1 //===-- Single-precision e^x - 1 function ---------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "src/math/expm1f.h"
10 #include "common_constants.h" // Lookup tables EXP_M1 and EXP_M2.
11 #include "src/__support/FPUtil/BasicOperations.h"
12 #include "src/__support/FPUtil/FEnvImpl.h"
13 #include "src/__support/FPUtil/FMA.h"
14 #include "src/__support/FPUtil/FPBits.h"
15 #include "src/__support/FPUtil/PolyEval.h"
16 #include "src/__support/common.h"
17 
18 #include <errno.h>
19 
20 namespace __llvm_libc {
21 
22 INLINE_FMA
23 LLVM_LIBC_FUNCTION(float, expm1f, (float x)) {
24   using FPBits = typename fputil::FPBits<float>;
25   FPBits xbits(x);
26 
27   // When x < log(2^-25) or nan
28   if (unlikely(xbits.uintval() >= 0xc18a'a123U)) {
29     // exp(-Inf) = 0
30     if (xbits.is_inf())
31       return -1.0f;
32     // exp(nan) = nan
33     if (xbits.is_nan())
34       return x;
35     int round_mode = fputil::get_round();
36     if (round_mode == FE_UPWARD || round_mode == FE_TOWARDZERO)
37       return -0x1.ffff'fep-1f; // -1.0f + 0x1.0p-24f
38     return -1.0f;
39   }
40   // x >= 89 or nan
41   if (unlikely(!xbits.get_sign() && (xbits.uintval() >= 0x42b2'0000))) {
42     if (xbits.uintval() < 0x7f80'0000U) {
43       int rounding = fputil::get_round();
44       if (rounding == FE_DOWNWARD || rounding == FE_TOWARDZERO)
45         return static_cast<float>(FPBits(FPBits::MAX_NORMAL));
46 
47       errno = ERANGE;
48     }
49     return x + static_cast<float>(FPBits::inf());
50   }
51 
52   int unbiased_exponent = static_cast<int>(xbits.get_unbiased_exponent());
53   // |x| < 2^-4
54   if (unbiased_exponent < 123) {
55     // |x| < 2^-25
56     if (unbiased_exponent < 102) {
57       // x = -0.0f
58       if (unlikely(xbits.uintval() == 0x8000'0000U))
59         return x;
60       // When |x| < 2^-25, the relative error:
61       //   |(e^x - 1) - x| / |x| < |x^2| / |x| = |x| < 2^-25 < epsilon(1)/2.
62       // So the correctly rounded values of expm1(x) are:
63       //   = x + eps(x) if rounding mode = FE_UPWARD,
64       //                   or (rounding mode = FE_TOWARDZERO and x is negative),
65       //   = x otherwise.
66       // To simplify the rounding decision and make it more efficient, we use
67       //   fma(x, x, x) ~ x + x^2 instead.
68       return fputil::fma(x, x, x);
69     }
70     // 2^-25 <= |x| < 2^-4
71     double xd = static_cast<double>(x);
72     double xsq = xd * xd;
73     // Degree-8 minimax polynomial generated by Sollya with:
74     // > display = hexadecimal;
75     // > P = fpminimax(expm1(x)/x, 7, [|D...|], [-2^-4, 2^-4]);
76     double r =
77         fputil::polyeval(xd, 0x1p-1, 0x1.55555555559abp-3, 0x1.55555555551a7p-5,
78                          0x1.111110f70f2a4p-7, 0x1.6c16c17639e82p-10,
79                          0x1.a02526febbea6p-13, 0x1.a01dc40888fcdp-16);
80     return static_cast<float>(fputil::fma(r, xsq, xd));
81   }
82 
83   // For -18 < x < 89, to compute exp(x), we perform the following range
84   // reduction: find hi, mid, lo such that:
85   //   x = hi + mid + lo, in which
86   //     hi is an integer,
87   //     mid * 2^7 is an integer
88   //     -2^(-8) <= lo < 2^-8.
89   // In particular,
90   //   hi + mid = round(x * 2^7) * 2^(-7).
91   // Then,
92   //   exp(x) = exp(hi + mid + lo) = exp(hi) * exp(mid) * exp(lo).
93   // We store exp(hi) and exp(mid) in the lookup tables EXP_M1 and EXP_M2
94   // respectively.  exp(lo) is computed using a degree-7 minimax polynomial
95   // generated by Sollya.
96 
97   // Exceptional value
98   if (xbits.uintval() == 0xbdc1'c6cbU) {
99     // x = -0x1.838d96p-4f
100     int round_mode = fputil::get_round();
101     if (round_mode == FE_TONEAREST || round_mode == FE_DOWNWARD)
102       return -0x1.71c884p-4f;
103     return -0x1.71c882p-4f;
104   }
105 
106   // x_hi = hi + mid.
107   int x_hi = static_cast<int>(x * 0x1.0p7f);
108   // Subtract (hi + mid) from x to get lo.
109   x -= static_cast<float>(x_hi) * 0x1.0p-7f;
110   double xd = static_cast<double>(x);
111   // Make sure that -2^(-8) <= lo < 2^-8.
112   if (x >= 0x1.0p-8f) {
113     ++x_hi;
114     xd -= 0x1.0p-7;
115   }
116   if (x < -0x1.0p-8f) {
117     --x_hi;
118     xd += 0x1.0p-7;
119   }
120   x_hi += 104 << 7;
121   // hi = x_hi >> 7
122   double exp_hi = EXP_M1[x_hi >> 7];
123   // lo = x_hi & 0x0000'007fU;
124   double exp_mid = EXP_M2[x_hi & 0x7f];
125   double exp_hi_mid = exp_hi * exp_mid;
126   // Degree-7 minimax polynomial generated by Sollya with the following
127   // commands:
128   //   > display = hexadecimal;
129   //   > Q = fpminimax(expm1(x)/x, 6, [|D...|], [-2^-8, 2^-8]);
130   //   > Q;
131   double exp_lo = fputil::polyeval(
132       xd, 0x1p0, 0x1p0, 0x1p-1, 0x1.5555555555555p-3, 0x1.55555555553ap-5,
133       0x1.1111111204dfcp-7, 0x1.6c16cb2da593ap-10, 0x1.9ff1648996d2ep-13);
134   return static_cast<float>(fputil::fma(exp_hi_mid, exp_lo, -1.0));
135 }
136 
137 } // namespace __llvm_libc
138