Skip to content

Commit 23d3caf

Browse files
bassiounixbassiounix
authored
[libc][math] Refactor exp10m1f implementation to header-only in src/__support/math folder. (#159897)
Part of #147386 in preparation for: https://discourse.llvm.org/t/rfc-make-clang-builtin-math-functions-constexpr-with-llvm-libc-to-support-c-23-constexpr-math-functions/86450
1 parent abffc54 commit 23d3caf

File tree

9 files changed

+297
-218
lines changed

9 files changed

+297
-218
lines changed

‎libc/shared/math.h‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -45,6 +45,7 @@
4545
#include "math/exp10.h"
4646
#include "math/exp10f.h"
4747
#include "math/exp10f16.h"
48+
#include "math/exp10m1f.h"
4849
#include "math/expf.h"
4950
#include "math/expf16.h"
5051
#include "math/frexpf.h"

‎libc/shared/math/exp10m1f.h‎

Lines changed: 23 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,23 @@
1+
//===-- Shared exp10m1f function --------------------------------*- C++ -*-===//
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+
#ifndef LLVM_LIBC_SHARED_MATH_EXP10M1F_H
10+
#define LLVM_LIBC_SHARED_MATH_EXP10M1F_H
11+
12+
#include "shared/libc_common.h"
13+
#include "src/__support/math/exp10m1f.h"
14+
15+
namespace LIBC_NAMESPACE_DECL {
16+
namespace shared {
17+
18+
using math::exp10m1f;
19+
20+
} // namespace shared
21+
} // namespace LIBC_NAMESPACE_DECL
22+
23+
#endif // LLVM_LIBC_SHARED_MATH_EXP10M1F_H

‎libc/src/__support/math/CMakeLists.txt‎

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -481,6 +481,23 @@ add_header_library(
481481
libc.src.__support.FPUtil.generic.sqrt
482482
)
483483

484+
add_header_library(
485+
exp10m1f
486+
HDRS
487+
exp10m1f.h
488+
DEPENDS
489+
.exp10f_utils
490+
libc.src.errno.errno
491+
libc.src.__support.common
492+
libc.src.__support.FPUtil.except_value_utils
493+
libc.src.__support.FPUtil.fenv_impl
494+
libc.src.__support.FPUtil.fp_bits
495+
libc.src.__support.FPUtil.multiply_add
496+
libc.src.__support.FPUtil.polyeval
497+
libc.src.__support.FPUtil.rounding_mode
498+
libc.src.__support.macros.optimization
499+
)
500+
484501
add_header_library(
485502
erff
486503
HDRS

‎libc/src/__support/math/exp10m1f.h‎

Lines changed: 234 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,234 @@
1+
//===-- Implementation header for exp10m1f ----------------------*- C++ -*-===//
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+
#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_EXP10M1F_H
10+
#define LLVM_LIBC_SRC___SUPPORT_MATH_EXP10M1F_H
11+
12+
#include "exp10f_utils.h"
13+
#include "src/__support/FPUtil/FEnvImpl.h"
14+
#include "src/__support/FPUtil/FPBits.h"
15+
#include "src/__support/FPUtil/PolyEval.h"
16+
#include "src/__support/FPUtil/except_value_utils.h"
17+
#include "src/__support/FPUtil/multiply_add.h"
18+
#include "src/__support/FPUtil/rounding_mode.h"
19+
#include "src/__support/common.h"
20+
#include "src/__support/libc_errno.h"
21+
#include "src/__support/macros/config.h"
22+
#include "src/__support/macros/optimization.h"
23+
24+
namespace LIBC_NAMESPACE_DECL {
25+
26+
namespace math {
27+
28+
namespace exp10m1f_internal {
29+
30+
#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
31+
static constexpr size_t N_EXCEPTS_LO = 11;
32+
33+
static constexpr fputil::ExceptValues<float, N_EXCEPTS_LO> EXP10M1F_EXCEPTS_LO =
34+
{{
35+
// x = 0x1.0fe54ep-11, exp10m1f(x) = 0x1.3937eep-10 (RZ)
36+
{0x3a07'f2a7U, 0x3a9c'9bf7U, 1U, 0U, 1U},
37+
// x = 0x1.80e6eap-11, exp10m1f(x) = 0x1.bb8272p-10 (RZ)
38+
{0x3a40'7375U, 0x3add'c139U, 1U, 0U, 1U},
39+
// x = -0x1.2a33bcp-51, exp10m1f(x) = -0x1.57515ep-50 (RZ)
40+
{0xa615'19deU, 0xa6ab'a8afU, 0U, 1U, 0U},
41+
// x = -0x0p+0, exp10m1f(x) = -0x0p+0 (RZ)
42+
{0x8000'0000U, 0x8000'0000U, 0U, 0U, 0U},
43+
// x = -0x1.b59e08p-31, exp10m1f(x) = -0x1.f7d356p-30 (RZ)
44+
{0xb05a'cf04U, 0xb0fb'e9abU, 0U, 1U, 1U},
45+
// x = -0x1.bf342p-12, exp10m1f(x) = -0x1.014e02p-10 (RZ)
46+
{0xb9df'9a10U, 0xba80'a701U, 0U, 1U, 0U},
47+
// x = -0x1.6207fp-11, exp10m1f(x) = -0x1.9746cap-10 (RZ)
48+
{0xba31'03f8U, 0xbacb'a365U, 0U, 1U, 1U},
49+
// x = -0x1.bd0c66p-11, exp10m1f(x) = -0x1.ffe168p-10 (RZ)
50+
{0xba5e'8633U, 0xbaff'f0b4U, 0U, 1U, 1U},
51+
// x = -0x1.ffd84cp-10, exp10m1f(x) = -0x1.25faf2p-8 (RZ)
52+
{0xbaff'ec26U, 0xbb92'fd79U, 0U, 1U, 0U},
53+
// x = -0x1.a74172p-9, exp10m1f(x) = -0x1.e57be2p-8 (RZ)
54+
{0xbb53'a0b9U, 0xbbf2'bdf1U, 0U, 1U, 1U},
55+
// x = -0x1.cb694cp-9, exp10m1f(x) = -0x1.0764e4p-7 (RZ)
56+
{0xbb65'b4a6U, 0xbc03'b272U, 0U, 1U, 0U},
57+
}};
58+
59+
static constexpr size_t N_EXCEPTS_HI = 19;
60+
61+
static constexpr fputil::ExceptValues<float, N_EXCEPTS_HI> EXP10M1F_EXCEPTS_HI =
62+
{{
63+
// (input, RZ output, RU offset, RD offset, RN offset)
64+
// x = 0x1.8d31eep-8, exp10m1f(x) = 0x1.cc7e4cp-7 (RZ)
65+
{0x3bc6'98f7U, 0x3c66'3f26U, 1U, 0U, 1U},
66+
// x = 0x1.915fcep-8, exp10m1f(x) = 0x1.d15f72p-7 (RZ)
67+
{0x3bc8'afe7U, 0x3c68'afb9U, 1U, 0U, 0U},
68+
// x = 0x1.bcf982p-8, exp10m1f(x) = 0x1.022928p-6 (RZ)
69+
{0x3bde'7cc1U, 0x3c81'1494U, 1U, 0U, 1U},
70+
// x = 0x1.99ff0ap-7, exp10m1f(x) = 0x1.dee416p-6 (RZ)
71+
{0x3c4c'ff85U, 0x3cef'720bU, 1U, 0U, 0U},
72+
// x = 0x1.75ea14p-6, exp10m1f(x) = 0x1.b9ff16p-5 (RZ)
73+
{0x3cba'f50aU, 0x3d5c'ff8bU, 1U, 0U, 0U},
74+
// x = 0x1.f81b64p-6, exp10m1f(x) = 0x1.2cb6bcp-4 (RZ)
75+
{0x3cfc'0db2U, 0x3d96'5b5eU, 1U, 0U, 0U},
76+
// x = 0x1.fafecp+3, exp10m1f(x) = 0x1.8c880ap+52 (RZ)
77+
{0x417d'7f60U, 0x59c6'4405U, 1U, 0U, 0U},
78+
// x = -0x1.3bf094p-8, exp10m1f(x) = -0x1.69ba4ap-7 (RZ)
79+
{0xbb9d'f84aU, 0xbc34'dd25U, 0U, 1U, 0U},
80+
// x = -0x1.4558bcp-8, exp10m1f(x) = -0x1.746fb8p-7 (RZ)
81+
{0xbba2'ac5eU, 0xbc3a'37dcU, 0U, 1U, 1U},
82+
// x = -0x1.4bb43p-8, exp10m1f(x) = -0x1.7babe4p-7 (RZ)
83+
{0xbba5'da18U, 0xbc3d'd5f2U, 0U, 1U, 1U},
84+
// x = -0x1.776cc8p-8, exp10m1f(x) = -0x1.ad62c4p-7 (RZ)
85+
{0xbbbb'b664U, 0xbc56'b162U, 0U, 1U, 0U},
86+
// x = -0x1.f024cp-8, exp10m1f(x) = -0x1.1b20d6p-6 (RZ)
87+
{0xbbf8'1260U, 0xbc8d'906bU, 0U, 1U, 1U},
88+
// x = -0x1.f510eep-8, exp10m1f(x) = -0x1.1de9aap-6 (RZ)
89+
{0xbbfa'8877U, 0xbc8e'f4d5U, 0U, 1U, 0U},
90+
// x = -0x1.0b43c4p-7, exp10m1f(x) = -0x1.30d418p-6 (RZ)
91+
{0xbc05'a1e2U, 0xbc98'6a0cU, 0U, 1U, 0U},
92+
// x = -0x1.245ee4p-7, exp10m1f(x) = -0x1.4d2b86p-6 (RZ)
93+
{0xbc12'2f72U, 0xbca6'95c3U, 0U, 1U, 0U},
94+
// x = -0x1.f9f2dap-7, exp10m1f(x) = -0x1.1e2186p-5 (RZ)
95+
{0xbc7c'f96dU, 0xbd0f'10c3U, 0U, 1U, 0U},
96+
// x = -0x1.08e42p-6, exp10m1f(x) = -0x1.2b5c4p-5 (RZ)
97+
{0xbc84'7210U, 0xbd15'ae20U, 0U, 1U, 1U},
98+
// x = -0x1.0cdc44p-5, exp10m1f(x) = -0x1.2a2152p-4 (RZ)
99+
{0xbd06'6e22U, 0xbd95'10a9U, 0U, 1U, 1U},
100+
// x = -0x1.ca4322p-5, exp10m1f(x) = -0x1.ef073p-4 (RZ)
101+
{0xbd65'2191U, 0xbdf7'8398U, 0U, 1U, 1U},
102+
}};
103+
#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
104+
105+
} // namespace exp10m1f_internal
106+
107+
LIBC_INLINE static constexpr float exp10m1f(float x) {
108+
using namespace exp10m1f_internal;
109+
using FPBits = fputil::FPBits<float>;
110+
FPBits xbits(x);
111+
112+
uint32_t x_u = xbits.uintval();
113+
uint32_t x_abs = x_u & 0x7fff'ffffU;
114+
115+
// When x >= log10(2^128), or x is nan
116+
if (LIBC_UNLIKELY(xbits.is_pos() && x_u >= 0x421a'209bU)) {
117+
if (xbits.is_finite()) {
118+
int rounding = fputil::quick_get_round();
119+
if (rounding == FE_DOWNWARD || rounding == FE_TOWARDZERO)
120+
return FPBits::max_normal().get_val();
121+
122+
fputil::set_errno_if_required(ERANGE);
123+
fputil::raise_except_if_required(FE_OVERFLOW);
124+
}
125+
126+
// x >= log10(2^128) and 10^x - 1 rounds to +inf, or x is +inf or nan
127+
return x + FPBits::inf().get_val();
128+
}
129+
130+
// When |x| <= log10(2) * 2^(-6)
131+
if (LIBC_UNLIKELY(x_abs <= 0x3b9a'209bU)) {
132+
#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
133+
if (auto r = EXP10M1F_EXCEPTS_LO.lookup(x_u); LIBC_UNLIKELY(r.has_value()))
134+
return r.value();
135+
#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
136+
137+
double dx = x;
138+
double dx_sq = dx * dx;
139+
double c0 = dx * Exp10Base::COEFFS[0];
140+
double c1 =
141+
fputil::multiply_add(dx, Exp10Base::COEFFS[2], Exp10Base::COEFFS[1]);
142+
double c2 =
143+
fputil::multiply_add(dx, Exp10Base::COEFFS[4], Exp10Base::COEFFS[3]);
144+
// 10^dx - 1 ~ (1 + COEFFS[0] * dx + ... + COEFFS[4] * dx^5) - 1
145+
// = COEFFS[0] * dx + ... + COEFFS[4] * dx^5
146+
return static_cast<float>(fputil::polyeval(dx_sq, c0, c1, c2));
147+
}
148+
149+
// When x <= log10(2^-25), or x is nan
150+
if (LIBC_UNLIKELY(x_u >= 0xc0f0d2f1)) {
151+
// exp10m1(-inf) = -1
152+
if (xbits.is_inf())
153+
return -1.0f;
154+
// exp10m1(nan) = nan
155+
if (xbits.is_nan())
156+
return x;
157+
158+
int rounding = fputil::quick_get_round();
159+
if (rounding == FE_UPWARD || rounding == FE_TOWARDZERO ||
160+
(rounding == FE_TONEAREST && x_u == 0xc0f0d2f1))
161+
return -0x1.ffff'fep-1f; // -1.0f + 0x1.0p-24f
162+
163+
fputil::set_errno_if_required(ERANGE);
164+
fputil::raise_except_if_required(FE_UNDERFLOW);
165+
return -1.0f;
166+
}
167+
168+
// Exact outputs when x = 1, 2, ..., 10.
169+
// Quick check mask: 0x800f'ffffU = ~(bits of 1.0f | ... | bits of 10.0f)
170+
if (LIBC_UNLIKELY((x_u & 0x800f'ffffU) == 0)) {
171+
switch (x_u) {
172+
case 0x3f800000U: // x = 1.0f
173+
return 9.0f;
174+
case 0x40000000U: // x = 2.0f
175+
return 99.0f;
176+
case 0x40400000U: // x = 3.0f
177+
return 999.0f;
178+
case 0x40800000U: // x = 4.0f
179+
return 9'999.0f;
180+
case 0x40a00000U: // x = 5.0f
181+
return 99'999.0f;
182+
case 0x40c00000U: // x = 6.0f
183+
return 999'999.0f;
184+
case 0x40e00000U: // x = 7.0f
185+
return 9'999'999.0f;
186+
case 0x41000000U: { // x = 8.0f
187+
int rounding = fputil::quick_get_round();
188+
if (rounding == FE_UPWARD || rounding == FE_TONEAREST)
189+
return 100'000'000.0f;
190+
return 99'999'992.0f;
191+
}
192+
case 0x41100000U: { // x = 9.0f
193+
int rounding = fputil::quick_get_round();
194+
if (rounding == FE_UPWARD || rounding == FE_TONEAREST)
195+
return 1'000'000'000.0f;
196+
return 999'999'936.0f;
197+
}
198+
case 0x41200000U: { // x = 10.0f
199+
int rounding = fputil::quick_get_round();
200+
if (rounding == FE_UPWARD || rounding == FE_TONEAREST)
201+
return 10'000'000'000.0f;
202+
return 9'999'998'976.0f;
203+
}
204+
}
205+
}
206+
207+
#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
208+
if (auto r = EXP10M1F_EXCEPTS_HI.lookup(x_u); LIBC_UNLIKELY(r.has_value()))
209+
return r.value();
210+
#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
211+
212+
// Range reduction: 10^x = 2^(mid + hi) * 10^lo
213+
// rr = (2^(mid + hi), lo)
214+
auto rr = exp_b_range_reduc<Exp10Base>(x);
215+
216+
// The low part is approximated by a degree-5 minimax polynomial.
217+
// 10^lo ~ 1 + COEFFS[0] * lo + ... + COEFFS[4] * lo^5
218+
double lo_sq = rr.lo * rr.lo;
219+
double c0 = fputil::multiply_add(rr.lo, Exp10Base::COEFFS[0], 1.0);
220+
double c1 =
221+
fputil::multiply_add(rr.lo, Exp10Base::COEFFS[2], Exp10Base::COEFFS[1]);
222+
double c2 =
223+
fputil::multiply_add(rr.lo, Exp10Base::COEFFS[4], Exp10Base::COEFFS[3]);
224+
double exp10_lo = fputil::polyeval(lo_sq, c0, c1, c2);
225+
// 10^x - 1 = 2^(mid + hi) * 10^lo - 1
226+
// ~ mh * exp10_lo - 1
227+
return static_cast<float>(fputil::multiply_add(exp10_lo, rr.mh, -1.0));
228+
}
229+
230+
} // namespace math
231+
232+
} // namespace LIBC_NAMESPACE_DECL
233+
234+
#endif // LLVM_LIBC_SRC___SUPPORT_MATH_EXP10M1F_H

‎libc/src/math/generic/CMakeLists.txt‎

Lines changed: 1 addition & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -1593,16 +1593,7 @@ add_entrypoint_object(
15931593
HDRS
15941594
../exp10m1f.h
15951595
DEPENDS
1596-
libc.src.errno.errno
1597-
libc.src.__support.common
1598-
libc.src.__support.FPUtil.except_value_utils
1599-
libc.src.__support.FPUtil.fenv_impl
1600-
libc.src.__support.FPUtil.fp_bits
1601-
libc.src.__support.FPUtil.multiply_add
1602-
libc.src.__support.FPUtil.polyeval
1603-
libc.src.__support.FPUtil.rounding_mode
1604-
libc.src.__support.macros.optimization
1605-
libc.src.__support.math.exp10f_utils
1596+
libc.src.__support.math.exp10m1f
16061597
)
16071598

16081599
add_entrypoint_object(

0 commit comments

Comments
 (0)