]>
Commit | Line | Data |
---|---|---|
2aa62f2b | 1 | /** @file\r |
2 | Compute the base 10 logrithm of x.\r | |
3 | \r | |
4 | Copyright (c) 2010 - 2011, Intel Corporation. All rights reserved.<BR>\r | |
5 | This program and the accompanying materials are licensed and made available under\r | |
6 | the terms and conditions of the BSD License that accompanies this distribution.\r | |
7 | The full text of the license may be found at\r | |
8 | http://opensource.org/licenses/bsd-license.\r | |
9 | \r | |
10 | THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,\r | |
11 | WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.\r | |
12 | \r | |
13 | * ====================================================\r | |
14 | * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.\r | |
15 | *\r | |
16 | * Developed at SunPro, a Sun Microsystems, Inc. business.\r | |
17 | * Permission to use, copy, modify, and distribute this\r | |
18 | * software is freely granted, provided that this notice\r | |
19 | * is preserved.\r | |
20 | * ====================================================\r | |
21 | \r | |
22 | e_pow.c 5.1 93/09/24\r | |
23 | NetBSD: e_pow.c,v 1.13 2004/06/30 18:43:15 drochner Exp\r | |
24 | **/\r | |
25 | #include <LibConfig.h>\r | |
26 | #include <sys/EfiCdefs.h>\r | |
27 | \r | |
28 | #if defined(_MSC_VER) /* Handle Microsoft VC++ compiler specifics. */\r | |
29 | // C4723: potential divide by zero.\r | |
30 | #pragma warning ( disable : 4723 )\r | |
31 | // C4756: overflow in constant arithmetic\r | |
32 | #pragma warning ( disable : 4756 )\r | |
33 | #endif\r | |
34 | \r | |
35 | /* __ieee754_pow(x,y) return x**y\r | |
36 | *\r | |
37 | * n\r | |
38 | * Method: Let x = 2 * (1+f)\r | |
39 | * 1. Compute and return log2(x) in two pieces:\r | |
40 | * log2(x) = w1 + w2,\r | |
41 | * where w1 has 53-24 = 29 bit trailing zeros.\r | |
42 | * 2. Perform y*log2(x) = n+y' by simulating multi-precision\r | |
43 | * arithmetic, where |y'|<=0.5.\r | |
44 | * 3. Return x**y = 2**n*exp(y'*log2)\r | |
45 | *\r | |
46 | * Special cases:\r | |
47 | * 1. (anything) ** 0 is 1\r | |
48 | * 2. (anything) ** 1 is itself\r | |
49 | * 3. (anything) ** NAN is NAN\r | |
50 | * 4. NAN ** (anything except 0) is NAN\r | |
51 | * 5. +-(|x| > 1) ** +INF is +INF\r | |
52 | * 6. +-(|x| > 1) ** -INF is +0\r | |
53 | * 7. +-(|x| < 1) ** +INF is +0\r | |
54 | * 8. +-(|x| < 1) ** -INF is +INF\r | |
55 | * 9. +-1 ** +-INF is NAN\r | |
56 | * 10. +0 ** (+anything except 0, NAN) is +0\r | |
57 | * 11. -0 ** (+anything except 0, NAN, odd integer) is +0\r | |
58 | * 12. +0 ** (-anything except 0, NAN) is +INF\r | |
59 | * 13. -0 ** (-anything except 0, NAN, odd integer) is +INF\r | |
60 | * 14. -0 ** (odd integer) = -( +0 ** (odd integer) )\r | |
61 | * 15. +INF ** (+anything except 0,NAN) is +INF\r | |
62 | * 16. +INF ** (-anything except 0,NAN) is +0\r | |
63 | * 17. -INF ** (anything) = -0 ** (-anything)\r | |
64 | * 18. (-anything) ** (integer) is (-1)**(integer)*(+anything**integer)\r | |
65 | * 19. (-anything except 0 and inf) ** (non-integer) is NAN\r | |
66 | *\r | |
67 | * Accuracy:\r | |
68 | * pow(x,y) returns x**y nearly rounded. In particular\r | |
69 | * pow(integer,integer)\r | |
70 | * always returns the correct integer provided it is\r | |
71 | * representable.\r | |
72 | *\r | |
73 | * Constants :\r | |
74 | * The hexadecimal values are the intended ones for the following\r | |
75 | * constants. The decimal values may be used, provided that the\r | |
76 | * compiler will convert from decimal to binary accurately enough\r | |
77 | * to produce the hexadecimal values shown.\r | |
78 | */\r | |
79 | \r | |
80 | #include "math.h"\r | |
81 | #include "math_private.h"\r | |
82 | #include <errno.h>\r | |
83 | \r | |
84 | static const double\r | |
85 | bp[] = {1.0, 1.5,},\r | |
86 | dp_h[] = { 0.0, 5.84962487220764160156e-01,}, /* 0x3FE2B803, 0x40000000 */\r | |
87 | dp_l[] = { 0.0, 1.35003920212974897128e-08,}, /* 0x3E4CFDEB, 0x43CFD006 */\r | |
88 | zero = 0.0,\r | |
89 | one = 1.0,\r | |
90 | two = 2.0,\r | |
91 | two53 = 9007199254740992.0, /* 0x43400000, 0x00000000 */\r | |
92 | huge = 1.0e300,\r | |
93 | tiny = 1.0e-300,\r | |
94 | /* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */\r | |
95 | L1 = 5.99999999999994648725e-01, /* 0x3FE33333, 0x33333303 */\r | |
96 | L2 = 4.28571428578550184252e-01, /* 0x3FDB6DB6, 0xDB6FABFF */\r | |
97 | L3 = 3.33333329818377432918e-01, /* 0x3FD55555, 0x518F264D */\r | |
98 | L4 = 2.72728123808534006489e-01, /* 0x3FD17460, 0xA91D4101 */\r | |
99 | L5 = 2.30660745775561754067e-01, /* 0x3FCD864A, 0x93C9DB65 */\r | |
100 | L6 = 2.06975017800338417784e-01, /* 0x3FCA7E28, 0x4A454EEF */\r | |
101 | P1 = 1.66666666666666019037e-01, /* 0x3FC55555, 0x5555553E */\r | |
102 | P2 = -2.77777777770155933842e-03, /* 0xBF66C16C, 0x16BEBD93 */\r | |
103 | P3 = 6.61375632143793436117e-05, /* 0x3F11566A, 0xAF25DE2C */\r | |
104 | P4 = -1.65339022054652515390e-06, /* 0xBEBBBD41, 0xC5D26BF1 */\r | |
105 | P5 = 4.13813679705723846039e-08, /* 0x3E663769, 0x72BEA4D0 */\r | |
106 | lg2 = 6.93147180559945286227e-01, /* 0x3FE62E42, 0xFEFA39EF */\r | |
107 | lg2_h = 6.93147182464599609375e-01, /* 0x3FE62E43, 0x00000000 */\r | |
108 | lg2_l = -1.90465429995776804525e-09, /* 0xBE205C61, 0x0CA86C39 */\r | |
109 | ovt = 8.0085662595372944372e-0017, /* -(1024-log2(ovfl+.5ulp)) */\r | |
110 | cp = 9.61796693925975554329e-01, /* 0x3FEEC709, 0xDC3A03FD =2/(3ln2) */\r | |
111 | cp_h = 9.61796700954437255859e-01, /* 0x3FEEC709, 0xE0000000 =(float)cp */\r | |
112 | cp_l = -7.02846165095275826516e-09, /* 0xBE3E2FE0, 0x145B01F5 =tail of cp_h*/\r | |
113 | ivln2 = 1.44269504088896338700e+00, /* 0x3FF71547, 0x652B82FE =1/ln2 */\r | |
114 | ivln2_h = 1.44269502162933349609e+00, /* 0x3FF71547, 0x60000000 =24b 1/ln2*/\r | |
115 | ivln2_l = 1.92596299112661746887e-08; /* 0x3E54AE0B, 0xF85DDF44 =1/ln2 tail*/\r | |
116 | \r | |
117 | double\r | |
118 | __ieee754_pow(double x, double y)\r | |
119 | {\r | |
120 | double z,ax,z_h,z_l,p_h,p_l;\r | |
121 | double y1,t1,t2,r,s,t,u,v,w;\r | |
122 | int32_t i,j,k,yisint,n;\r | |
123 | int32_t hx,hy,ix,iy;\r | |
124 | u_int32_t lx,ly;\r | |
125 | \r | |
126 | EXTRACT_WORDS(hx,lx,x);\r | |
127 | EXTRACT_WORDS(hy,ly,y);\r | |
128 | ix = hx&0x7fffffff; iy = hy&0x7fffffff;\r | |
129 | \r | |
130 | /* y==zero: x**0 = 1 */\r | |
131 | if((iy|ly)==0) return one;\r | |
132 | \r | |
133 | /* +-NaN return x+y */\r | |
134 | if(ix > 0x7ff00000 || ((ix==0x7ff00000)&&(lx!=0)) ||\r | |
135 | iy > 0x7ff00000 || ((iy==0x7ff00000)&&(ly!=0)))\r | |
136 | return x+y;\r | |
137 | \r | |
138 | /* determine if y is an odd int when x < 0\r | |
139 | * yisint = 0 ... y is not an integer\r | |
140 | * yisint = 1 ... y is an odd int\r | |
141 | * yisint = 2 ... y is an even int\r | |
142 | */\r | |
143 | yisint = 0;\r | |
144 | if(hx<0) {\r | |
145 | if(iy>=0x43400000) yisint = 2; /* even integer y */\r | |
146 | else if(iy>=0x3ff00000) {\r | |
147 | k = (iy>>20)-0x3ff; /* exponent */\r | |
148 | if(k>20) {\r | |
149 | j = ly>>(52-k);\r | |
150 | if((u_int32_t)(j<<(52-k))==ly) yisint = 2-(j&1);\r | |
151 | } else if(ly==0) {\r | |
152 | j = iy>>(20-k);\r | |
153 | if((j<<(20-k))==iy) yisint = 2-(j&1);\r | |
154 | }\r | |
155 | }\r | |
156 | }\r | |
157 | \r | |
158 | /* special value of y */\r | |
159 | if(ly==0) {\r | |
160 | if (iy==0x7ff00000) { /* y is +-inf */\r | |
161 | if(((ix-0x3ff00000)|lx)==0)\r | |
162 | return y - y; /* inf**+-1 is NaN */\r | |
163 | else if (ix >= 0x3ff00000)/* (|x|>1)**+-inf = inf,0 */\r | |
164 | return (hy>=0)? y: zero;\r | |
165 | else /* (|x|<1)**-,+inf = inf,0 */\r | |
166 | return (hy<0)?-y: zero;\r | |
167 | }\r | |
168 | if(iy==0x3ff00000) { /* y is +-1 */\r | |
169 | if(hy<0) return one/x; else return x;\r | |
170 | }\r | |
171 | if(hy==0x40000000) return x*x; /* y is 2 */\r | |
172 | if(hy==0x3fe00000) { /* y is 0.5 */\r | |
173 | if(hx>=0) /* x >= +0 */\r | |
174 | return __ieee754_sqrt(x);\r | |
175 | }\r | |
176 | }\r | |
177 | \r | |
178 | ax = fabs(x);\r | |
179 | /* special value of x */\r | |
180 | if(lx==0) {\r | |
181 | if(ix==0x7ff00000||ix==0||ix==0x3ff00000){\r | |
182 | z = ax; /*x is +-0,+-inf,+-1*/\r | |
183 | if(hy<0) z = one/z; /* z = (1/|x|) */\r | |
184 | if(hx<0) {\r | |
185 | if(((ix-0x3ff00000)|yisint)==0) {\r | |
186 | z = (z-z)/(z-z); /* (-1)**non-int is NaN */\r | |
187 | } else if(yisint==1)\r | |
188 | z = -z; /* (x<0)**odd = -(|x|**odd) */\r | |
189 | }\r | |
190 | return z;\r | |
191 | }\r | |
192 | }\r | |
193 | \r | |
194 | n = (hx>>31)+1;\r | |
195 | \r | |
196 | /* (x<0)**(non-int) is NaN */\r | |
197 | if((n|yisint)==0) {\r | |
198 | errno = EDOM;\r | |
199 | return (x-x)/(x-x);\r | |
200 | }\r | |
201 | \r | |
202 | s = one; /* s (sign of result -ve**odd) = -1 else = 1 */\r | |
203 | if((n|(yisint-1))==0) s = -one;/* (-ve)**(odd int) */\r | |
204 | \r | |
205 | /* |y| is huge */\r | |
206 | if(iy>0x41e00000) { /* if |y| > 2**31 */\r | |
207 | if(iy>0x43f00000){ /* if |y| > 2**64, must o/uflow */\r | |
208 | if(ix<=0x3fefffff) return (hy<0)? huge*huge:tiny*tiny;\r | |
209 | if(ix>=0x3ff00000) return (hy>0)? huge*huge:tiny*tiny;\r | |
210 | }\r | |
211 | /* over/underflow if x is not close to one */\r | |
212 | if(ix<0x3fefffff) return (hy<0)? s*huge*huge:s*tiny*tiny;\r | |
213 | if(ix>0x3ff00000) return (hy>0)? s*huge*huge:s*tiny*tiny;\r | |
214 | /* now |1-x| is tiny <= 2**-20, suffice to compute\r | |
215 | log(x) by x-x^2/2+x^3/3-x^4/4 */\r | |
216 | t = ax-one; /* t has 20 trailing zeros */\r | |
217 | w = (t*t)*(0.5-t*(0.3333333333333333333333-t*0.25));\r | |
218 | u = ivln2_h*t; /* ivln2_h has 21 sig. bits */\r | |
219 | v = t*ivln2_l-w*ivln2;\r | |
220 | t1 = u+v;\r | |
221 | SET_LOW_WORD(t1,0);\r | |
222 | t2 = v-(t1-u);\r | |
223 | } else {\r | |
224 | double ss,s2,s_h,s_l,t_h,t_l;\r | |
225 | n = 0;\r | |
226 | /* take care subnormal number */\r | |
227 | if(ix<0x00100000)\r | |
228 | {ax *= two53; n -= 53; GET_HIGH_WORD(ix,ax); }\r | |
229 | n += ((ix)>>20)-0x3ff;\r | |
230 | j = ix&0x000fffff;\r | |
231 | /* determine interval */\r | |
232 | ix = j|0x3ff00000; /* normalize ix */\r | |
233 | if(j<=0x3988E) k=0; /* |x|<sqrt(3/2) */\r | |
234 | else if(j<0xBB67A) k=1; /* |x|<sqrt(3) */\r | |
235 | else {k=0;n+=1;ix -= 0x00100000;}\r | |
236 | SET_HIGH_WORD(ax,ix);\r | |
237 | \r | |
238 | /* compute ss = s_h+s_l = (x-1)/(x+1) or (x-1.5)/(x+1.5) */\r | |
239 | u = ax-bp[k]; /* bp[0]=1.0, bp[1]=1.5 */\r | |
240 | v = one/(ax+bp[k]);\r | |
241 | ss = u*v;\r | |
242 | s_h = ss;\r | |
243 | SET_LOW_WORD(s_h,0);\r | |
244 | /* t_h=ax+bp[k] High */\r | |
245 | t_h = zero;\r | |
246 | SET_HIGH_WORD(t_h,((ix>>1)|0x20000000)+0x00080000+(k<<18));\r | |
247 | t_l = ax - (t_h-bp[k]);\r | |
248 | s_l = v*((u-s_h*t_h)-s_h*t_l);\r | |
249 | /* compute log(ax) */\r | |
250 | s2 = ss*ss;\r | |
251 | r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6)))));\r | |
252 | r += s_l*(s_h+ss);\r | |
253 | s2 = s_h*s_h;\r | |
254 | t_h = 3.0+s2+r;\r | |
255 | SET_LOW_WORD(t_h,0);\r | |
256 | t_l = r-((t_h-3.0)-s2);\r | |
257 | /* u+v = ss*(1+...) */\r | |
258 | u = s_h*t_h;\r | |
259 | v = s_l*t_h+t_l*ss;\r | |
260 | /* 2/(3log2)*(ss+...) */\r | |
261 | p_h = u+v;\r | |
262 | SET_LOW_WORD(p_h,0);\r | |
263 | p_l = v-(p_h-u);\r | |
264 | z_h = cp_h*p_h; /* cp_h+cp_l = 2/(3*log2) */\r | |
265 | z_l = cp_l*p_h+p_l*cp+dp_l[k];\r | |
266 | /* log2(ax) = (ss+..)*2/(3*log2) = n + dp_h + z_h + z_l */\r | |
267 | t = (double)n;\r | |
268 | t1 = (((z_h+z_l)+dp_h[k])+t);\r | |
269 | SET_LOW_WORD(t1,0);\r | |
270 | t2 = z_l-(((t1-t)-dp_h[k])-z_h);\r | |
271 | }\r | |
272 | \r | |
273 | /* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */\r | |
274 | y1 = y;\r | |
275 | SET_LOW_WORD(y1,0);\r | |
276 | p_l = (y-y1)*t1+y*t2;\r | |
277 | p_h = y1*t1;\r | |
278 | z = p_l+p_h;\r | |
279 | EXTRACT_WORDS(j,i,z);\r | |
280 | if (j>=0x40900000) { /* z >= 1024 */\r | |
281 | if(((j-0x40900000)|i)!=0) /* if z > 1024 */\r | |
282 | return s*huge*huge; /* overflow */\r | |
283 | else {\r | |
284 | if(p_l+ovt>z-p_h) return s*huge*huge; /* overflow */\r | |
285 | }\r | |
286 | } else if((j&0x7fffffff)>=0x4090cc00 ) { /* z <= -1075 */\r | |
287 | if(((j-0xc090cc00)|i)!=0) /* z < -1075 */\r | |
288 | return s*tiny*tiny; /* underflow */\r | |
289 | else {\r | |
290 | if(p_l<=z-p_h) return s*tiny*tiny; /* underflow */\r | |
291 | }\r | |
292 | }\r | |
293 | /*\r | |
294 | * compute 2**(p_h+p_l)\r | |
295 | */\r | |
296 | i = j&0x7fffffff;\r | |
297 | k = (i>>20)-0x3ff;\r | |
298 | n = 0;\r | |
299 | if(i>0x3fe00000) { /* if |z| > 0.5, set n = [z+0.5] */\r | |
300 | n = j+(0x00100000>>(k+1));\r | |
301 | k = ((n&0x7fffffff)>>20)-0x3ff; /* new k for n */\r | |
302 | t = zero;\r | |
303 | SET_HIGH_WORD(t,n&~(0x000fffff>>k));\r | |
304 | n = ((n&0x000fffff)|0x00100000)>>(20-k);\r | |
305 | if(j<0) n = -n;\r | |
306 | p_h -= t;\r | |
307 | }\r | |
308 | t = p_l+p_h;\r | |
309 | SET_LOW_WORD(t,0);\r | |
310 | u = t*lg2_h;\r | |
311 | v = (p_l-(t-p_h))*lg2+t*lg2_l;\r | |
312 | z = u+v;\r | |
313 | w = v-(z-u);\r | |
314 | t = z*z;\r | |
315 | t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5))));\r | |
316 | r = (z*t1)/(t1-two)-(w+z*w);\r | |
317 | z = one-(r-z);\r | |
318 | GET_HIGH_WORD(j,z);\r | |
319 | j += (n<<20);\r | |
320 | if((j>>20)<=0) z = scalbn(z,n); /* subnormal output */\r | |
321 | else SET_HIGH_WORD(z,j);\r | |
322 | return s*z;\r | |
323 | }\r |