hexadecimal values are the intended ones for the following + * constants. The decimal values may be used, provided that the + * compiler will convert from decimal to binary accurately enough + * to produce the hexadecimal values shown. + */ + +#include "libm.h" + +static const double +bp[] = {1.0, 1.5,}, +dp_h[] = { 0.0, 5.84962487220764160156e-01,}, /* 0x3FE2B803, 0x40000000 */ +dp_l[] = { 0.0, 1.35003920212974897128e-08,}, /* 0x3E4CFDEB, 0x43CFD006 */ +two53 = 9007199254740992.0, /* 0x43400000, 0x00000000 */ +huge = 1.0e300, +tiny = 1.0e-300, +/* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */ +L1 = 5.99999999999994648725e-01, /* 0x3FE33333, 0x33333303 */ +L2 = 4.28571428578550184252e-01, /* 0x3FDB6DB6, 0xDB6FABFF */ +L3 = 3.33333329818377432918e-01, /* 0x3FD55555, 0x518F264D */ +L4 = 2.72728123808534006489e-01, /* 0x3FD17460, 0xA91D4101 */ +L5 = 2.30660745775561754067e-01, /* 0x3FCD864A, 0x93C9DB65 */ +L6 = 2.06975017800338417784e-01, /* 0x3FCA7E28, 0x4A454EEF */ +P1 = 1.66666666666666019037e-01, /* 0x3FC55555, 0x5555553E */ +P2 = -2.77777777770155933842e-03, /* 0xBF66C16C, 0x16BEBD93 */ +P3 = 6.61375632143793436117e-05, /* 0x3F11566A, 0xAF25DE2C */ +P4 = -1.65339022054652515390e-06, /* 0xBEBBBD41, 0xC5D26BF1 */ +P5 = 4.13813679705723846039e-08, /* 0x3E663769, 0x72BEA4D0 */ +lg2 = 6.93147180559945286227e-01, /* 0x3FE62E42, 0xFEFA39EF */ +lg2_h = 6.93147182464599609375e-01, /* 0x3FE62E43, 0x00000000 */ +lg2_l = -1.90465429995776804525e-09, /* 0xBE205C61, 0x0CA86C39 */ +ovt = 8.0085662595372944372e-017, /* -(1024-log2(ovfl+.5ulp)) */ +cp = 9.61796693925975554329e-01, /* 0x3FEEC709, 0xDC3A03FD =2/(3ln2) */ +cp_h = 9.61796700954437255859e-01, /* 0x3FEEC709, 0xE0000000 =(float)cp */ +cp_l = -7.02846165095275826516e-09, /* 0xBE3E2FE0, 0x145B01F5 =tail of cp_h*/ +ivln2 = 1.44269504088896338700e+00, /* 0x3FF71547, 0x652B82FE =1/ln2 */ +ivln2_h = 1.44269502162933349609e+00, /* 0x3FF71547, 0x60000000 =24b 1/ln2*/ +ivln2_l = 1.92596299112661746887e-08; /* 0x3E54AE0B, 0xF85DDF44 =1/ln2 tail*/ + +double pow(double x, double y) +{ + double z,ax,z_h,z_l,p_h,p_l; + double y1,t1,t2,r,s,t,u,v,w; + int32_t i,j,k,yisint,n; + int32_t hx,hy,ix,iy; + uint32_t lx,ly; + + EXTRACT_WORDS(hx, lx, x); + EXTRACT_WORDS(hy, ly, y); + ix = hx & 0x7fffffff; + iy = hy & 0x7fffffff; + + /* x**0 = 1, even if x is NaN */ + if ((iy|ly) == 0) + return 1.0; + /* 1**y = 1, even if y is NaN */ + if (hx == 0x3ff00000 && lx == 0) + return 1.0; + /* NaN if either arg is NaN */ + if (ix > 0x7ff00000 || (ix == 0x7ff00000 && lx != 0) || + iy > 0x7ff00000 || (iy == 0x7ff00000 && ly != 0)) + return x + y; + + /* determine if y is an odd int when x < 0 + * yisint = 0 ... y is not an integer + * yisint = 1 ... y is an odd int + * yisint = 2 ... y is an even int + */ + yisint = 0; + if (hx < 0) { + if (iy >= 0x43400000) + yisint = 2; /* even integer y */ + else if (iy >= 0x3ff00000) { + k = (iy>>20) - 0x3ff; /* exponent */ + if (k > 20) { + j = ly>>(52-k); + if ((j<<(52-k)) == ly) + yisint = 2 - (j&1); + } else if (ly == 0) { + j = iy>>(20-k); + if ((j<<(20-k)) == iy) + yisint = 2 - (j&1); + } + } + } + + /* special value of y */ + if (ly == 0) { + if (iy == 0x7ff00000) { /* y is +-inf */ + if (((ix-0x3ff00000)|lx) == 0) /* (-1)**+-inf is 1 */ + return 1.0; + else if (ix >= 0x3ff00000) /* (|x|>1)**+-inf = inf,0 */ + return hy >= 0 ? y : 0.0; + else /* (|x|<1)**+-inf = 0,inf */ + return hy >= 0 ? 0.0 : -y; + } + if (iy == 0x3ff00000) { /* y is +-1 */ + if (hy >= 0) + return x; + y = 1/x; +#if FLT_EVAL_METHOD!=0 + { + union {double f; uint64_t i;} u = {y}; + uint64_t i = u.i & -1ULL/2; + if (i>>52 == 0 && (i&(i-1))) + FORCE_EVAL((float)y); + } +#endif + return y; + } + if (hy == 0x40000000) /* y is 2 */ + return x*x; + if (hy == 0x3fe00000) { /* y is 0.5 */ + if (hx >= 0) /* x >= +0 */ + return sqrt(x); + } + } + + ax = fabs(x); + /* special value of x */ + if (lx == 0) { + if (ix == 0x7ff00000 || ix == 0 || ix == 0x3ff00000) { /* x is +-0,+-inf,+-1 */ + z = ax; + if (hy < 0) /* z = (1/|x|) */ + z = 1.0/z; + if (hx < 0) { + if (((ix-0x3ff00000)|yisint) == 0) { + z = (z-z)/(z-z); /* (-1)**non-int is NaN */ + } else if (yisint == 1) + z = -z; /* (x<0)**odd = -(|x|**odd) */ + } + return z; + } + } + + s = 1.0; /* sign of result */ + if (hx < 0) { + if (yisint == 0) /* (x<0)**(non-int) is NaN */ + return (x-x)/(x-x); + if (yisint == 1) /* (x<0)**(odd int) */ + s = -1.0; + } + + /* |y| is huge */ + if (iy > 0x41e00000) { /* if |y| > 2**31 */ + if (iy > 0x43f00000) { /* if |y| > 2**64, must o/uflow */ + if (ix <= 0x3fefffff) + return hy < 0 ? huge*huge : tiny*tiny; + if (ix >= 0x3ff00000) + return hy > 0 ? huge*huge : tiny*tiny; + } + /* over/underflow if x is not close to one */ + if (ix < 0x3fefffff) + return hy < 0 ? s*huge*huge : s*tiny*tiny; + if (ix > 0x3ff00000) + return hy > 0 ? s*huge*huge : s*tiny*tiny; + /* now |1-x| is tiny <= 2**-20, suffice to compute + log(x) by x-x^2/2+x^3/3-x^4/4 */ + t = ax - 1.0; /* t has 20 trailing zeros */ + w = (t*t)*(0.5 - t*(0.3333333333333333333333-t*0.25)); + u = ivln2_h*t; /* ivln2_h has 21 sig. bits */ + v = t*ivln2_l - w*ivln2; + t1 = u + v; + SET_LOW_WORD(t1, 0); + t2 = v - (t1-u); + } else { + double ss,s2,s_h,s_l,t_h,t_l; + n = 0; + /* take care subnormal number */ + if (ix < 0x00100000) { + ax *= two53; + n -= 53; + GET_HIGH_WORD(ix,ax); + } + n += ((ix)>>20) - 0x3ff; + j = ix & 0x000fffff; + /* determine interval */ + ix = j | 0x3ff00000; /* normalize ix */ + if (j <= 0x3988E) /* |x|>1)|0x20000000) + 0x00080000 + (k<<18)); + t_l = ax - (t_h-bp[k]); + s_l = v*((u-s_h*t_h)-s_h*t_l); + /* compute log(ax) */ + s2 = ss*ss; + r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6))))); + r += s_l*(s_h+ss); + s2 = s_h*s_h; + t_h = 3.0 + s2 + r; + SET_LOW_WORD(t_h, 0); + t_l = r - ((t_h-3.0)-s2); + /* u+v = ss*(1+...) */ + u = s_h*t_h; + v = s_l*t_h + t_l*ss; + /* 2/(3log2)*(ss+...) */ + p_h = u + v; + SET_LOW_WORD(p_h, 0); + p_l = v - (p_h-u); + z_h = cp_h*p_h; /* cp_h+cp_l = 2/(3*log2) */ + z_l = cp_l*p_h+p_l*cp + dp_l[k]; + /* log2(ax) = (ss+..)*2/(3*log2) = n + dp_h + z_h + z_l */ + t = (double)n; + t1 = ((z_h + z_l) + dp_h[k]) + t; + SET_LOW_WORD(t1, 0); + t2 = z_l - (((t1 - t) - dp_h[k]) - z_h); + } + + /* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */ + y1 = y; + SET_LOW_WORD(y1, 0); + p_l = (y-y1)*t1 + y*t2; + p_h = y1*t1; + z = p_l + p_h; + EXTRACT_WORDS(j, i, z); + if (j >= 0x40900000) { /* z >= 1024 */ + if (((j-0x40900000)|i) != 0) /* if z > 1024 */ + return s*huge*huge; /* overflow */ + if (p_l + ovt > z - p_h) + return s*huge*huge; /* overflow */ + } else if ((j&0x7fffffff) >= 0x4090cc00) { /* z <= -1075 */ // FIXME: instead of abs(j) use unsigned j + if (((j-0xc090cc00)|i) != 0) /* z < -1075 */ + return s*tiny*tiny; /* underflow */ + if (p_l <= z - p_h) + return s*tiny*tiny; /* underflow */ + } + /* + * compute 2**(p_h+p_l) + */ + i = j & 0x7fffffff; + k = (i>>20) - 0x3ff; + n = 0; + if (i > 0x3fe00000) { /* if |z| > 0.5, set n = [z+0.5] */ + n = j + (0x00100000>>(k+1)); + k = ((n&0x7fffffff)>>20) - 0x3ff; /* new k for n */ + t = 0.0; + SET_HIGH_WORD(t, n & ~(0x000fffff>>k)); + n = ((n&0x000fffff)|0x00100000)>>(20-k); + if (j < 0) + n = -n; + p_h -= t; + } + t = p_l + p_h; + SET_LOW_WORD(t, 0); + u = t*lg2_h; + v = (p_l-(t-p_h))*lg2 + t*lg2_l; + z = u + v; + w = v - (z-u); + t = z*z; + t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5)))); + r = (z*t1)/(t1-2.0) - (w + z*w); + z = 1.0 - (r-z); + GET_HIGH_WORD(j, z); + j += n<<20; + if ((j>>20) <= 0) /* subnormal output */ + z = scalbn(z,n); + else + SET_HIGH_WORD(z, j); + return s*z; +} diff --git a/tools/system_libs.py b/tools/system_libs.py index f040aeeb47abe..ba074d2f14e64 100644 --- a/tools/system_libs.py +++ b/tools/system_libs.py @@ -86,7 +86,7 @@ def create_lib(libname, inputs): building.emar('cr', libname, inputs) -def get_wasm_libc_rt_files(): +def get_wasm_libc_rt_files(is_optz=False, all=False): # Combining static linking with LTO is tricky under LLVM. The codegen that # happens during LTO can generate references to new symbols that didn't exist # in the linker inputs themselves. @@ -102,29 +102,31 @@ def get_wasm_libc_rt_files(): # Note that this also includes things that may be depended on by those # functions - fmin uses signbit, for example, so signbit must be here (so if # fmin is added by codegen, it will have all it needs). - math_files = files_in_path( - path='system/lib/libc/musl/src/math', - filenames=[ - 'fmin.c', 'fminf.c', 'fminl.c', - 'fmax.c', 'fmaxf.c', 'fmaxl.c', - 'fmod.c', 'fmodf.c', 'fmodl.c', - 'log.c', 'log_data.c', - 'logf.c', 'logf_data.c', - 'log2.c', 'log2_data.c', - 'log2f.c', 'log2f_data.c', - 'log10.c', 'log10f.c', - 'exp.c', 'exp_data.c', - 'exp2.c', - 'exp2f.c', 'exp2f_data.c', - 'exp10.c', 'exp10f.c', - 'scalbn.c', '__fpclassifyl.c', - '__signbitl.c', '__signbitf.c', '__signbit.c', - '__math_divzero.c', '__math_divzerof.c', - '__math_oflow.c', '__math_oflowf.c', - '__math_uflow.c', '__math_uflowf.c', - '__math_invalid.c', '__math_invalidf.c', '__math_invalidl.c', - - ]) + math_files = [ + 'fmin.c', 'fminf.c', 'fminl.c', + 'fmax.c', 'fmaxf.c', 'fmaxl.c', + 'fmod.c', 'fmodf.c', 'fmodl.c', + 'logf.c', 'logf_data.c', + 'log2f.c', 'log2f_data.c', + 'log10.c', 'log10f.c', + 'exp.c', 'exp_data.c', + 'exp2.c', + 'exp2f.c', 'exp2f_data.c', + 'exp10.c', 'exp10f.c', + 'scalbn.c', '__fpclassifyl.c', + '__signbitl.c', '__signbitf.c', '__signbit.c', + '__math_divzero.c', '__math_divzerof.c', + '__math_oflow.c', '__math_oflowf.c', + '__math_uflow.c', '__math_uflowf.c', + '__math_invalid.c', '__math_invalidf.c', '__math_invalidl.c', + ] + if all or is_optz: + math_files += ['pow_small.c', 'log_small.c', 'log2_small.c'] + if all or not is_optz: + math_files += ['pow.c', 'pow_data.c', 'log.c', 'log_data.c', 'log2.c', 'log2_data.c'] + + math_files = files_in_path(path='system/lib/libc/musl/src/math', filenames=math_files) + other_files = files_in_path( path='system/lib/libc', filenames=['emscripten_memcpy.c', 'emscripten_memset.c', @@ -835,7 +837,7 @@ def get_files(self): ]) # These are included in wasm_libc_rt instead - ignore += [os.path.basename(f) for f in get_wasm_libc_rt_files()] + ignore += [os.path.basename(f) for f in get_wasm_libc_rt_files(all=True)] ignore = set(ignore) # TODO: consider using more math code from musl, doing so makes box2d faster @@ -1375,7 +1377,7 @@ class libc_rt(OptimizedAggressivelyForSizeLibrary, AsanInstrumentedLibrary, Comp name = 'libc_rt' def get_files(self): - return get_wasm_libc_rt_files() + return get_wasm_libc_rt_files(is_optz=self.is_optz) class libubsan_minimal_rt(CompilerRTLibrary, MTLibrary):