[15/23] math: Use acoshf from CORE-MATH

Message ID 20241129132032.476978-16-adhemerval.zanella@linaro.org
State Superseded
Delegated to: DJ Delorie
Headers
Series Add remaining CORE-MATH binary32 implementations to libm |

Commit Message

Adhemerval Zanella Netto Nov. 29, 2024, 1:17 p.m. UTC
  The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows slight better performance to the generic acoshf.

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                      61.2471        58.7742         4.04%
x86_64-v2                   62.6519        59.0523         5.75%
x86_64-v3                   58.7408        50.1393        14.64%
aarch64                     24.8580        21.3317        14.19%
power10                     17.0469        13.1345        22.95%

reciprocal-throughput        master        patched   improvement
x86_64                      16.1618        15.1864         6.04%
x86_64-v2                   15.7729        14.7563         6.45%
x86_64-v3                   14.1669        11.9568        15.60%
aarch64                      10.911        9.5486         12.49%
power10                     6.38196        5.06734        20.60%

Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
---
 SHARED-FILES                           |   4 +
 sysdeps/aarch64/libm-test-ulps         |   4 -
 sysdeps/alpha/fpu/libm-test-ulps       |   4 -
 sysdeps/arc/fpu/libm-test-ulps         |   4 -
 sysdeps/arc/nofpu/libm-test-ulps       |   1 -
 sysdeps/arm/libm-test-ulps             |   4 -
 sysdeps/csky/fpu/libm-test-ulps        |   4 -
 sysdeps/csky/nofpu/libm-test-ulps      |   4 -
 sysdeps/hppa/fpu/libm-test-ulps        |   4 -
 sysdeps/i386/fpu/e_acoshf.S            | 101 -----------
 sysdeps/ieee754/flt-32/e_acoshf.c      | 230 ++++++++++++++++++++-----
 sysdeps/loongarch/lp64/libm-test-ulps  |   4 -
 sysdeps/m68k/m680x0/fpu/libm-test-ulps |   4 -
 sysdeps/microblaze/libm-test-ulps      |   1 -
 sysdeps/mips/mips32/libm-test-ulps     |   4 -
 sysdeps/mips/mips64/libm-test-ulps     |   4 -
 sysdeps/or1k/fpu/libm-test-ulps        |   4 -
 sysdeps/or1k/nofpu/libm-test-ulps      |   4 -
 sysdeps/powerpc/fpu/libm-test-ulps     |   4 -
 sysdeps/powerpc/nofpu/libm-test-ulps   |   4 -
 sysdeps/riscv/nofpu/libm-test-ulps     |   4 -
 sysdeps/riscv/rvd/libm-test-ulps       |   4 -
 sysdeps/s390/fpu/libm-test-ulps        |   4 -
 sysdeps/sh/libm-test-ulps              |   2 -
 sysdeps/sparc/fpu/libm-test-ulps       |   4 -
 sysdeps/x86_64/fpu/libm-test-ulps      |   4 -
 26 files changed, 194 insertions(+), 225 deletions(-)
 delete mode 100644 sysdeps/i386/fpu/e_acoshf.S
  

Patch

diff --git a/SHARED-FILES b/SHARED-FILES
index 3d32f0f3b6..c29c1c8106 100644
--- a/SHARED-FILES
+++ b/SHARED-FILES
@@ -298,3 +298,7 @@  sysdeps/ieee754/flt-32/e_acosf.c:
   (src/binary32/acos/acosf.c in CORE-MATH)
   - The code was adapted to use glibc code style and internal
     functions to handle errno, overflow, and underflow.
+sysdeps/ieee754/flt-32/e_acoshf.c:
+  (src/binary32/acosh/acoshf.c in CORE-MATH)
+  - The code was adapted to use glibc code style and internal
+    functions to handle errno, overflow, and underflow.
diff --git a/sysdeps/aarch64/libm-test-ulps b/sysdeps/aarch64/libm-test-ulps
index a593a4f8cb..902740f848 100644
--- a/sysdeps/aarch64/libm-test-ulps
+++ b/sysdeps/aarch64/libm-test-ulps
@@ -27,7 +27,6 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_advsimd":
@@ -36,7 +35,6 @@  float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_sve":
@@ -45,12 +43,10 @@  float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/alpha/fpu/libm-test-ulps b/sysdeps/alpha/fpu/libm-test-ulps
index aac2e31951..d26fcb40c9 100644
--- a/sysdeps/alpha/fpu/libm-test-ulps
+++ b/sysdeps/alpha/fpu/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/arc/fpu/libm-test-ulps b/sysdeps/arc/fpu/libm-test-ulps
index b80f97b5ed..0057716df0 100644
--- a/sysdeps/arc/fpu/libm-test-ulps
+++ b/sysdeps/arc/fpu/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 3
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 3
-float: 3
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/arc/nofpu/libm-test-ulps b/sysdeps/arc/nofpu/libm-test-ulps
index de1fdd9b1e..2fa0c82346 100644
--- a/sysdeps/arc/nofpu/libm-test-ulps
+++ b/sysdeps/arc/nofpu/libm-test-ulps
@@ -6,7 +6,6 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/arm/libm-test-ulps b/sysdeps/arm/libm-test-ulps
index a2e6549e65..968efaf299 100644
--- a/sysdeps/arm/libm-test-ulps
+++ b/sysdeps/arm/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/csky/fpu/libm-test-ulps b/sysdeps/csky/fpu/libm-test-ulps
index 3e7b07e81a..ebde72f4d7 100644
--- a/sysdeps/csky/fpu/libm-test-ulps
+++ b/sysdeps/csky/fpu/libm-test-ulps
@@ -14,19 +14,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 float: 1
diff --git a/sysdeps/csky/nofpu/libm-test-ulps b/sysdeps/csky/nofpu/libm-test-ulps
index 58bd5eaa15..73b6a2042a 100644
--- a/sysdeps/csky/nofpu/libm-test-ulps
+++ b/sysdeps/csky/nofpu/libm-test-ulps
@@ -14,19 +14,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 float: 1
diff --git a/sysdeps/hppa/fpu/libm-test-ulps b/sysdeps/hppa/fpu/libm-test-ulps
index ecef024df5..8a84b303b7 100644
--- a/sysdeps/hppa/fpu/libm-test-ulps
+++ b/sysdeps/hppa/fpu/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/i386/fpu/e_acoshf.S b/sysdeps/i386/fpu/e_acoshf.S
deleted file mode 100644
index 6eda9dc584..0000000000
--- a/sysdeps/i386/fpu/e_acoshf.S
+++ /dev/null
@@ -1,101 +0,0 @@ 
-/* ix87 specific implementation of arcsinh.
-   Copyright (C) 1996-2024 Free Software Foundation, Inc.
-   This file is part of the GNU C Library.
-
-   The GNU C Library is free software; you can redistribute it and/or
-   modify it under the terms of the GNU Lesser General Public
-   License as published by the Free Software Foundation; either
-   version 2.1 of the License, or (at your option) any later version.
-
-   The GNU C Library is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Lesser General Public License for more details.
-
-   You should have received a copy of the GNU Lesser General Public
-   License along with the GNU C Library; if not, see
-   <https://www.gnu.org/licenses/>.  */
-
-#include <machine/asm.h>
-#include <libm-alias-finite.h>
-
-	.section .rodata.cst8,"aM",@progbits,8
-
-	.p2align 3
-	.type one,@object
-one:	.double 1.0
-	ASM_SIZE_DIRECTIVE(one)
-	.type limit,@object
-limit:	.double 0.29
-	ASM_SIZE_DIRECTIVE(limit)
-
-#ifdef PIC
-#define MO(op) op##@GOTOFF(%edx)
-#else
-#define MO(op) op
-#endif
-
-	.text
-ENTRY(__ieee754_acoshf)
-	movl	4(%esp), %ecx
-	cmpl	$0x3f800000, %ecx
-	jl	5f			// < 1 => invalid
-	fldln2				// log(2)
-	flds	4(%esp)			// x : log(2)
-	cmpl	$0x47000000, %ecx
-	ja	3f			// x > 2^14
-#ifdef	PIC
-	LOAD_PIC_REG (dx)
-#endif
-	cmpl	$0x40000000, %ecx
-	ja	4f			// x > 2
-
-	// 1 <= x <= 2 => y = log1p(x-1+sqrt(2*(x-1)+(x-1)^2))
-	fsubl	MO(one)			// x-1 : log(2)
-	fabs				// acosh(1) is +0 in all rounding modes
-	fld	%st			// x-1 : x-1 : log(2)
-	fmul	%st(1)			// (x-1)^2 : x-1 : log(2)
-	fadd	%st(1)			// x-1+(x-1)^2 : x-1 : log(2)
-	fadd	%st(1)			// 2*(x-1)+(x-1)^2 : x-1 : log(2)
-	fsqrt				// sqrt(2*(x-1)+(x-1)^2) : x-1 : log(2)
-	faddp				// x-1+sqrt(2*(x-1)+(x-1)^2) : log(2)
-	fcoml	MO(limit)
-	fnstsw
-	sahf
-	ja	2f
-	fyl2xp1				// log1p(x-1+sqrt(2*(x-1)+(x-1)^2))
-	ret
-
-2:	faddl	MO(one)			// x+sqrt(2*(x-1)+(x-1)^2) : log(2)
-	fyl2x				// log(x+sqrt(2*(x-1)+(x-1)^2))
-	ret
-
-	// x > 2^14 => y = log(x) + log(2)
-	.align ALIGNARG(4)
-3:	fyl2x				// log(x)
-	fldln2				// log(2) : log(x)
-	faddp				// log(x)+log(2)
-	ret
-
-	// 2^28 > x > 2 => y = log(2*x - 1/(x+sqrt(x*x-1)))
-	.align ALIGNARG(4)
-4:	fld	%st			// x : x : log(2)
-	fadd	%st, %st(1)		// x : 2*x : log(2)
-	fld	%st			// x : x : 2*x : log(2)
-	fmul	%st(1)			// x^2 : x : 2*x : log(2)
-	fsubl	MO(one)			// x^2-1 : x : 2*x : log(2)
-	fsqrt				// sqrt(x^2-1) : x : 2*x : log(2)
-	faddp				// x+sqrt(x^2-1) : 2*x : log(2)
-	fdivrl	MO(one)			// 1/(x+sqrt(x^2-1)) : 2*x : log(2)
-	fsubrp				// 2*x+1/(x+sqrt(x^2)-1) : log(2)
-	fyl2x				// log(2*x+1/(x+sqrt(x^2-1)))
-	ret
-
-	// x < 1 (or -NaN) => NaN
-	.align ALIGNARG(4)
-5:	flds	4(%esp)
-	fsub	%st
-	fdiv	%st, %st(0)
-	ret
-END(__ieee754_acoshf)
-libm_alias_finite (__ieee754_acoshf, __acoshf)
diff --git a/sysdeps/ieee754/flt-32/e_acoshf.c b/sysdeps/ieee754/flt-32/e_acoshf.c
index 2237af4ba4..c92e66d604 100644
--- a/sysdeps/ieee754/flt-32/e_acoshf.c
+++ b/sysdeps/ieee754/flt-32/e_acoshf.c
@@ -1,49 +1,199 @@ 
-/* e_acoshf.c -- float version of e_acosh.c.
- */
-
-/*
- * ====================================================
- * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
- *
- * Developed at SunPro, a Sun Microsystems, Inc. business.
- * Permission to use, copy, modify, and distribute this
- * software is freely granted, provided that this notice
- * is preserved.
- * ====================================================
- */
-
-#if defined(LIBM_SCCS) && !defined(lint)
-static char rcsid[] = "$NetBSD: e_acoshf.c,v 1.5 1995/05/12 04:57:20 jtc Exp $";
-#endif
+/* Correctly-rounded inverse hyperbolic cosine function for binary32 format.
+
+Copyright (c) 2023-2024 Alexei Sibidanov.
+
+The original version of this file was copied from the CORE-MATH
+project (file src/binary32/acosh/acoshf.c, revision 927eb45).
+
+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 <math_private.h>
+#include <stdint.h>
 #include <libm-alias-finite.h>
+#include "math_config.h"
 
-static const float
-one	= 1.0,
-ln2	= 6.9314718246e-01;  /* 0x3f317218 */
+static __attribute__ ((noinline)) float
+as_special (float x)
+{
+  uint32_t t = asuint (x);
+  if (t == 0x3f800000u)
+    return 0.0f;
+  if ((t << 1) > 0xff000000u)
+    return x + x; /* nan */
+  if (t == 0x7f800000u)
+    return x; /* inf */
+  return __math_invalidf (x);
+}
 
-float __ieee754_acoshf(float x)
+float
+__ieee754_acoshf (float x)
 {
-	float t;
-	int32_t hx;
-	GET_FLOAT_WORD(hx,x);
-	if(hx<0x3f800000) {		/* x < 1 */
-	    return (x-x)/(x-x);
-	} else if(hx >=0x4d800000) {	/* x > 2**28 */
-	    if(hx >=0x7f800000) {	/* x is inf of NaN */
-		return x+x;
-	    } else
-		return __ieee754_logf(x)+ln2;	/* acosh(huge)=log(2x) */
-	} else if (hx==0x3f800000) {
-	    return 0.0;			/* acosh(1) = 0 */
-	} else if (hx > 0x40000000) {	/* 2**28 > x > 2 */
-	    t=x*x;
-	    return __ieee754_logf((float)2.0*x-one/(x+sqrtf(t-one)));
-	} else {			/* 1<x<2 */
-	    t = x-one;
-	    return __log1pf(t+sqrtf((float)2.0*t+t*t));
+  static const double ix[] =
+    {
+      0x1p+0,           0x1.fc07f01fcp-1, 0x1.f81f81f82p-1, 0x1.f44659e4ap-1,
+      0x1.f07c1f07cp-1, 0x1.ecc07b302p-1, 0x1.e9131abfp-1,  0x1.e573ac902p-1,
+      0x1.e1e1e1e1ep-1, 0x1.de5d6e3f8p-1, 0x1.dae6076bap-1, 0x1.d77b654b8p-1,
+      0x1.d41d41d42p-1, 0x1.d0cb58f6ep-1, 0x1.cd8568904p-1, 0x1.ca4b3055ep-1,
+      0x1.c71c71c72p-1, 0x1.c3f8f01c4p-1, 0x1.c0e070382p-1, 0x1.bdd2b8994p-1,
+      0x1.bacf914c2p-1, 0x1.b7d6c3ddap-1, 0x1.b4e81b4e8p-1, 0x1.b2036406cp-1,
+      0x1.af286bca2p-1, 0x1.ac5701ac6p-1, 0x1.a98ef606ap-1, 0x1.a6d01a6dp-1,
+      0x1.a41a41a42p-1, 0x1.a16d3f97ap-1, 0x1.9ec8e951p-1,  0x1.9c2d14ee4p-1,
+      0x1.99999999ap-1, 0x1.970e4f80cp-1, 0x1.948b0fcd6p-1, 0x1.920fb49dp-1,
+      0x1.8f9c18f9cp-1, 0x1.8d3018d3p-1,  0x1.8acb90f6cp-1, 0x1.886e5f0acp-1,
+      0x1.861861862p-1, 0x1.83c977ab2p-1, 0x1.818181818p-1, 0x1.7f405fd02p-1,
+      0x1.7d05f417ep-1, 0x1.7ad2208ep-1,  0x1.78a4c8178p-1, 0x1.767dce434p-1,
+      0x1.745d1745ep-1, 0x1.724287f46p-1, 0x1.702e05c0cp-1, 0x1.6e1f76b44p-1,
+      0x1.6c16c16c2p-1, 0x1.6a13cd154p-1, 0x1.681681682p-1, 0x1.661ec6a52p-1,
+      0x1.642c8590cp-1, 0x1.623fa7702p-1, 0x1.605816058p-1, 0x1.5e75bb8dp-1,
+      0x1.5c9882b94p-1, 0x1.5ac056b02p-1, 0x1.58ed23082p-1, 0x1.571ed3c5p-1,
+      0x1.555555556p-1, 0x1.5390948f4p-1, 0x1.51d07eae2p-1, 0x1.501501502p-1,
+      0x1.4e5e0a73p-1,  0x1.4cab88726p-1, 0x1.4afd6a052p-1, 0x1.49539e3b2p-1,
+      0x1.47ae147aep-1, 0x1.460cbc7f6p-1, 0x1.446f86562p-1, 0x1.42d6625d6p-1,
+      0x1.414141414p-1, 0x1.3fb013fbp-1,  0x1.3e22cbce4p-1, 0x1.3c995a47cp-1,
+      0x1.3b13b13b2p-1, 0x1.3991c2c18p-1, 0x1.381381382p-1, 0x1.3698df3dep-1,
+      0x1.3521cfb2cp-1, 0x1.33ae45b58p-1, 0x1.323e34a2cp-1, 0x1.30d19013p-1,
+      0x1.2f684bda2p-1, 0x1.2e025c04cp-1, 0x1.2c9fb4d82p-1, 0x1.2b404ad02p-1,
+      0x1.29e4129e4p-1, 0x1.288b01288p-1, 0x1.27350b882p-1, 0x1.25e22708p-1,
+      0x1.24924924ap-1, 0x1.23456789ap-1, 0x1.21fb78122p-1, 0x1.20b470c68p-1,
+      0x1.1f7047dc2p-1, 0x1.1e2ef3b4p-1,  0x1.1cf06ada2p-1, 0x1.1bb4a4046p-1,
+      0x1.1a7b9611ap-1, 0x1.19453808cp-1, 0x1.181181182p-1, 0x1.16e068942p-1,
+      0x1.15b1e5f76p-1, 0x1.1485f0e0ap-1, 0x1.135c81136p-1, 0x1.12358e75ep-1,
+      0x1.111111112p-1, 0x1.0fef010fep-1, 0x1.0ecf56be6p-1, 0x1.0db20a89p-1,
+      0x1.0c9714fbcp-1, 0x1.0b7e6ec26p-1, 0x1.0a6810a68p-1, 0x1.0953f3902p-1,
+      0x1.084210842p-1, 0x1.073260a48p-1, 0x1.0624dd2f2p-1, 0x1.05197f7d8p-1,
+      0x1.041041042p-1, 0x1.03091b52p-1,  0x1.020408102p-1, 0x1.01010101p-1,
+      0x1p-1
+    };
+
+  static const double lix[] =
+    {
+      0x0p+0,               0x1.fe02a6b146789p-8, 0x1.fc0a8b0fa03e4p-7,
+      0x1.7b91b07de311bp-6, 0x1.f829b0e7c33p-6,   0x1.39e87b9fd7d6p-5,
+      0x1.77458f63edcfcp-5, 0x1.b42dd7117b1bfp-5, 0x1.f0a30c01362a6p-5,
+      0x1.16536eea7fae1p-4, 0x1.341d7961791d1p-4, 0x1.51b073f07983fp-4,
+      0x1.6f0d28ae3eb4cp-4, 0x1.8c345d6383b21p-4, 0x1.a926d3a475563p-4,
+      0x1.c5e548f63a743p-4, 0x1.e27076e28f2e6p-4, 0x1.fec9131dbaabbp-4,
+      0x1.0d77e7ccf6e59p-3, 0x1.1b72ad52f87ap-3,  0x1.29552f81eb523p-3,
+      0x1.371fc201f7f74p-3, 0x1.44d2b6ccbfd1ep-3, 0x1.526e5e3a41438p-3,
+      0x1.5ff3070a613d4p-3, 0x1.6d60fe717221dp-3, 0x1.7ab890212b909p-3,
+      0x1.87fa065214911p-3, 0x1.9525a9cf296b4p-3, 0x1.a23bc1fe42563p-3,
+      0x1.af3c94e81bff3p-3, 0x1.bc2867430acd6p-3, 0x1.c8ff7c7989a22p-3,
+      0x1.d5c216b535b91p-3, 0x1.e27076e2f92e6p-3, 0x1.ef0adcbe0d936p-3,
+      0x1.fb9186d5ebe2bp-3, 0x1.0402594b51041p-2, 0x1.0a324e27370e3p-2,
+      0x1.1058bf9ad7ad5p-2, 0x1.1675cabaa660ep-2, 0x1.1c898c16b91fbp-2,
+      0x1.22941fbcfb966p-2, 0x1.2895a13dd2ea3p-2, 0x1.2e8e2bade7d31p-2,
+      0x1.347dd9a9afd55p-2, 0x1.3a64c556b05eap-2, 0x1.40430868877e4p-2,
+      0x1.4618bc219dec2p-2, 0x1.4be5f9579e0a1p-2, 0x1.51aad872c982dp-2,
+      0x1.5767717432a6cp-2, 0x1.5d1bdbf5669cap-2, 0x1.62c82f2b83795p-2,
+      0x1.686c81e9964afp-2, 0x1.6e08eaa2929e4p-2, 0x1.739d7f6b95007p-2,
+      0x1.792a55fdb7fa2p-2, 0x1.7eaf83b82efc3p-2, 0x1.842d1da1ecb17p-2,
+      0x1.89a3386be825bp-2, 0x1.8f11e87347ac7p-2, 0x1.947941c1f26fbp-2,
+      0x1.99d958119208bp-2, 0x1.9f323ecbd984cp-2, 0x1.a484090e5eb0ap-2,
+      0x1.a9cec9a9cf84ap-2, 0x1.af1293245606bp-2, 0x1.b44f77bc98f63p-2,
+      0x1.b9858969218fbp-2, 0x1.beb4d9da96b7cp-2, 0x1.c3dd7a7d0354dp-2,
+      0x1.c8ff7c79ada22p-2, 0x1.ce1af0b855bebp-2, 0x1.d32fe7e039bd5p-2,
+      0x1.d83e72587673ep-2, 0x1.dd46a04c204a1p-2, 0x1.e24881a7cac26p-2,
+      0x1.e744261d8a788p-2, 0x1.ec399d2457ccp-2,  0x1.f128f5fac86edp-2,
+      0x1.f6123fa71c8acp-2, 0x1.faf588f76631fp-2, 0x1.ffd2e08580c98p-2,
+      0x1.02552a5a4f0ffp-1, 0x1.04bdf9da8b6d2p-1, 0x1.0723e5c1b4f4p-1,
+      0x1.0986f4f589521p-1, 0x1.0be72e423ca83p-1, 0x1.0e44985d0f48cp-1,
+      0x1.109f39e2be497p-1, 0x1.12f71959283bcp-1, 0x1.154c3d2f4f5eap-1,
+      0x1.179eabbd9c9a1p-1, 0x1.19ee6b466516fp-1, 0x1.1c3b81f723c25p-1,
+      0x1.1e85f5e6ec0dp-1,  0x1.20cdcd193f76ep-1, 0x1.23130d7beb743p-1,
+      0x1.2555bce9887cbp-1, 0x1.2795e1288211bp-1, 0x1.29d37fec2308bp-1,
+      0x1.2c0e9ed45768cp-1, 0x1.2e47436e5ae68p-1, 0x1.307d7334ff0bep-1,
+      0x1.32b1339134571p-1, 0x1.34e289d9b39d3p-1, 0x1.37117b5481bb6p-1,
+      0x1.393e0d3549a1ap-1, 0x1.3b6844a017823p-1, 0x1.3d9026a70eefbp-1,
+      0x1.3fb5b84cfeb42p-1, 0x1.41d8fe844b2aep-1, 0x1.43f9fe2fb9267p-1,
+      0x1.4618bc21d86c2p-1, 0x1.48353d1e928dfp-1, 0x1.4a4f85db1debbp-1,
+      0x1.4c679afcc323ap-1, 0x1.4e7d811b77bb1p-1, 0x1.50913cbff8c6bp-1,
+      0x1.52a2d265be5abp-1, 0x1.54b2467998498p-1, 0x1.56bf9d5b34b99p-1,
+      0x1.58cadb5cbe989p-1, 0x1.5ad404c33af2dp-1, 0x1.5cdb1dc6ad765p-1,
+      0x1.5ee02a9241e75p-1, 0x1.60e32f447a8d9p-1, 0x1.62e42fefa39efp-1
+    };
+
+  uint32_t t = asuint (x);
+  if (__glibc_unlikely (t <= 0x3f800000u))
+    return as_special (x);
+  else if (__glibc_unlikely (t <= 0x3f99db23u))   /* x <= 0x1.33b646p+0 */
+    {
+      float zf = x - 1.0f;
+      double z = zf;
+      double a = sqrt (2.0 * z);
+      static const double c[] =
+         {
+	   -0x1.555555555491ep-4,  0x1.333333319c2p-6,
+	   -0x1.6db6da26e5e0ep-8,  0x1.f1c698b7100cep-10,
+	   -0x1.6e7ca9e152dc2p-11, 0x1.1b4cf187158bdp-12,
+	   -0x1.b5137f11a88e5p-14, 0x1.056c6c57152b2p-15
+	 };
+      double z2 = z * z, z4 = z2 * z2;
+      double f = ((c[0] + z * c[1]) + z2 * (c[2] + z * c[3]))
+		 + z4 * ((c[4] + z * c[5]) + z2 * (c[6] + z * c[7]));
+      double r = a + (a * z) * f;
+      return r;
+    }
+  else if (__glibc_likely (t < 0x7f800000u))
+    {
+      double xd = x, x2 = xd * xd;
+      uint64_t tp = asuint64 (xd + sqrt (x2 - 1));
+      uint64_t m = tp & (~UINT64_C(0) >> 12);
+      int j = (m + (UINT64_C(1) << (52 - 8))) >> (52 - 7);
+      int e = (tp >> 52) - 0x3ff;
+      double w = asdouble (m | UINT64_C(0x3ff) << 52);
+      double z = w * ix[j] - 1.0;
+      static const double c[] =
+	{
+	  0x1.0000000066947p+0, -0x1.00007f053d8cbp-1, 0x1.555280111d914p-2
+	};
+      double z2 = z * z;
+      double r = ((lix[128] * e + lix[j]) + z * c[0])
+		  + z2 * (c[1] + z * c[2]);
+      if (__glibc_unlikely (((asuint64 (r) + 259000) & INT64_C(0xfffffff))
+			    < 260000))
+	{
+	  static const double cp[] =
+	    {
+	       0x1p+0,               -0x1p-1,
+	       0x1.55555555030bcp-2, -0x1.ffffffff2b4e5p-3,
+	       0x1.999b5076a42f2p-3, -0x1.55570c45a647dp-3
+	    };
+	  z2 = z * z;
+	  double c0 = cp[0] + z * cp[1];
+	  double c2 = cp[2] + z * cp[3];
+	  double c4 = cp[4] + z * cp[5];
+	  c0 += z2 * (c2 + z2 * c4);
+	  const double ln2l = 0x1.7f7d1cf79abcap-20;
+	  const double ln2h = 0x1.62e4p-1;
+	  double Lh = ln2h * e;
+	  double Ll = ln2l * e;
+	  r = fma (z, c0, Ll + lix[j]) + Lh;
+	  if (__glibc_unlikely ((asuint64 (r) & UINT64_C(0xfffffff)) == 0))
+	    {
+	      double h = fma (z, c0, Ll + lix[j]) + (Lh - r);
+	      r = r + 64.0 * h;
+	    }
 	}
+      return r;
+    }
+  else
+    return as_special (x);
 }
 libm_alias_finite (__ieee754_acoshf, __acoshf)
diff --git a/sysdeps/loongarch/lp64/libm-test-ulps b/sysdeps/loongarch/lp64/libm-test-ulps
index 67bf2d2f9f..63fd07fbe0 100644
--- a/sysdeps/loongarch/lp64/libm-test-ulps
+++ b/sysdeps/loongarch/lp64/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/m68k/m680x0/fpu/libm-test-ulps b/sysdeps/m68k/m680x0/fpu/libm-test-ulps
index 50080f46b3..73019999e0 100644
--- a/sysdeps/m68k/m680x0/fpu/libm-test-ulps
+++ b/sysdeps/m68k/m680x0/fpu/libm-test-ulps
@@ -3,22 +3,18 @@ 
 # Maximal error of functions:
 Function: "acosh":
 double: 1
-float: 1
 ldouble: 1
 
 Function: "acosh_downward":
 double: 2
-float: 1
 ldouble: 1
 
 Function: "acosh_towardzero":
 double: 2
-float: 1
 ldouble: 1
 
 Function: "acosh_upward":
 double: 2
-float: 1
 ldouble: 1
 
 Function: "asin":
diff --git a/sysdeps/microblaze/libm-test-ulps b/sysdeps/microblaze/libm-test-ulps
index c93ebfa21b..a9e3f3a318 100644
--- a/sysdeps/microblaze/libm-test-ulps
+++ b/sysdeps/microblaze/libm-test-ulps
@@ -5,7 +5,6 @@  Function: "acos":
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "asin":
 float: 1
diff --git a/sysdeps/mips/mips32/libm-test-ulps b/sysdeps/mips/mips32/libm-test-ulps
index d23762642b..fc0d001750 100644
--- a/sysdeps/mips/mips32/libm-test-ulps
+++ b/sysdeps/mips/mips32/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/mips/mips64/libm-test-ulps b/sysdeps/mips/mips64/libm-test-ulps
index aece335d6b..855b48e800 100644
--- a/sysdeps/mips/mips64/libm-test-ulps
+++ b/sysdeps/mips/mips64/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/or1k/fpu/libm-test-ulps b/sysdeps/or1k/fpu/libm-test-ulps
index 862ab6b559..0cc7beccfc 100644
--- a/sysdeps/or1k/fpu/libm-test-ulps
+++ b/sysdeps/or1k/fpu/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/or1k/nofpu/libm-test-ulps b/sysdeps/or1k/nofpu/libm-test-ulps
index 3c364480f4..a0c3179fc3 100644
--- a/sysdeps/or1k/nofpu/libm-test-ulps
+++ b/sysdeps/or1k/nofpu/libm-test-ulps
@@ -15,19 +15,15 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 
 Function: "asin":
 double: 1
diff --git a/sysdeps/powerpc/fpu/libm-test-ulps b/sysdeps/powerpc/fpu/libm-test-ulps
index 135853f213..05facea37f 100644
--- a/sysdeps/powerpc/fpu/libm-test-ulps
+++ b/sysdeps/powerpc/fpu/libm-test-ulps
@@ -23,25 +23,21 @@  ldouble: 2
 
 Function: "acosh":
 double: 2
-float: 2
 float128: 4
 ldouble: 2
 
 Function: "acosh_downward":
 double: 2
-float: 2
 float128: 3
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 float128: 2
 ldouble: 5
 
 Function: "acosh_upward":
 double: 2
-float: 2
 float128: 3
 ldouble: 4
 
diff --git a/sysdeps/powerpc/nofpu/libm-test-ulps b/sysdeps/powerpc/nofpu/libm-test-ulps
index 44cf9983dc..f553da0caa 100644
--- a/sysdeps/powerpc/nofpu/libm-test-ulps
+++ b/sysdeps/powerpc/nofpu/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 2
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 1
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 5
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "add_ldouble":
diff --git a/sysdeps/riscv/nofpu/libm-test-ulps b/sysdeps/riscv/nofpu/libm-test-ulps
index 8d3b048e84..3bf997c5d8 100644
--- a/sysdeps/riscv/nofpu/libm-test-ulps
+++ b/sysdeps/riscv/nofpu/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "asin":
diff --git a/sysdeps/riscv/rvd/libm-test-ulps b/sysdeps/riscv/rvd/libm-test-ulps
index a1d85be3b6..f85d9caf03 100644
--- a/sysdeps/riscv/rvd/libm-test-ulps
+++ b/sysdeps/riscv/rvd/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/s390/fpu/libm-test-ulps b/sysdeps/s390/fpu/libm-test-ulps
index 7a3d526d69..b788f5730b 100644
--- a/sysdeps/s390/fpu/libm-test-ulps
+++ b/sysdeps/s390/fpu/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/sh/libm-test-ulps b/sysdeps/sh/libm-test-ulps
index d1e5b05531..d46c722ea1 100644
--- a/sysdeps/sh/libm-test-ulps
+++ b/sysdeps/sh/libm-test-ulps
@@ -8,11 +8,9 @@  double: 1
 
 Function: "acosh":
 double: 2
-float: 2
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 
 Function: "asin":
 float: 1
diff --git a/sysdeps/sparc/fpu/libm-test-ulps b/sysdeps/sparc/fpu/libm-test-ulps
index 6a70b6f5ea..8faf5357df 100644
--- a/sysdeps/sparc/fpu/libm-test-ulps
+++ b/sysdeps/sparc/fpu/libm-test-ulps
@@ -19,22 +19,18 @@  ldouble: 1
 
 Function: "acosh":
 double: 2
-float: 2
 ldouble: 4
 
 Function: "acosh_downward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 ldouble: 2
 
 Function: "acosh_upward":
 double: 2
-float: 2
 ldouble: 3
 
 Function: "asin":
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index ff532cea1e..85d5e8fcd8 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -43,25 +43,21 @@  float: 1
 
 Function: "acosh":
 double: 2
-float: 2
 float128: 4
 ldouble: 3
 
 Function: "acosh_downward":
 double: 2
-float: 2
 float128: 3
 ldouble: 4
 
 Function: "acosh_towardzero":
 double: 2
-float: 2
 float128: 2
 ldouble: 4
 
 Function: "acosh_upward":
 double: 2
-float: 2
 float128: 3
 ldouble: 3