Skip to content

Commit f04329e

Browse files
dpgeorgedpgeorge
committed
lib/libm: Add acosh, asinh, atanh, tan; get working with stmhal.
acoshf, asinhf, atanhf were added from musl. mathsincos.c was split up into its original, separate files (from newlibe-nano-2). tan was added. All of the important missing float functions are now implemented, and pyboard now passes tests/float/math_fun.py (finally!).
1 parent 6936f46 commit f04329e

18 files changed

Lines changed: 1295 additions & 638 deletions

File tree

‎lib/libm/acoshf.c‎

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,32 @@
1+
/*****************************************************************************/
2+
/*****************************************************************************/
3+
// acoshf from musl-0.9.15
4+
/*****************************************************************************/
5+
/*****************************************************************************/
6+
7+
#include "libm.h"
8+
9+
#if FLT_EVAL_METHOD==2
10+
#undef sqrtf
11+
#define sqrtf sqrtl
12+
#elif FLT_EVAL_METHOD==1
13+
#undef sqrtf
14+
#define sqrtf sqrt
15+
#endif
16+
17+
/* acosh(x) = log(x + sqrt(x*x-1)) */
18+
float acoshf(float x)
19+
{
20+
union {float f; uint32_t i;} u = {x};
21+
uint32_t a = u.i & 0x7fffffff;
22+
23+
if (a < 0x3f800000+(1<<23))
24+
/* |x| < 2, invalid if x < 1 or nan */
25+
/* up to 2ulp error in [1,1.125] */
26+
return log1pf(x-1 + sqrtf((x-1)*(x-1)+2*(x-1)));
27+
if (a < 0x3f800000+(12<<23))
28+
/* |x| < 0x1p12 */
29+
return logf(2*x - 1/(x+sqrtf(x*x-1)));
30+
/* x >= 0x1p12 */
31+
return logf(x) + 0.693147180559945309417232121458176568f;
32+
}

‎lib/libm/asinhf.c‎

Lines changed: 34 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
/*****************************************************************************/
2+
/*****************************************************************************/
3+
// asinhf from musl-0.9.15
4+
/*****************************************************************************/
5+
/*****************************************************************************/
6+
7+
#include "libm.h"
8+
9+
/* asinh(x) = sign(x)*log(|x|+sqrt(x*x+1)) ~= x - x^3/6 + o(x^5) */
10+
float asinhf(float x)
11+
{
12+
union {float f; uint32_t i;} u = {.f = x};
13+
uint32_t i = u.i & 0x7fffffff;
14+
unsigned s = u.i >> 31;
15+
16+
/* |x| */
17+
u.i = i;
18+
x = u.f;
19+
20+
if (i >= 0x3f800000 + (12<<23)) {
21+
/* |x| >= 0x1p12 or inf or nan */
22+
x = logf(x) + 0.693147180559945309417232121458176568f;
23+
} else if (i >= 0x3f800000 + (1<<23)) {
24+
/* |x| >= 2 */
25+
x = logf(2*x + 1/(sqrtf(x*x+1)+x));
26+
} else if (i >= 0x3f800000 - (12<<23)) {
27+
/* |x| >= 0x1p-12, up to 1.6ulp error in [0.125,0.5] */
28+
x = log1pf(x + x*x/(sqrtf(x*x+1)+1));
29+
} else {
30+
/* |x| < 0x1p-12, raise inexact if x!=0 */
31+
FORCE_EVAL(x + 0x1p120f);
32+
}
33+
return s ? -x : x;
34+
}

‎lib/libm/atanhf.c‎

Lines changed: 34 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
/*****************************************************************************/
2+
/*****************************************************************************/
3+
// atanhf from musl-0.9.15
4+
/*****************************************************************************/
5+
/*****************************************************************************/
6+
7+
#include "libm.h"
8+
9+
/* atanh(x) = log((1+x)/(1-x))/2 = log1p(2x/(1-x))/2 ~= x + x^3/3 + o(x^5) */
10+
float atanhf(float x)
11+
{
12+
union {float f; uint32_t i;} u = {.f = x};
13+
unsigned s = u.i >> 31;
14+
float_t y;
15+
16+
/* |x| */
17+
u.i &= 0x7fffffff;
18+
y = u.f;
19+
20+
if (u.i < 0x3f800000 - (1<<23)) {
21+
if (u.i < 0x3f800000 - (32<<23)) {
22+
/* handle underflow */
23+
if (u.i < (1<<23))
24+
FORCE_EVAL((float)(y*y));
25+
} else {
26+
/* |x| < 0.5, up to 1.7ulp error */
27+
y = 0.5f*log1pf(2*y + 2*y*y/(1-y));
28+
}
29+
} else {
30+
/* avoid overflow */
31+
y = 0.5f*log1pf(2*(y/(1-y)));
32+
}
33+
return s ? -y : y;
34+
}

‎lib/libm/ef_rem_pio2.c‎

Lines changed: 202 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,202 @@
1+
/*
2+
* This file is part of the Micro Python project, http://micropython.org/
3+
*
4+
* These math functions are taken from newlib-nano-2, the newlib/libm/math
5+
* directory, available from https://github.com/32bitmicro/newlib-nano-2.
6+
*
7+
* Appropriate copyright headers are reproduced below.
8+
*/
9+
10+
/* ef_rem_pio2.c -- float version of e_rem_pio2.c
11+
* Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
12+
*/
13+
14+
/*
15+
* ====================================================
16+
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
17+
*
18+
* Developed at SunPro, a Sun Microsystems, Inc. business.
19+
* Permission to use, copy, modify, and distribute this
20+
* software is freely granted, provided that this notice
21+
* is preserved.
22+
* ====================================================
23+
*
24+
*/
25+
26+
/* __ieee754_rem_pio2f(x,y)
27+
*
28+
* return the remainder of x rem pi/2 in y[0]+y[1]
29+
* use __kernel_rem_pio2f()
30+
*/
31+
32+
#include "fdlibm.h"
33+
34+
/*
35+
* Table of constants for 2/pi, 396 Hex digits (476 decimal) of 2/pi
36+
*/
37+
#ifdef __STDC__
38+
static const __int32_t two_over_pi[] = {
39+
#else
40+
static __int32_t two_over_pi[] = {
41+
#endif
42+
0xA2, 0xF9, 0x83, 0x6E, 0x4E, 0x44, 0x15, 0x29, 0xFC,
43+
0x27, 0x57, 0xD1, 0xF5, 0x34, 0xDD, 0xC0, 0xDB, 0x62,
44+
0x95, 0x99, 0x3C, 0x43, 0x90, 0x41, 0xFE, 0x51, 0x63,
45+
0xAB, 0xDE, 0xBB, 0xC5, 0x61, 0xB7, 0x24, 0x6E, 0x3A,
46+
0x42, 0x4D, 0xD2, 0xE0, 0x06, 0x49, 0x2E, 0xEA, 0x09,
47+
0xD1, 0x92, 0x1C, 0xFE, 0x1D, 0xEB, 0x1C, 0xB1, 0x29,
48+
0xA7, 0x3E, 0xE8, 0x82, 0x35, 0xF5, 0x2E, 0xBB, 0x44,
49+
0x84, 0xE9, 0x9C, 0x70, 0x26, 0xB4, 0x5F, 0x7E, 0x41,
50+
0x39, 0x91, 0xD6, 0x39, 0x83, 0x53, 0x39, 0xF4, 0x9C,
51+
0x84, 0x5F, 0x8B, 0xBD, 0xF9, 0x28, 0x3B, 0x1F, 0xF8,
52+
0x97, 0xFF, 0xDE, 0x05, 0x98, 0x0F, 0xEF, 0x2F, 0x11,
53+
0x8B, 0x5A, 0x0A, 0x6D, 0x1F, 0x6D, 0x36, 0x7E, 0xCF,
54+
0x27, 0xCB, 0x09, 0xB7, 0x4F, 0x46, 0x3F, 0x66, 0x9E,
55+
0x5F, 0xEA, 0x2D, 0x75, 0x27, 0xBA, 0xC7, 0xEB, 0xE5,
56+
0xF1, 0x7B, 0x3D, 0x07, 0x39, 0xF7, 0x8A, 0x52, 0x92,
57+
0xEA, 0x6B, 0xFB, 0x5F, 0xB1, 0x1F, 0x8D, 0x5D, 0x08,
58+
0x56, 0x03, 0x30, 0x46, 0xFC, 0x7B, 0x6B, 0xAB, 0xF0,
59+
0xCF, 0xBC, 0x20, 0x9A, 0xF4, 0x36, 0x1D, 0xA9, 0xE3,
60+
0x91, 0x61, 0x5E, 0xE6, 0x1B, 0x08, 0x65, 0x99, 0x85,
61+
0x5F, 0x14, 0xA0, 0x68, 0x40, 0x8D, 0xFF, 0xD8, 0x80,
62+
0x4D, 0x73, 0x27, 0x31, 0x06, 0x06, 0x15, 0x56, 0xCA,
63+
0x73, 0xA8, 0xC9, 0x60, 0xE2, 0x7B, 0xC0, 0x8C, 0x6B,
64+
};
65+
66+
/* This array is like the one in e_rem_pio2.c, but the numbers are
67+
single precision and the last 8 bits are forced to 0. */
68+
#ifdef __STDC__
69+
static const __int32_t npio2_hw[] = {
70+
#else
71+
static __int32_t npio2_hw[] = {
72+
#endif
73+
0x3fc90f00, 0x40490f00, 0x4096cb00, 0x40c90f00, 0x40fb5300, 0x4116cb00,
74+
0x412fed00, 0x41490f00, 0x41623100, 0x417b5300, 0x418a3a00, 0x4196cb00,
75+
0x41a35c00, 0x41afed00, 0x41bc7e00, 0x41c90f00, 0x41d5a000, 0x41e23100,
76+
0x41eec200, 0x41fb5300, 0x4203f200, 0x420a3a00, 0x42108300, 0x4216cb00,
77+
0x421d1400, 0x42235c00, 0x4229a500, 0x422fed00, 0x42363600, 0x423c7e00,
78+
0x4242c700, 0x42490f00
79+
};
80+
81+
/*
82+
* invpio2: 24 bits of 2/pi
83+
* pio2_1: first 17 bit of pi/2
84+
* pio2_1t: pi/2 - pio2_1
85+
* pio2_2: second 17 bit of pi/2
86+
* pio2_2t: pi/2 - (pio2_1+pio2_2)
87+
* pio2_3: third 17 bit of pi/2
88+
* pio2_3t: pi/2 - (pio2_1+pio2_2+pio2_3)
89+
*/
90+
91+
#ifdef __STDC__
92+
static const float
93+
#else
94+
static float
95+
#endif
96+
zero = 0.0000000000e+00, /* 0x00000000 */
97+
half = 5.0000000000e-01, /* 0x3f000000 */
98+
two8 = 2.5600000000e+02, /* 0x43800000 */
99+
invpio2 = 6.3661980629e-01, /* 0x3f22f984 */
100+
pio2_1 = 1.5707855225e+00, /* 0x3fc90f80 */
101+
pio2_1t = 1.0804334124e-05, /* 0x37354443 */
102+
pio2_2 = 1.0804273188e-05, /* 0x37354400 */
103+
pio2_2t = 6.0770999344e-11, /* 0x2e85a308 */
104+
pio2_3 = 6.0770943833e-11, /* 0x2e85a300 */
105+
pio2_3t = 6.1232342629e-17; /* 0x248d3132 */
106+
107+
#ifdef __STDC__
108+
__int32_t __ieee754_rem_pio2f(float x, float *y)
109+
#else
110+
__int32_t __ieee754_rem_pio2f(x,y)
111+
float x,y[];
112+
#endif
113+
{
114+
float z,w,t,r,fn;
115+
float tx[3];
116+
__int32_t i,j,n,ix,hx;
117+
int e0,nx;
118+
119+
GET_FLOAT_WORD(hx,x);
120+
ix = hx&0x7fffffff;
121+
if(ix<=0x3f490fd8) /* |x| ~<= pi/4 , no need for reduction */
122+
{y[0] = x; y[1] = 0; return 0;}
123+
if(ix<0x4016cbe4) { /* |x| < 3pi/4, special case with n=+-1 */
124+
if(hx>0) {
125+
z = x - pio2_1;
126+
if((ix&0xfffffff0)!=0x3fc90fd0) { /* 24+24 bit pi OK */
127+
y[0] = z - pio2_1t;
128+
y[1] = (z-y[0])-pio2_1t;
129+
} else { /* near pi/2, use 24+24+24 bit pi */
130+
z -= pio2_2;
131+
y[0] = z - pio2_2t;
132+
y[1] = (z-y[0])-pio2_2t;
133+
}
134+
return 1;
135+
} else { /* negative x */
136+
z = x + pio2_1;
137+
if((ix&0xfffffff0)!=0x3fc90fd0) { /* 24+24 bit pi OK */
138+
y[0] = z + pio2_1t;
139+
y[1] = (z-y[0])+pio2_1t;
140+
} else { /* near pi/2, use 24+24+24 bit pi */
141+
z += pio2_2;
142+
y[0] = z + pio2_2t;
143+
y[1] = (z-y[0])+pio2_2t;
144+
}
145+
return -1;
146+
}
147+
}
148+
if(ix<=0x43490f80) { /* |x| ~<= 2^7*(pi/2), medium size */
149+
t = fabsf(x);
150+
n = (__int32_t) (t*invpio2+half);
151+
fn = (float)n;
152+
r = t-fn*pio2_1;
153+
w = fn*pio2_1t; /* 1st round good to 40 bit */
154+
if(n<32&&(ix&0xffffff00)!=npio2_hw[n-1]) {
155+
y[0] = r-w; /* quick check no cancellation */
156+
} else {
157+
__uint32_t high;
158+
j = ix>>23;
159+
y[0] = r-w;
160+
GET_FLOAT_WORD(high,y[0]);
161+
i = j-((high>>23)&0xff);
162+
if(i>8) { /* 2nd iteration needed, good to 57 */
163+
t = r;
164+
w = fn*pio2_2;
165+
r = t-w;
166+
w = fn*pio2_2t-((t-r)-w);
167+
y[0] = r-w;
168+
GET_FLOAT_WORD(high,y[0]);
169+
i = j-((high>>23)&0xff);
170+
if(i>25) { /* 3rd iteration need, 74 bits acc */
171+
t = r; /* will cover all possible cases */
172+
w = fn*pio2_3;
173+
r = t-w;
174+
w = fn*pio2_3t-((t-r)-w);
175+
y[0] = r-w;
176+
}
177+
}
178+
}
179+
y[1] = (r-y[0])-w;
180+
if(hx<0) {y[0] = -y[0]; y[1] = -y[1]; return -n;}
181+
else return n;
182+
}
183+
/*
184+
* all other (large) arguments
185+
*/
186+
if(!FLT_UWORD_IS_FINITE(ix)) {
187+
y[0]=y[1]=x-x; return 0;
188+
}
189+
/* set z = scalbn(|x|,ilogb(x)-7) */
190+
e0 = (int)((ix>>23)-134); /* e0 = ilogb(z)-7; */
191+
SET_FLOAT_WORD(z, ix - ((__int32_t)e0<<23));
192+
for(i=0;i<2;i++) {
193+
tx[i] = (float)((__int32_t)(z));
194+
z = (z-tx[i])*two8;
195+
}
196+
tx[2] = z;
197+
nx = 3;
198+
while(tx[nx-1]==zero) nx--; /* skip zero term */
199+
n = __kernel_rem_pio2f(tx,y,e0,nx,2,two_over_pi);
200+
if(hx<0) {y[0] = -y[0]; y[1] = -y[1]; return -n;}
201+
return n;
202+
}

0 commit comments

Comments
 (0)