From patchwork Wed Dec 4 16:37:57 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adhemerval Zanella X-Patchwork-Id: 102407 Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 48A4F3858C66 for ; Wed, 4 Dec 2024 16:56:21 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 48A4F3858C66 Authentication-Results: sourceware.org; dkim=pass (2048-bit key, unprotected) header.d=linaro.org header.i=@linaro.org header.a=rsa-sha256 header.s=google header.b=U2OxHome X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mail-pl1-x636.google.com (mail-pl1-x636.google.com [IPv6:2607:f8b0:4864:20::636]) by sourceware.org (Postfix) with ESMTPS id 36D5F3858431 for ; Wed, 4 Dec 2024 16:41:11 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 36D5F3858431 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=linaro.org Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=linaro.org ARC-Filter: OpenARC Filter v1.0.0 sourceware.org 36D5F3858431 Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=2607:f8b0:4864:20::636 ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1733330471; cv=none; b=wmXXOhIzKYuMtQcszQeI3xfHp/BBq2AOPuv0bPhiQ06TRGjiKKC8+HWROdmPMbL6/EFz4A+8JXUgcdmy0hhfkF3ZbS9Z7EEKhxDTr+LegUAT2MRE1meKRT8rtNGG4HoVX8rpE/wIvYEx54W9oljEDbmscEPVMGhlIALzIeXCQYo= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1733330471; c=relaxed/simple; bh=/4U7X0FTzM8RBeBfDGXno+DGp4k7OnTW0ccudZnlzh4=; h=DKIM-Signature:From:To:Subject:Date:Message-ID:MIME-Version; b=rtQFDd9qvSdqgFx+W0IrOZ9r1+lJkcJDFOzAbZ0i/sIZTMXZ2bVEgFeJozf2cuXWirjn9eE3JDG89ppRD+tvj3SMUUp/9CxO/5ZgEo2EhfreeZ6521OEpwpEtlL9hL2gW5TAsGgy9WuFrrGBEuik1lZPCvKbGCLf87cK9Ko9nI4= ARC-Authentication-Results: i=1; server2.sourceware.org Received: by mail-pl1-x636.google.com with SMTP id d9443c01a7336-2154e3af730so45108035ad.3 for ; Wed, 04 Dec 2024 08:41:11 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; t=1733330444; x=1733935244; darn=sourceware.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to; bh=MXxe3ZyGE5hpAGXJTDF/AckArrG5bA/J0/jeDkwfd5w=; b=U2OxHomeXPG5Iy6IzTblfEvSVNIHfNV4gLrAM0k536Rkkkde0Ctb2ud9UqBf0CqxRx 5GFIZUdV0a8kkT3NlpouB4X73NGZbFOcX+CE4jLRpvQmAvFolHJ0Wf+QMxM8vX1/ZW3d GiLhmMZPruKkmNpLJn9heQTLt0cFaLaXi7VAOdFq0Q285BD78KcrBXSMOaAj8pJ8jH33 BybN/JTHYz1Cacd+G2XrDKbs48L6ZyrI/uOD138kmXdWWJfeom3+yhtps2RTOc4iX9om pk2uM5fjZtabaohF6dUln0D7FJtTj3NosCZPpNdG8s7eY9Ym4BaCammPlUP1iivE5c3J IMhA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1733330444; x=1733935244; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=MXxe3ZyGE5hpAGXJTDF/AckArrG5bA/J0/jeDkwfd5w=; b=j7QAquc5F4Z6YT56VaV03FhkxKALN+dozmDjstnbPmdvBJHsiBhhTmMD99wE4KZNDa ZEYiPvuhTpHEJo2KN5gIS0bcu6gVNGfyPyA6QyaXAATUG/gkt5kcOauA4EvFe2CMlFqJ 1nakJkQDWtfRS9a0WCzEhfvsznDmZ3eccdd+ObSxc1IFMY7jQRCW6Os5WWd+fhqTlIKK aouMRiWx8g+jz1ajKr+muTkBGF0Dl6JjWGM0xHSxEwjOS9VAzWUrR6znZuIVZGue3iQ5 0cpD90m8cGdreIPUNAbfLFWNuVbUNtKNu130ExBpwf1OX7gYIYOnRH3i/KJYYEPpomMx 4eFA== X-Gm-Message-State: AOJu0YwwSBUDQF9dq3p+y2GQGUy7LHrpQMFI7V3E0tkHHI0+5KAXtpTg zbNPDbyYNNDFeEfdqurmI/quOXiB4DLje7VtXJIjITGoW8oWWGOOBkVnbE/FFT9m2eltUE8Bhp1 s X-Gm-Gg: ASbGncsNyDo3MP3UUeCDDF7pFU7cEmxJgFwT0YLLH2X2XTxvVqVe2rQR4xwqZ8tSdJj peKAwisSlZixbqYdWkY6nXLMn/6lmKNrirZJEKmf2Z65QfrFW8iiYmSBn7kbxFW3I71U3G6yKE2 Lq3Biqq66DEbYm9qknmgHebbcQbYnPly0N3GMoSGcfglsruVgdoQY3ektEK1IOpFKr9Jr3c1o1J prpF8Idb2v14CHgCj8DzcU2R6EdO/70/CT53mSOdfvvgJmWJNwULCwJt7myHQ== X-Google-Smtp-Source: AGHT+IE3K5F1yA2CxvWuTRkz9ibz3BAqAhaXNrA6RqXcOZNBM+eAX7xW69QQFQgk3XJcUaZCovovmg== X-Received: by 2002:a17:902:d48a:b0:215:9894:5670 with SMTP id d9443c01a7336-215d003e8c2mr73542325ad.16.1733330443672; Wed, 04 Dec 2024 08:40:43 -0800 (PST) Received: from mandiga.. ([2804:1b3:a7c2:2d1:40a3:b587:af7c:a5cd]) by smtp.gmail.com with ESMTPSA id d9443c01a7336-21521967714sm114029225ad.140.2024.12.04.08.40.41 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 04 Dec 2024 08:40:43 -0800 (PST) From: Adhemerval Zanella To: libc-alpha@sourceware.org Cc: DJ Delorie , Joseph Myers , Alexei Sibidanov , Paul Zimmermann Subject: [PATCH v2 22/25] math: Use atanhf from CORE-MATH Date: Wed, 4 Dec 2024 13:37:57 -0300 Message-ID: <20241204163949.1408676-23-adhemerval.zanella@linaro.org> X-Mailer: git-send-email 2.43.0 In-Reply-To: <20241204163949.1408676-1-adhemerval.zanella@linaro.org> References: <20241204163949.1408676-1-adhemerval.zanella@linaro.org> MIME-Version: 1.0 X-Spam-Status: No, score=-12.8 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP, URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.30 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: libc-alpha-bounces~patchwork=sourceware.org@sourceware.org The CORE-MATH implementation is correctly rounded (for any rounding mode) and shows slight better performance to the generic atanhf. 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 59.4930 45.8568 22.92% x86_64v2 59.5705 45.5804 23.48% x86_64v3 53.1838 37.7155 29.08% i686 169.354 133.5940 21.12% aarch64 (Neoverse) 26.0781 16.9829 34.88% power10 15.6591 10.7623 31.27% reciprocal-throughput master patched improvement x86_64 23.5903 18.5766 21.25% x86_64v2 22.6489 18.2683 19.34% x86_64v3 19.0401 13.9474 26.75% i686 97.6034 107.3260 -9.96% aarch64 (Neoverse) 15.3664 9.57846 37.67% power10 6.8877 4.6242 32.86% Signed-off-by: Alexei Sibidanov Signed-off-by: Paul Zimmermann Signed-off-by: Adhemerval Zanella --- 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_atanhf.S | 110 --------- sysdeps/i386/fpu/libm-test-ulps | 3 - .../i386/i686/fpu/multiarch/libm-test-ulps | 3 - sysdeps/ieee754/flt-32/e_atanhf.c | 212 +++++++++++++----- sysdeps/loongarch/lp64/libm-test-ulps | 4 - sysdeps/m68k/coldfire/fpu/libm-test-ulps | 1 - 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 - 28 files changed, 161 insertions(+), 252 deletions(-) delete mode 100644 sysdeps/i386/fpu/e_atanhf.S diff --git a/SHARED-FILES b/SHARED-FILES index 99f8554393..d32c837b46 100644 --- a/SHARED-FILES +++ b/SHARED-FILES @@ -318,3 +318,7 @@ sysdeps/ieee754/flt-32/e_atan2f.c: (src/binary32/atan2/atan2f.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_atanhf.c: + (src/binary32/atanh/atanhf.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 9ba57d0d0d..7e9d23ab6f 100644 --- a/sysdeps/aarch64/libm-test-ulps +++ b/sysdeps/aarch64/libm-test-ulps @@ -146,7 +146,6 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_advsimd": @@ -155,7 +154,6 @@ float: 1 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_sve": @@ -164,12 +162,10 @@ float: 1 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/alpha/fpu/libm-test-ulps b/sysdeps/alpha/fpu/libm-test-ulps index d098eaf820..af11a87641 100644 --- a/sysdeps/alpha/fpu/libm-test-ulps +++ b/sysdeps/alpha/fpu/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/arc/fpu/libm-test-ulps b/sysdeps/arc/fpu/libm-test-ulps index 5b89ee2696..ef93b0bb21 100644 --- a/sysdeps/arc/fpu/libm-test-ulps +++ b/sysdeps/arc/fpu/libm-test-ulps @@ -75,19 +75,15 @@ double: 2 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/arc/nofpu/libm-test-ulps b/sysdeps/arc/nofpu/libm-test-ulps index ac43ab8dcc..0d2e660e09 100644 --- a/sysdeps/arc/nofpu/libm-test-ulps +++ b/sysdeps/arc/nofpu/libm-test-ulps @@ -20,7 +20,6 @@ Function: "atan2": Function: "atanh": double: 2 -float: 2 Function: "cabs": double: 1 diff --git a/sysdeps/arm/libm-test-ulps b/sysdeps/arm/libm-test-ulps index b0b0b3ab05..0a95004f36 100644 --- a/sysdeps/arm/libm-test-ulps +++ b/sysdeps/arm/libm-test-ulps @@ -74,19 +74,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/csky/fpu/libm-test-ulps b/sysdeps/csky/fpu/libm-test-ulps index b8de7af6b9..f1b62e3da4 100644 --- a/sysdeps/csky/fpu/libm-test-ulps +++ b/sysdeps/csky/fpu/libm-test-ulps @@ -71,19 +71,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/csky/nofpu/libm-test-ulps b/sysdeps/csky/nofpu/libm-test-ulps index a7594b6b1c..9c2bfc6a4a 100644 --- a/sysdeps/csky/nofpu/libm-test-ulps +++ b/sysdeps/csky/nofpu/libm-test-ulps @@ -71,19 +71,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/hppa/fpu/libm-test-ulps b/sysdeps/hppa/fpu/libm-test-ulps index 61ce339fcc..5730dd3acb 100644 --- a/sysdeps/hppa/fpu/libm-test-ulps +++ b/sysdeps/hppa/fpu/libm-test-ulps @@ -74,19 +74,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/i386/fpu/e_atanhf.S b/sysdeps/i386/fpu/e_atanhf.S deleted file mode 100644 index c94dc7ba51..0000000000 --- a/sysdeps/i386/fpu/e_atanhf.S +++ /dev/null @@ -1,110 +0,0 @@ -/* ix87 specific implementation of arctanh function. - 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 - . */ - -#include -#include -#include - - .section .rodata - - .align ALIGNARG(4) - .type half,@object -half: .double 0.5 - ASM_SIZE_DIRECTIVE(half) - .type one,@object -one: .double 1.0 - ASM_SIZE_DIRECTIVE(one) - .type limit,@object -limit: .double 0.29 - ASM_SIZE_DIRECTIVE(limit) - .align ALIGNARG(4) - .type ln2_2,@object -ln2_2: .quad 0xb17217f7d1cf79ac /* 0.3465735902799726547086160 */ - .short 0x3ffd - ASM_SIZE_DIRECTIVE(ln2_2) - -DEFINE_FLT_MIN - -#ifdef PIC -#define MO(op) op##@GOTOFF(%edx) -#else -#define MO(op) op -#endif - - .text -ENTRY(__ieee754_atanhf) - movl 4(%esp), %ecx - - movl %ecx, %eax - andl $0x7fffffff, %eax - cmpl $0x7f800000, %eax - ja 5f - -#ifdef PIC - LOAD_PIC_REG (dx) -#endif - - andl $0x80000000, %ecx // ECX == 0 iff X >= 0 - - fldt MO(ln2_2) // 0.5*ln2 - xorl %ecx, 4(%esp) - flds 4(%esp) // |x| : 0.5*ln2 - fcoml MO(half) // |x| : 0.5*ln2 - fld %st(0) // |x| : |x| : 0.5*ln2 - fnstsw // |x| : |x| : 0.5*ln2 - sahf - jae 2f - fadd %st, %st(1) // |x| : 2*|x| : 0.5*ln2 - fld %st // |x| : |x| : 2*|x| : 0.5*ln2 - fsubrl MO(one) // 1-|x| : |x| : 2*|x| : 0.5*ln2 - fxch // |x| : 1-|x| : 2*|x| : 0.5*ln2 - fmul %st(2) // 2*|x|^2 : 1-|x| : 2*|x| : 0.5*ln2 - fdivp // (2*|x|^2)/(1-|x|) : 2*|x| : 0.5*ln2 - faddp // 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2 - fcoml MO(limit) // 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2 - fnstsw // 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2 - sahf - jae 4f - fyl2xp1 // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|)) - FLT_CHECK_FORCE_UFLOW_NONNEG - jecxz 3f - fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x)) -3: ret - - .align ALIGNARG(4) -4: faddl MO(one) // 1+2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2 - fyl2x // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|)) - jecxz 3f - fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x)) -3: ret - - .align ALIGNARG(4) -2: faddl MO(one) // 1+|x| : |x| : 0.5*ln2 - fxch // |x| : 1+|x| : 0.5*ln2 - fsubrl MO(one) // 1-|x| : 1+|x| : 0.5*ln2 - fdivrp // (1+|x|)/(1-|x|) : 0.5*ln2 - fyl2x // 0.5*ln2*ld((1+|x|)/(1-|x|)) - jecxz 3f - fchs // 0.5*ln2*ld((1+x)/(1-x)) -3: ret - - // x == NaN -5: flds 4(%esp) - ret -END(__ieee754_atanhf) -libm_alias_finite (__ieee754_atanhf, __atanhf) diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps index 5873895497..cb390de754 100644 --- a/sysdeps/i386/fpu/libm-test-ulps +++ b/sysdeps/i386/fpu/libm-test-ulps @@ -126,19 +126,16 @@ ldouble: 3 Function: "atanh_downward": double: 1 -float: 1 float128: 4 ldouble: 4 Function: "atanh_towardzero": double: 1 -float: 1 float128: 2 ldouble: 3 Function: "atanh_upward": double: 1 -float: 1 float128: 4 ldouble: 5 diff --git a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps index 8fa9053fe4..e4f273d557 100644 --- a/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps +++ b/sysdeps/i386/i686/fpu/multiarch/libm-test-ulps @@ -126,19 +126,16 @@ ldouble: 3 Function: "atanh_downward": double: 1 -float: 1 float128: 4 ldouble: 4 Function: "atanh_towardzero": double: 1 -float: 1 float128: 2 ldouble: 3 Function: "atanh_upward": double: 1 -float: 1 float128: 4 ldouble: 5 diff --git a/sysdeps/ieee754/flt-32/e_atanhf.c b/sysdeps/ieee754/flt-32/e_atanhf.c index d36eb15160..5138408324 100644 --- a/sysdeps/ieee754/flt-32/e_atanhf.c +++ b/sysdeps/ieee754/flt-32/e_atanhf.c @@ -1,76 +1,178 @@ -/* Copyright (C) 2011-2024 Free Software Foundation, Inc. - This file is part of the GNU C Library. +/* Correctly-rounded inverse hyperbolic tangent function for binary32 value. - 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. +Copyright (c) 2023-2024 Alexei Sibidanov. - 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. +The original version of this file was copied from the CORE-MATH +project (file src/binary32/acosh/acoshf.c, revision bc385c2). - You should have received a copy of the GNU Lesser General Public - License along with the GNU C Library; if not, see - . */ +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. -/* __ieee754_atanh(x) - Method : - 1.Reduced x to positive by atanh(-x) = -atanh(x) - 2.For x>=0.5 - 1 2x x - atanh(x) = --- * log(1 + -------) = 0.5 * log1p(2 * --------) - 2 1 - x 1 - x +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. +*/ - For x<0.5 - atanh(x) = 0.5*log1p(2x+2x*x/(1-x)) - - Special cases: - atanh(x) is NaN if |x| > 1 with signal; - atanh(NaN) is that NaN with no signal; - atanh(+-1) is +-INF with signal. - - */ - -#include -#include #include -#include -#include -#include +#include #include +#include "math_config.h" -static const float huge = 1e30; +static __attribute__((noinline)) float +as_special (float x) +{ + uint32_t ix = asuint (x); + uint32_t ax = ix << 1; + if (ax == 0x7f000000u) /* +-1 */ + return __math_divzerof (ix >> 31); + if (ax > 0xff000000u) + return x + x; /* nan */ + return __math_invalidf (x); +} float __ieee754_atanhf (float x) { - float xa = fabsf (x); - float t; - if (isless (xa, 0.5f)) + /* Calculate atanh(x) using the difference of two logarithms -- atanh(x) = + (ln(1+x) - ln(1-x))/2 */ + static const double tr[] = + { + 0x1.fc07f02p-1, 0x1.f44659ep-1, 0x1.ecc07b3p-1, 0x1.e573ac9p-1, + 0x1.de5d6e4p-1, 0x1.d77b655p-1, 0x1.d0cb58fp-1, 0x1.ca4b305p-1, + 0x1.c3f8f02p-1, 0x1.bdd2b8ap-1, 0x1.b7d6c3ep-1, 0x1.b20364p-1, + 0x1.ac5701bp-1, 0x1.a6d01a7p-1, 0x1.a16d3f9p-1, 0x1.9c2d14fp-1, + 0x1.970e4f8p-1, 0x1.920fb4ap-1, 0x1.8d3018dp-1, 0x1.886e5f1p-1, + 0x1.83c977bp-1, 0x1.7f405fdp-1, 0x1.7ad2209p-1, 0x1.767dce4p-1, + 0x1.724287fp-1, 0x1.6e1f76bp-1, 0x1.6a13cd1p-1, 0x1.661ec6ap-1, + 0x1.623fa77p-1, 0x1.5e75bb9p-1, 0x1.5ac056bp-1, 0x1.571ed3cp-1, + 0x1.5390949p-1, 0x1.5015015p-1, 0x1.4cab887p-1, 0x1.49539e4p-1, + 0x1.460cbc8p-1, 0x1.42d6626p-1, 0x1.3fb014p-1, 0x1.3c995a4p-1, + 0x1.3991c2cp-1, 0x1.3698df4p-1, 0x1.33ae45bp-1, 0x1.30d1901p-1, + 0x1.2e025cp-1, 0x1.2b404adp-1, 0x1.288b013p-1, 0x1.25e2271p-1, + 0x1.2345679p-1, 0x1.20b470cp-1, 0x1.1e2ef3bp-1, 0x1.1bb4a4p-1, + 0x1.1945381p-1, 0x1.16e0689p-1, 0x1.1485f0ep-1, 0x1.12358e7p-1, + 0x1.0fef011p-1, 0x1.0db20a9p-1, 0x1.0b7e6ecp-1, 0x1.0953f39p-1, + 0x1.073260ap-1, 0x1.05197f8p-1, 0x1.03091b5p-1, 0x1.010101p-1 + }; + static const double tl[] = + { + 0x1.fe02a69106789p-9, 0x1.7b91b1155b11bp-7, 0x1.39e87ba1ebd6p-6, + 0x1.b42dd713971bfp-6, 0x1.16536ee637ae1p-5, 0x1.51b073c96183fp-5, + 0x1.8c345da019b21p-5, 0x1.c5e5492abc743p-5, 0x1.fec912fbbeabbp-5, + 0x1.1b72ad33f67ap-4, 0x1.371fc1f6e8f74p-4, 0x1.526e5e5a1b438p-4, + 0x1.6d60fe601d21dp-4, 0x1.87fa06438c911p-4, 0x1.a23bc223ab563p-4, + 0x1.bc28673a58cd6p-4, 0x1.d5c216b8fbb91p-4, 0x1.ef0adcaec5936p-4, + 0x1.040259530d041p-3, 0x1.1058bf8d24ad5p-3, 0x1.1c898c09d99fbp-3, + 0x1.2895a13e286a3p-3, 0x1.347dd9a447d55p-3, 0x1.404308716a7e4p-3, + 0x1.4be5f963b78a1p-3, 0x1.5767718015a6cp-3, 0x1.62c82f3a5c795p-3, + 0x1.6e08eab13a1e4p-3, 0x1.792a55fe147a2p-3, 0x1.842d1d9928b17p-3, + 0x1.8f11e873a62c7p-3, 0x1.99d958207e08bp-3, 0x1.a484090c1bb0ap-3, + 0x1.af129324b786bp-3, 0x1.b9858970710fbp-3, 0x1.c3dd7a6ddad4dp-3, + 0x1.ce1af0b65f3ebp-3, 0x1.d83e725022f3ep-3, 0x1.e2488197c6c26p-3, + 0x1.ec399d3d68ccp-3, 0x1.f6123fac028acp-3, 0x1.ffd2e07e7f498p-3, + 0x1.04bdf9e3b26d2p-2, 0x1.0986f4fa93521p-2, 0x1.0e4498651cc8cp-2, + 0x1.12f719595efbcp-2, 0x1.179eabb0a99a1p-2, 0x1.1c3b81e933c25p-2, + 0x1.20cdcd0e0ab6ep-2, 0x1.2555bcf50f7cbp-2, 0x1.29d37ff34b08bp-2, + 0x1.2e47437640268p-2, 0x1.32b1338401d71p-2, 0x1.37117b5c147b6p-2, + 0x1.3b6844a13fc23p-2, 0x1.3fb5b857f6f42p-2, 0x1.43f9fe2f7ce67p-2, + 0x1.48353d11488dfp-2, 0x1.4c679b014ee3ap-2, 0x1.50913cc03686bp-2, + 0x1.54b2468259498p-2, 0x1.58cadb57d7989p-2, 0x1.5cdb1dcaa1765p-2, + 0x1.60e32f46788d9p-2 + }; + static const double ln2n[] = + { + 0x1.62e42fedb2a44p-2, 0x1.62e42feeab21ap-1, 0x1.0a2b23f33e789p+0, + 0x1.62e42fef27604p+0, 0x1.bb9d3beb1048p+0, 0x1.0a2b23f37c97ep+1, + 0x1.3687a9f1710bcp+1, 0x1.62e42fef657fap+1, 0x1.8f40b5ed59f38p+1, + 0x1.bb9d3beb4e676p+1, 0x1.e7f9c1e942db4p+1, 0x1.0a2b23f39ba79p+2, + 0x1.205966f295e18p+2, 0x1.3687a9f1901b7p+2, 0x1.4cb5ecf08a556p+2, + 0x1.62e42fef848f5p+2, 0x1.791272ee7ec93p+2, 0x1.8f40b5ed79032p+2, + 0x1.a56ef8ec733d1p+2, 0x1.bb9d3beb6d77p+2, 0x1.d1cb7eea67b0fp+2, + 0x1.e7f9c1e961eaep+2, 0x1.fe2804e85c24dp+2, 0x1.0a2b23f3ab2f6p+3 + }; + static const double b[] = { - if (__glibc_unlikely (xa < 0x1.0p-28f)) + 0x1.fffffffce5a6ap-2, -0x1.0001f81ec0ab8p-2, 0x1.555a0f53d79a5p-3 + }; + static const double s[] = { 1, -1 }; + + uint32_t ux = asuint (x); + uint32_t ax = ux << 1; + if (__glibc_unlikely (ax < 0x7a300000u || ax >= 0x7f000000u)) + { + if (__glibc_unlikely (ax >= 0x7f000000u)) + return as_special (x); + if (__glibc_unlikely (ax < 0x73713744u)) { - math_force_eval (huge + x); - math_check_force_underflow (x); - return x; + if (ax == 0) + return x; /* x = +-0 */ + return fmaf (x, 0x1p-25f, x); /* |x| < 0.000352112(0x1.713744p-12) */ + } + else + { /* |x| < 0x1.3p-5 */ + static const double c[] = + { + 0x1.5555555555527p-2, 0x1.9999999ba4ee8p-3, + 0x1.24922c280990ap-3, 0x1.c8236aae809c6p-4 + }; + double z = x; + double z2 = z * z; + double z4 = z2 * z2; + double r = c[0] + z2 * c[1] + z4 * (c[2] + z2 * c[3]); + return z + (z * z2) * r; } - - t = xa + xa; - t = 0.5f * __log1pf (t + t * xa / (1.0f - xa)); } - else if (__glibc_likely (isless (xa, 1.0f))) - t = 0.5f * __log1pf ((xa + xa) / (1.0f - xa)); - else + double sgn = s[ux >> 31]; + unsigned int e = ax >> 24; + unsigned int md = ((ux << 8) | 1 << 31) >> (126 - e); + unsigned int mn = -md; + int nz = __builtin_clz (mn) + 1; + mn <<= nz; + unsigned int jn = mn >> 26; + unsigned int jd = md >> 26; + double tn = asdouble (((int64_t) mn << 20) | (INT64_C(1023) << 52)); + double td = asdouble (((int64_t) md << 20) | (INT64_C(1023) << 52)); + double zn = tn * tr[jn] - 1; + double zd = td * tr[jd] - 1; + double zn2 = zn * zn; + double zd2 = zd * zd; + double rn = ((tl[jn] - ln2n[nz - 1]) + zn * b[0]) + zn2 * (b[1] + zn * b[2]); + double rd = (tl[jd] + zd * b[0]) + zd2 * (b[1] + zd * b[2]); + double r = sgn * (rd - rn); + float ub = r; + float lb = r + sgn * 0.226e-9; + if (__glibc_unlikely (ub != lb)) { - if (isgreater (xa, 1.0f)) - return (x - x) / (x - x); - - return x / 0.0f; + static const double c[] = + { + 0x1p-1, -0x1.000000000001bp-2, 0x1.55555555555bap-3, + -0x1.fffffff26d72ep-4, 0x1.99999989035p-4, -0x1.555c39cb9ee8p-4, + 0x1.24992d8b014a1p-4 + }; + double zn4 = zn2 * zn2; + double zd4 = zd2 * zd2; + double fn = zn * (((c[0] + zn * c[1]) + zn2 * (c[2] + zn * c[3])) + + zn4 * ((c[4] + zn * c[5]) + zn2 * c[6])); + fn += 0x1.0ca86c3898dp-50 * nz; + fn += tl[jn]; + double en = nz * 0x1.62e42fefa3ap-2; + double fd = zd * (((c[0] + zd * c[1]) + zd2 * (c[2] + zd * c[3])) + + zd4 * ((c[4] + zd * c[5]) + zd2 * c[6])); + fd += tl[jd]; + r = fd - fn + en; + ub = sgn * r; } - - return copysignf (t, x); + return ub; } libm_alias_finite (__ieee754_atanhf, __atanhf) diff --git a/sysdeps/loongarch/lp64/libm-test-ulps b/sysdeps/loongarch/lp64/libm-test-ulps index a34539753c..1e1a289169 100644 --- a/sysdeps/loongarch/lp64/libm-test-ulps +++ b/sysdeps/loongarch/lp64/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/m68k/coldfire/fpu/libm-test-ulps b/sysdeps/m68k/coldfire/fpu/libm-test-ulps index 989015ff9a..998088f295 100644 --- a/sysdeps/m68k/coldfire/fpu/libm-test-ulps +++ b/sysdeps/m68k/coldfire/fpu/libm-test-ulps @@ -4,7 +4,6 @@ Function: "atan2": Function: "atanh": -float: 1 Function: Imaginary part of "cacosh": float: 1 diff --git a/sysdeps/microblaze/libm-test-ulps b/sysdeps/microblaze/libm-test-ulps index 4fa242daca..77018f4f72 100644 --- a/sysdeps/microblaze/libm-test-ulps +++ b/sysdeps/microblaze/libm-test-ulps @@ -17,7 +17,6 @@ Function: "atan2": Function: "atanh": double: 2 -float: 2 Function: "cabs": double: 1 diff --git a/sysdeps/mips/mips32/libm-test-ulps b/sysdeps/mips/mips32/libm-test-ulps index 2f6f7fff23..2191d57515 100644 --- a/sysdeps/mips/mips32/libm-test-ulps +++ b/sysdeps/mips/mips32/libm-test-ulps @@ -74,19 +74,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/mips/mips64/libm-test-ulps b/sysdeps/mips/mips64/libm-test-ulps index 037fb885ed..7d789b90ad 100644 --- a/sysdeps/mips/mips64/libm-test-ulps +++ b/sysdeps/mips/mips64/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/or1k/fpu/libm-test-ulps b/sysdeps/or1k/fpu/libm-test-ulps index 3fee2b50cc..357d0a6946 100644 --- a/sysdeps/or1k/fpu/libm-test-ulps +++ b/sysdeps/or1k/fpu/libm-test-ulps @@ -74,19 +74,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/or1k/nofpu/libm-test-ulps b/sysdeps/or1k/nofpu/libm-test-ulps index 801d2acb80..b3519a006d 100644 --- a/sysdeps/or1k/nofpu/libm-test-ulps +++ b/sysdeps/or1k/nofpu/libm-test-ulps @@ -74,19 +74,15 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_downward": double: 3 -float: 3 Function: "atanh_towardzero": double: 2 -float: 2 Function: "atanh_upward": double: 3 -float: 3 Function: "cabs": double: 1 diff --git a/sysdeps/powerpc/fpu/libm-test-ulps b/sysdeps/powerpc/fpu/libm-test-ulps index 1ef04d301b..00e3c516d0 100644 --- a/sysdeps/powerpc/fpu/libm-test-ulps +++ b/sysdeps/powerpc/fpu/libm-test-ulps @@ -126,25 +126,21 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 float128: 4 ldouble: 2 Function: "atanh_downward": double: 3 -float: 3 float128: 4 ldouble: 3 Function: "atanh_towardzero": double: 2 -float: 2 float128: 2 ldouble: 4 Function: "atanh_upward": double: 3 -float: 3 float128: 4 ldouble: 4 diff --git a/sysdeps/powerpc/nofpu/libm-test-ulps b/sysdeps/powerpc/nofpu/libm-test-ulps index 8b7ed72386..b13e465745 100644 --- a/sysdeps/powerpc/nofpu/libm-test-ulps +++ b/sysdeps/powerpc/nofpu/libm-test-ulps @@ -102,22 +102,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 2 Function: "atanh_downward": double: 3 -float: 3 ldouble: 3 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 4 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/riscv/nofpu/libm-test-ulps b/sysdeps/riscv/nofpu/libm-test-ulps index 19e36742aa..dc78136bc6 100644 --- a/sysdeps/riscv/nofpu/libm-test-ulps +++ b/sysdeps/riscv/nofpu/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/riscv/rvd/libm-test-ulps b/sysdeps/riscv/rvd/libm-test-ulps index ad3242fefd..9477c4a101 100644 --- a/sysdeps/riscv/rvd/libm-test-ulps +++ b/sysdeps/riscv/rvd/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/s390/fpu/libm-test-ulps b/sysdeps/s390/fpu/libm-test-ulps index 668f947473..60c7e17f46 100644 --- a/sysdeps/s390/fpu/libm-test-ulps +++ b/sysdeps/s390/fpu/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/sh/libm-test-ulps b/sysdeps/sh/libm-test-ulps index 15655ad3a8..4b308af5f5 100644 --- a/sysdeps/sh/libm-test-ulps +++ b/sysdeps/sh/libm-test-ulps @@ -35,11 +35,9 @@ double: 1 Function: "atanh": double: 2 -float: 2 Function: "atanh_towardzero": double: 2 -float: 2 Function: "cabs": double: 1 diff --git a/sysdeps/sparc/fpu/libm-test-ulps b/sysdeps/sparc/fpu/libm-test-ulps index 7059a017e8..6ca72eb9de 100644 --- a/sysdeps/sparc/fpu/libm-test-ulps +++ b/sysdeps/sparc/fpu/libm-test-ulps @@ -98,22 +98,18 @@ ldouble: 2 Function: "atanh": double: 2 -float: 2 ldouble: 4 Function: "atanh_downward": double: 3 -float: 3 ldouble: 4 Function: "atanh_towardzero": double: 2 -float: 2 ldouble: 2 Function: "atanh_upward": double: 3 -float: 3 ldouble: 4 Function: "cabs": diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps index 5c169fb9e6..a659ba2f95 100644 --- a/sysdeps/x86_64/fpu/libm-test-ulps +++ b/sysdeps/x86_64/fpu/libm-test-ulps @@ -239,25 +239,21 @@ float: 1 Function: "atanh": double: 2 -float: 2 float128: 4 ldouble: 3 Function: "atanh_downward": double: 3 -float: 3 float128: 4 ldouble: 5 Function: "atanh_towardzero": double: 2 -float: 2 float128: 2 ldouble: 4 Function: "atanh_upward": double: 3 -float: 3 float128: 4 ldouble: 5