math: Use acospif from CORE-MATH

The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows better performance to the generic acospif.

The code was adapted to glibc style and to use the definition of
math_config.h (to handle errno, overflow, and underflow).

Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (Neoverse-N1,
gcc 13.3.1), and powerpc (POWER10, gcc 13.2.1):

latency                  master        patched   improvement
x86_64                  54.8281        42.9070        21.74%
x86_64v2                54.1717        42.7497        21.08%
x86_64v3                49.3552        34.1512        30.81%
aarch64 (Neoverse)      17.9395        14.3733        19.88%
power8                  20.3110         8.8609        56.37%
power10                 11.3113        8.84067        21.84%

reciprocal-throughput    master        patched   improvement
x86_64                  21.2301        14.4803        31.79%
x86_64v2                20.6858        13.9506        32.56%
x86_64v3                16.1944        11.3377        29.99%
aarch64 (Neoverse)      11.4474        7.13282        37.69%
power8                  10.6916        3.57547        66.56%
power10                 4.64269        3.54145        23.72%

Reviewed-by: DJ Delorie <dj@redhat.com>

18 files changed, 137 insertions, 59 deletions

diff --git a/sysdeps/aarch64/libm-test-ulps b/sysdeps/aarch64/libm-test-ulps
index 59ec929..1a403d9 100644
--- a/sysdeps/aarch64/libm-test-ulps
+++ b/sysdeps/aarch64/libm-test-ulps
@@ -51,22 +51,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/arc/fpu/libm-test-ulps b/sysdeps/arc/fpu/libm-test-ulps
index 82bc156..c0c5daa 100644
--- a/sysdeps/arc/fpu/libm-test-ulps
+++ b/sysdeps/arc/fpu/libm-test-ulps
@@ -27,19 +27,15 @@ double: 3
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 
 Function: "acospi_upward":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/arc/nofpu/libm-test-ulps b/sysdeps/arc/nofpu/libm-test-ulps
index aa93d71..2b34f5a 100644
--- a/sysdeps/arc/nofpu/libm-test-ulps
+++ b/sysdeps/arc/nofpu/libm-test-ulps
@@ -9,7 +9,6 @@ double: 2
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/arm/libm-test-ulps b/sysdeps/arm/libm-test-ulps
index 218ffa8..afb0532 100644
--- a/sysdeps/arm/libm-test-ulps
+++ b/sysdeps/arm/libm-test-ulps
@@ -27,19 +27,15 @@ double: 2
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 
 Function: "acospi_upward":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/hppa/fpu/libm-test-ulps b/sysdeps/hppa/fpu/libm-test-ulps
index 2b8fa35..b9959c8 100644
--- a/sysdeps/hppa/fpu/libm-test-ulps
+++ b/sysdeps/hppa/fpu/libm-test-ulps
@@ -27,19 +27,15 @@ double: 2
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 
 Function: "acospi_upward":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index b898344..85c58f3 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -41,25 +41,21 @@ ldouble: 3
 
 Function: "acospi":
 double: 1
-float: 1
 float128: 2
 ldouble: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_upward":
 double: 2
-float: 1
 float128: 2
 ldouble: 2
 
diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
index 750d519..bc14e7e 100644
--- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
+++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps
@@ -41,25 +41,21 @@ ldouble: 3
 
 Function: "acospi":
 double: 1
-float: 1
 float128: 2
 ldouble: 3
 
 Function: "acospi_downward":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_upward":
 double: 2
-float: 1
 float128: 2
 ldouble: 2
 
diff --git a/sysdeps/ieee754/flt-32/s_acospif.c b/sysdeps/ieee754/flt-32/s_acospif.c
new file mode 100644
index 0000000..03d63a7
--- /dev/null
+++ b/sysdeps/ieee754/flt-32/s_acospif.c
@@ -0,0 +1,137 @@
+/* Correctly-rounded half-revolution arc-cosine function for binary32 value.
+
+Copyright (c) 2022-2025 Alexei Sibidanov.
+
+The original version of this file was copied from the CORE-MATH
+project (file src/binary32/acospi/acospif.c, revision 1a6a9ab).
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+*/
+
+#include <math.h>
+#include <stdint.h>
+#include <libm-alias-float.h>
+#include "math_config.h"
+
+float
+__acospif (float x)
+{
+  float ax = fabsf (x);
+  double az = ax;
+  double z = x;
+  uint32_t t = asuint (x);
+  int e = (t >> 23) & 0xff;
+  if (__glibc_unlikely (e >= 127))
+    {
+      if (x == 1.0f)
+	return 0.0f;
+      if (x == -1.0f)
+	return 1.0f;
+      if (e == 0xff && (t << 9))
+	return x + x; /* nan */
+      return __math_edomf ((x - x) / (x - x)); /* nan */
+    }
+  int s = 146 - e;
+  int i = 0;
+  if (__glibc_likely (s < 32))
+    i = ((t & (~0u >> 9)) | 1 << 23) >> s;
+  static const double ch[][8] = {
+    { 0x1.45f306dc9c882p-2, 0x1.b2995e7b7dc2fp-5, 0x1.8723a1cf50c7ep-6,
+      0x1.d1a4591d16a29p-7, 0x1.3ce3aa68ddaeep-7, 0x1.d3182ab0cc1bfp-8,
+      0x1.62b379a8b88e3p-8, 0x1.6811411fcfec2p-8 },
+    { 0x1.ffffffffd3cd9p-2, -0x1.17cc1b3355fd5p-4, 0x1.d067a1e8d5a99p-6,
+      -0x1.08e16fb09314ap-6, 0x1.5eed43d42dcb2p-7, -0x1.f58baca7acc71p-8,
+      0x1.5dab64e2dcf15p-8, -0x1.59270e30797acp-9 },
+    { 0x1.fffffff7c4622p-2, -0x1.17cc149ded3a2p-4, 0x1.d0654d4cb2c1ap-6,
+      -0x1.08c3ba713d33ap-6, 0x1.5d2053481079cp-7, -0x1.e485ebc545e7ep-8,
+      0x1.303baca167dddp-8, -0x1.dee8d16d06b38p-10 },
+    { 0x1.ffffffa749848p-2, -0x1.17cbe7155935p-4, 0x1.d05a312269adfp-6,
+      -0x1.0862b3ee617d7p-6, 0x1.5920708db2a73p-7, -0x1.cb0463b3862c3p-8,
+      0x1.02b82478f95d7p-8, -0x1.52a7b8579e729p-10 },
+    { 0x1.fffffe1f92bb5p-2, -0x1.17cb3e74c64e3p-4, 0x1.d03af67311cbfp-6,
+      -0x1.079441cbfc7ap-6, 0x1.52b4287805a61p-7, -0x1.ac3286d604a98p-8,
+      0x1.b2f1210d9701bp-9, -0x1.e740ddc25afd6p-11 },
+    { 0x1.fffff92beb6e2p-2, -0x1.17c986fe9518bp-4, 0x1.cff98167c9a5ep-6,
+      -0x1.0638b591eae52p-6, 0x1.4a0803828959ep-7, -0x1.8adeca229f11dp-8,
+      0x1.6b9a7ba05dfcep-9, -0x1.640521a43b2dp-11 },
+    { 0x1.ffffeccee5bfcp-2, -0x1.17c5f1753f5eap-4, 0x1.cf874e4fe258fp-6,
+      -0x1.043e6cf77b256p-6, 0x1.3f7db42227d92p-7, -0x1.691a6fa2a2882p-8,
+      0x1.2f6543162bc61p-9, -0x1.07d5da05822b6p-11 },
+    { 0x1.ffffd2f64431dp-2, -0x1.17bf8208c10c1p-4, 0x1.ced7487cdb124p-6,
+      -0x1.01a0d30932905p-6, 0x1.3388f99b254dap-7, -0x1.4844e245c65bdp-8,
+      0x1.fa777150197c6p-10, -0x1.8c1ecf16a05c8p-12 },
+    { 0x1.ffffa36d1712ep-2, -0x1.17b523971bd4ep-4, 0x1.cddee26de2deep-6,
+      -0x1.fccb00abaaabcp-7, 0x1.269afc3622342p-7, -0x1.2933152686752p-8,
+      0x1.a76d4956cc9a3p-10, -0x1.2ce7d6dc651cep-12 },
+    { 0x1.ffff5402ab3a1p-2, -0x1.17a5ba85da77ap-4, 0x1.cc96894e05c02p-6,
+      -0x1.f532143cb832ep-7, 0x1.19180b660ff09p-7, -0x1.0c57417a78b3cp-8,
+      0x1.62e26cbd7bb1ep-10, -0x1.ce28d33fe1df3p-13 },
+    { 0x1.fffed8d639751p-2, -0x1.1790349f3ae76p-4, 0x1.caf9a4fd1b398p-6,
+      -0x1.ec986b111342ep-7, 0x1.0b53c3ad4baa4p-7, -0x1.e3c2282eeace4p-9,
+      0x1.2a55369f55bbep-10, -0x1.667fe48c396e8p-13 },
+    { 0x1.fffe24b714161p-2, -0x1.177394fbcb719p-4, 0x1.c90652d920ebdp-6,
+      -0x1.e3239197bddf1p-7, 0x1.fb2188525b025p-8, -0x1.b3aadd451afc7p-9,
+      0x1.f74020f31fdabp-11, -0x1.18b0cb246768dp-13 },
+    { 0x1.fffd298bec9e2p-2, -0x1.174efbfd34648p-4, 0x1.c6bcfe48ea92bp-6,
+      -0x1.d8f9f2a16157cp-7, 0x1.e0044f56c8864p-8, -0x1.883e2347fe76cp-9,
+      0x1.a9f0e3c1b7af5p-11, -0x1.bb5acc0e60825p-14 },
+    { 0x1.fffbd8b784c4dp-2, -0x1.1721abdd3722ep-4, 0x1.c41fee756d4bp-6,
+      -0x1.ce40bccf8065fp-7, 0x1.c59b684b70ef9p-8, -0x1.6133d027996b3p-9,
+      0x1.69cad01106397p-11, -0x1.60f8e45494156p-14 },
+    { 0x1.fffa23749cf88p-2, -0x1.16eb0a8285c06p-4, 0x1.c132d762e1b0dp-6,
+      -0x1.c31a959398f4ep-7, 0x1.ac1c5b46bc8ap-8, -0x1.3e34f1abe51dcp-9,
+      0x1.346738737c0b9p-11, -0x1.1b227a3f5c75p-14 },
+    { 0x1.fff7fb25bb407p-2, -0x1.16aaa14d7564p-4, 0x1.bdfa75fca5ff2p-6,
+      -0x1.b7a6e260d079cp-7, 0x1.93ab06911033cp-8, -0x1.1ee5560967fd5p-9,
+      0x1.07d31060838bfp-11, -0x1.c96f33a283115p-15 },
+  };
+  const double *c = ch[i];
+  double z2 = z * z;
+  double z4 = z2 * z2;
+  if (__glibc_unlikely (i == 0))
+    {
+      double c0 = c[0] + z2 * c[1];
+      double c2 = c[2] + z2 * c[3];
+      double c4 = c[4] + z2 * c[5];
+      double c6 = c[6] + z2 * c[7];
+      c0 += c2 * z4;
+      c4 += c6 * z4;
+      /* For |x| <= 0x1.0fd288p-127, c0 += c4*(z4*z4) would raise a spurious
+	 underflow exception, we use an FMA instead, where c4 * z4 does not
+	 underflow. */
+      c0 = fma (c4 * z4, z4, c0);
+      return 0.5 - z * c0;
+    }
+  else
+    {
+      double f = sqrt (1 - az);
+      double c0 = c[0] + az * c[1];
+      double c2 = c[2] + az * c[3];
+      double c4 = c[4] + az * c[5];
+      double c6 = c[6] + az * c[7];
+      c0 += c2 * z2;
+      c4 += c6 * z2;
+      c0 += c4 * z4;
+      static const double o[] = { 0, 1 };
+      double r = o[t >> 31] + c0 * copysign (f, x);
+      return r;
+    }
+}
+libm_alias_float (__acospi, acospi)
diff --git a/sysdeps/loongarch/lp64/libm-test-ulps b/sysdeps/loongarch/lp64/libm-test-ulps
index f8bf089..ce84ddf 100644
--- a/sysdeps/loongarch/lp64/libm-test-ulps
+++ b/sysdeps/loongarch/lp64/libm-test-ulps
@@ -35,22 +35,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/mips/mips64/libm-test-ulps b/sysdeps/mips/mips64/libm-test-ulps
index 98079e0..67c37df 100644
--- a/sysdeps/mips/mips64/libm-test-ulps
+++ b/sysdeps/mips/mips64/libm-test-ulps
@@ -35,22 +35,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/or1k/fpu/libm-test-ulps b/sysdeps/or1k/fpu/libm-test-ulps
index b0de024..d3b1036 100644
--- a/sysdeps/or1k/fpu/libm-test-ulps
+++ b/sysdeps/or1k/fpu/libm-test-ulps
@@ -27,19 +27,15 @@ double: 2
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 
 Function: "acospi_upward":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/or1k/nofpu/libm-test-ulps b/sysdeps/or1k/nofpu/libm-test-ulps
index aa047f3..14b7e0f 100644
--- a/sysdeps/or1k/nofpu/libm-test-ulps
+++ b/sysdeps/or1k/nofpu/libm-test-ulps
@@ -27,7 +27,6 @@ double: 2
 
 Function: "acospi":
 double: 2
-float: 1
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/powerpc/fpu/libm-test-ulps b/sysdeps/powerpc/fpu/libm-test-ulps
index cf3dec3..c9c86de 100644
--- a/sysdeps/powerpc/fpu/libm-test-ulps
+++ b/sysdeps/powerpc/fpu/libm-test-ulps
@@ -43,25 +43,21 @@ ldouble: 4
 
 Function: "acospi":
 double: 2
-float: 1
 float128: 1
 ldouble: 1
 
 Function: "acospi_downward":
 double: 1
-float: 2
 float128: 1
 ldouble: 4
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 float128: 1
 ldouble: 4
 
 Function: "acospi_upward":
 double: 2
-float: 1
 float128: 2
 ldouble: 4
 
diff --git a/sysdeps/riscv/nofpu/libm-test-ulps b/sysdeps/riscv/nofpu/libm-test-ulps
index d971ee2..6206a95 100644
--- a/sysdeps/riscv/nofpu/libm-test-ulps
+++ b/sysdeps/riscv/nofpu/libm-test-ulps
@@ -35,7 +35,6 @@ ldouble: 2
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/riscv/rvd/libm-test-ulps b/sysdeps/riscv/rvd/libm-test-ulps
index 0f84906..124ca4b 100644
--- a/sysdeps/riscv/rvd/libm-test-ulps
+++ b/sysdeps/riscv/rvd/libm-test-ulps
@@ -35,22 +35,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/s390/fpu/libm-test-ulps b/sysdeps/s390/fpu/libm-test-ulps
index 76a1f3c..364ccf3 100644
--- a/sysdeps/s390/fpu/libm-test-ulps
+++ b/sysdeps/s390/fpu/libm-test-ulps
@@ -35,22 +35,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/sparc/fpu/libm-test-ulps b/sysdeps/sparc/fpu/libm-test-ulps
index 02a80c4..1174972 100644
--- a/sysdeps/sparc/fpu/libm-test-ulps
+++ b/sysdeps/sparc/fpu/libm-test-ulps
@@ -35,22 +35,18 @@ ldouble: 3
 
 Function: "acospi":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "acospi_downward":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 ldouble: 1
 
 Function: "acospi_upward":
 double: 2
-float: 1
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index e454a63..5ed5112 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -83,25 +83,21 @@ float: 2
 
 Function: "acospi":
 double: 2
-float: 1
 float128: 2
 ldouble: 3
 
 Function: "acospi_downward":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_towardzero":
 double: 1
-float: 2
 float128: 1
 ldouble: 3
 
 Function: "acospi_upward":
 double: 2
-float: 1
 float128: 2
 ldouble: 2