From patchwork Mon Dec 6 22:14:17 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Khem Raj X-Patchwork-Id: 690 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from aws-us-west-2-korg-lkml-1.web.codeaurora.org (localhost.localdomain [127.0.0.1]) by smtp.lore.kernel.org (Postfix) with ESMTP id 240A5C433EF for ; Mon, 6 Dec 2021 22:14:24 +0000 (UTC) Received: from mail-pj1-f41.google.com (mail-pj1-f41.google.com [209.85.216.41]) by mx.groups.io with SMTP id smtpd.web12.58888.1638828863400227702 for ; Mon, 06 Dec 2021 14:14:23 -0800 Authentication-Results: mx.groups.io; dkim=pass header.i=@gmail.com header.s=20210112 header.b=DkjVsetY; spf=pass (domain: gmail.com, ip: 209.85.216.41, mailfrom: raj.khem@gmail.com) Received: by mail-pj1-f41.google.com with SMTP id h24so8802867pjq.2 for ; Mon, 06 Dec 2021 14:14:23 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:cc:subject:date:message-id:mime-version :content-transfer-encoding; bh=52GDssYABJfVLBYvKwYeAJH6K7ypE3VWKpZt6kCHGvA=; b=DkjVsetYE2xKZmnuCVFjDaWB0FyQrZBVb6Juy6NWKBXl7KFmI2UK/ZHuRAxPRX6ED0 rtucevcKub1rP5Q1YTRF9y6Y3sjazN/zNpGxFb90NlaagSu5potac/6ZxbMD4ZPE1rgj XSj56KOnNcJyzczEnA4NFmjGnpGAmAb3aq1XstKLHZwlLCrdmcfFy3Lyv0XO55ZQZTND W+1qKJH8lD3b+p2YZyE9cU9SKcNNAsZl4Z1jsAhqRrgdBMRAS3x3+cOWlvGU5lkukCJ6 RjcbVB2EzE6QSC5F4JdRHZ+tOmqiCBcIq4YwaZAK1/2ywLzNvM7TbS1sRZLeAFB07OTJ 2YtA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:mime-version :content-transfer-encoding; bh=52GDssYABJfVLBYvKwYeAJH6K7ypE3VWKpZt6kCHGvA=; b=MpM2xhNHlFWq2F7qDoAhcItUpsPvG5GpWFyEMIvyZst0CbxYGflx82nilHPVFNDEbl chkot4CjMmC4a46Qy1oVjfBUqLZouJqzODswqIaNV4t3l+1AStSb943ejoyTC4Zi99WL QEsnhGQaNWlqmmfV/LHk4DxftH/9Z5ZsM1pEl0nSLTqV4aYhCiah/phmdu7w0t33JH2/ CbgHDWeJNnAJRN3QSIxqPpJjGG+TwmUE6wcjHKjZe/kFVtBOsWXSFIKuLnPcbIUYB5ZP voMMK4FfgudcrKqx9w7DG/O1v/8hOQinIihFGOdymKF73hOEwxdtFOxpMkOD/+zYJRfn kHJQ== X-Gm-Message-State: AOAM532/klIDh06aQ1giSnj50ri3DTNi7ALLCQzRzWgkKBbGxvcI/KQh FTdPG374rMvQqpv5dvDW3xy9JatemG3vWQ== X-Google-Smtp-Source: ABdhPJzI6ZIS0fB3ERDA2GxjFSIS17pRXbnLEWAWO+klaLG9obWpHPMB6ByHZrYGWnqdjqPFQ2/Jrw== X-Received: by 2002:a17:90b:3b8d:: with SMTP id pc13mr1636919pjb.112.1638828860306; Mon, 06 Dec 2021 14:14:20 -0800 (PST) Received: from apollo.hsd1.ca.comcast.net ([2601:646:9200:a0f0::ce68]) by smtp.gmail.com with ESMTPSA id il7sm318687pjb.54.2021.12.06.14.14.19 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 06 Dec 2021 14:14:19 -0800 (PST) From: Khem Raj To: openembedded-core@lists.openembedded.org Cc: Khem Raj Subject: [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Date: Mon, 6 Dec 2021 14:14:17 -0800 Message-Id: <20211206221418.2281162-1-raj.khem@gmail.com> X-Mailer: git-send-email 2.34.1 MIME-Version: 1.0 List-Id: X-Webhook-Received: from li982-79.members.linode.com [45.33.32.79] by aws-us-west-2-korg-lkml-1.web.codeaurora.org with HTTPS for ; Mon, 06 Dec 2021 22:14:24 -0000 X-Groupsio-URL: https://lists.openembedded.org/g/openembedded-core/message/159271 Signed-off-by: Khem Raj --- ...5500-e6500-603e-fsqrt-implementation.patch | 1581 ----------------- ...undefined-reference-to-__sqrt_finite.patch | 205 --- ...-are-now-inline-functions-and-call-o.patch | 384 ---- ...-are-now-inline-functions-and-call-o.patch | 58 - meta/recipes-core/glibc/glibc_2.34.bb | 4 - 5 files changed, 2232 deletions(-) delete mode 100644 meta/recipes-core/glibc/glibc/0009-fsl-e500-e5500-e6500-603e-fsqrt-implementation.patch delete mode 100644 meta/recipes-core/glibc/glibc/0010-ppc-sqrt-Fix-undefined-reference-to-__sqrt_finite.patch delete mode 100644 meta/recipes-core/glibc/glibc/0011-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch delete mode 100644 meta/recipes-core/glibc/glibc/0014-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch diff --git a/meta/recipes-core/glibc/glibc/0009-fsl-e500-e5500-e6500-603e-fsqrt-implementation.patch b/meta/recipes-core/glibc/glibc/0009-fsl-e500-e5500-e6500-603e-fsqrt-implementation.patch deleted file mode 100644 index 01de227a023..00000000000 --- a/meta/recipes-core/glibc/glibc/0009-fsl-e500-e5500-e6500-603e-fsqrt-implementation.patch +++ /dev/null @@ -1,1581 +0,0 @@ -From 74923ca4b1ae0ed5a2478e7d265b37534f6815d7 Mon Sep 17 00:00:00 2001 -From: Khem Raj -Date: Wed, 18 Mar 2015 00:01:50 +0000 -Subject: [PATCH] fsl e500/e5500/e6500/603e fsqrt implementation - -Upstream-Status: Pending -Signed-off-by: Edmar Wienskoski -Signed-off-by: Khem Raj ---- - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c | 134 ++++++++++++++++++ - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c | 101 +++++++++++++ - sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c | 134 ++++++++++++++++++ - .../powerpc/powerpc32/e500mc/fpu/e_sqrtf.c | 101 +++++++++++++ - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c | 134 ++++++++++++++++++ - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c | 101 +++++++++++++ - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c | 134 ++++++++++++++++++ - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c | 101 +++++++++++++ - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c | 134 ++++++++++++++++++ - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c | 101 +++++++++++++ - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c | 134 ++++++++++++++++++ - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c | 101 +++++++++++++ - .../linux/powerpc/powerpc32/603e/fpu/Implies | 1 + - .../powerpc/powerpc32/e300c3/fpu/Implies | 2 + - .../powerpc/powerpc32/e500mc/fpu/Implies | 1 + - .../linux/powerpc/powerpc32/e5500/fpu/Implies | 1 + - .../linux/powerpc/powerpc32/e6500/fpu/Implies | 1 + - .../linux/powerpc/powerpc64/e5500/fpu/Implies | 1 + - .../linux/powerpc/powerpc64/e6500/fpu/Implies | 1 + - 19 files changed, 1418 insertions(+) - create mode 100644 sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c - create mode 100644 sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c - create mode 100644 sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c - create mode 100644 sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c - create mode 100644 sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c - create mode 100644 sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c - create mode 100644 sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc32/603e/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc32/e300c3/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc32/e500mc/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc32/e5500/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc32/e6500/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc64/e5500/fpu/Implies - create mode 100644 sysdeps/unix/sysv/linux/powerpc/powerpc64/e6500/fpu/Implies - -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -new file mode 100644 -index 0000000000..71e516d1c8 ---- /dev/null -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -@@ -0,0 +1,134 @@ -+/* Double-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float two108 = 3.245185536584267269e+32; -+static const float twom54 = 5.551115123125782702e-17; -+static const float half = 0.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the actual square root and half of its reciprocal -+ simultaneously. */ -+ -+#ifdef __STDC__ -+double -+__ieee754_sqrt (double b) -+#else -+double -+__ieee754_sqrt (b) -+ double b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+ double y, g, h, d, r; -+ ieee_double_shape_type u; -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ u.value = b; -+ -+ relax_fenv_state (); -+ -+ __asm__ ("frsqrte %[estimate], %[x]\n" -+ : [estimate] "=f" (y) : [x] "f" (b)); -+ -+ /* Following Muller et al, page 168, equation 5.20. -+ -+ h goes to 1/(2*sqrt(b)) -+ g goes to sqrt(b). -+ -+ We need three iterations to get within 1ulp. */ -+ -+ /* Indicate that these can be performed prior to the branch. GCC -+ insists on sinking them below the branch, however; it seems like -+ they'd be better before the branch so that we can cover any latency -+ from storing the argument and loading its high word. Oh well. */ -+ -+ g = b * y; -+ h = 0.5 * y; -+ -+ /* Handle small numbers by scaling. */ -+ if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -+ return __ieee754_sqrt (b * two108) * twom54; -+ -+#define FMADD(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmadd %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ r = FNMSUB (g, h, half); -+ g = FMADD (g, r, g); -+ h = FMADD (h, r, h); -+ -+ /* g is now +/- 1ulp, or exactly equal to, the square root of b. */ -+ -+ /* Final refinement. */ -+ d = FNMSUB (g, g, b); -+ -+ fesetenv_register (fe); -+ return FMADD (d, h, g); -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_wash (b); -+} -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -new file mode 100644 -index 0000000000..26fa067abf ---- /dev/null -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -@@ -0,0 +1,101 @@ -+/* Single-precision floating point square root. -+ Copyright (C) 2010 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, write to the Free -+ Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA -+ 02111-1307 USA. */ -+ -+#include -+#include -+#include -+#include -+ -+#include -+#include -+ -+static const ieee_float_shape_type a_nan = {.word = 0x7fc00000 }; -+static const ieee_float_shape_type a_inf = {.word = 0x7f800000 }; -+static const float threehalf = 1.5; -+ -+/* The method is based on the descriptions in: -+ -+ _The Handbook of Floating-Pointer Arithmetic_ by Muller et al., chapter 5; -+ _IA-64 and Elementary Functions: Speed and Precision_ by Markstein, chapter 9 -+ -+ We find the reciprocal square root and use that to compute the actual -+ square root. */ -+ -+#ifdef __STDC__ -+float -+__ieee754_sqrtf (float b) -+#else -+float -+__ieee754_sqrtf (b) -+ float b; -+#endif -+{ -+ if (__builtin_expect (b > 0, 1)) -+ { -+#define FMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+#define FNMSUB(a_, c_, b_) \ -+ ({ double __r; \ -+ __asm__ ("fnmsub %[r], %[a], %[c], %[b]\n" \ -+ : [r] "=f" (__r) : [a] "f" (a_), [c] "f" (c_), [b] "f" (b_)); \ -+ __r;}) -+ -+ if (__builtin_expect (b != a_inf.value, 1)) -+ { -+ double y, x; -+ fenv_t fe; -+ -+ fe = fegetenv_register (); -+ -+ relax_fenv_state (); -+ -+ /* Compute y = 1.5 * b - b. Uses fewer constants than y = 0.5 * b. */ -+ y = FMSUB (threehalf, b, b); -+ -+ /* Initial estimate. */ -+ __asm__ ("frsqrte %[x], %[b]\n" : [x] "=f" (x) : [b] "f" (b)); -+ -+ /* Iterate. x_{n+1} = x_n * (1.5 - y * (x_n * x_n)). */ -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ x = x * FNMSUB (y, x * x, threehalf); -+ -+ /* All done. */ -+ fesetenv_register (fe); -+ return x * b; -+ } -+ } -+ else if (b < 0) -+ { -+ /* For some reason, some PowerPC32 processors don't implement -+ FE_INVALID_SQRT. */ -+#ifdef FE_INVALID_SQRT -+ feraiseexcept (FE_INVALID_SQRT); -+ -+ fenv_union_t u = { .fenv = fegetenv_register () }; -+ if ((u.l & FE_INVALID) == 0) -+#endif -+ feraiseexcept (FE_INVALID); -+ b = a_nan.value; -+ } -+ return f_washf (b); -+} -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/603e/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc32/603e/fpu/Implies -new file mode 100644 -index 0000000000..b103b4dea5 ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/603e/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc32/603e/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/e300c3/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e300c3/fpu/Implies -new file mode 100644 -index 0000000000..64db17fada ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e300c3/fpu/Implies -@@ -0,0 +1,2 @@ -+# e300c3 is a variant of 603e so use the same optimizations for sqrt -+powerpc/powerpc32/603e/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/e500mc/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e500mc/fpu/Implies -new file mode 100644 -index 0000000000..7eac5fcf02 ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e500mc/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc32/e500mc/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/e5500/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e5500/fpu/Implies -new file mode 100644 -index 0000000000..264b2a7700 ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e5500/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc32/e5500/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc32/e6500/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e6500/fpu/Implies -new file mode 100644 -index 0000000000..a25934467b ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc32/e6500/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc32/e6500/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc64/e5500/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc64/e5500/fpu/Implies -new file mode 100644 -index 0000000000..a7bc854be8 ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc64/e5500/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc64/e5500/fpu -diff --git a/sysdeps/unix/sysv/linux/powerpc/powerpc64/e6500/fpu/Implies b/sysdeps/unix/sysv/linux/powerpc/powerpc64/e6500/fpu/Implies -new file mode 100644 -index 0000000000..04ff8cc181 ---- /dev/null -+++ b/sysdeps/unix/sysv/linux/powerpc/powerpc64/e6500/fpu/Implies -@@ -0,0 +1 @@ -+powerpc/powerpc64/e6500/fpu diff --git a/meta/recipes-core/glibc/glibc/0010-ppc-sqrt-Fix-undefined-reference-to-__sqrt_finite.patch b/meta/recipes-core/glibc/glibc/0010-ppc-sqrt-Fix-undefined-reference-to-__sqrt_finite.patch deleted file mode 100644 index 5c1130cf10f..00000000000 --- a/meta/recipes-core/glibc/glibc/0010-ppc-sqrt-Fix-undefined-reference-to-__sqrt_finite.patch +++ /dev/null @@ -1,205 +0,0 @@ -From 5da3da7f2d276c2a6ae1b04419b28e96953803ec Mon Sep 17 00:00:00 2001 -From: Khem Raj -Date: Wed, 18 Mar 2015 00:15:07 +0000 -Subject: [PATCH] ppc/sqrt: Fix undefined reference to `__sqrt_finite' - -on ppc fixes the errors like below -| ./.libs/libpulsecore-1.1.so: undefined reference to `__sqrt_finite' -| collect2: ld returned 1 exit status - -Upstream-Status: Pending - -ChangeLog - -2012-01-06 Khem Raj - - * sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c: Add __*_finite alias. - Remove cruft. - * sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c: Ditto. - * sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c: Ditto. - * sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c: Ditto. - -Signed-off-by: Khem Raj ---- - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c | 7 +------ - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c | 7 +------ - sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c | 1 + - sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c | 1 + - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c | 1 + - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c | 1 + - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c | 1 + - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c | 1 + - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c | 7 +------ - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c | 7 +------ - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c | 1 + - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c | 1 + - 12 files changed, 12 insertions(+), 24 deletions(-) - -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -index 71e516d1c8..1795fd6c3e 100644 ---- a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -@@ -39,14 +39,8 @@ static const float half = 0.5; - We find the actual square root and half of its reciprocal - simultaneously. */ - --#ifdef __STDC__ - double - __ieee754_sqrt (double b) --#else --double --__ieee754_sqrt (b) -- double b; --#endif - { - if (__builtin_expect (b > 0, 1)) - { -@@ -132,3 +126,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -index 26fa067abf..a917f313ab 100644 ---- a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -@@ -37,14 +37,8 @@ static const float threehalf = 1.5; - We find the reciprocal square root and use that to compute the actual - square root. */ - --#ifdef __STDC__ - float - __ieee754_sqrtf (float b) --#else --float --__ieee754_sqrtf (b) -- float b; --#endif - { - if (__builtin_expect (b > 0, 1)) - { -@@ -99,3 +93,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -index 71e516d1c8..fc4a74990e 100644 ---- a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -@@ -132,3 +132,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -index 26fa067abf..9d175122a8 100644 ---- a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -@@ -99,3 +99,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -index 71e516d1c8..fc4a74990e 100644 ---- a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -@@ -132,3 +132,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -index 26fa067abf..9d175122a8 100644 ---- a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -@@ -99,3 +99,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -index 71e516d1c8..fc4a74990e 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -@@ -132,3 +132,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -index 26fa067abf..9d175122a8 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -@@ -99,3 +99,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -index 71e516d1c8..1795fd6c3e 100644 ---- a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -@@ -39,14 +39,8 @@ static const float half = 0.5; - We find the actual square root and half of its reciprocal - simultaneously. */ - --#ifdef __STDC__ - double - __ieee754_sqrt (double b) --#else --double --__ieee754_sqrt (b) -- double b; --#endif - { - if (__builtin_expect (b > 0, 1)) - { -@@ -132,3 +126,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -index 26fa067abf..a917f313ab 100644 ---- a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -@@ -37,14 +37,8 @@ static const float threehalf = 1.5; - We find the reciprocal square root and use that to compute the actual - square root. */ - --#ifdef __STDC__ - float - __ieee754_sqrtf (float b) --#else --float --__ieee754_sqrtf (b) -- float b; --#endif - { - if (__builtin_expect (b > 0, 1)) - { -@@ -99,3 +93,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -index 71e516d1c8..fc4a74990e 100644 ---- a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -@@ -132,3 +132,4 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -index 26fa067abf..9d175122a8 100644 ---- a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -@@ -99,3 +99,4 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+strong_alias (__ieee754_sqrtf, __sqrtf_finite) diff --git a/meta/recipes-core/glibc/glibc/0011-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch b/meta/recipes-core/glibc/glibc/0011-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch deleted file mode 100644 index b72e79057b0..00000000000 --- a/meta/recipes-core/glibc/glibc/0011-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch +++ /dev/null @@ -1,384 +0,0 @@ -From 77f1c90d67a2f8852184fb8fd95cb0ed63065dc7 Mon Sep 17 00:00:00 2001 -From: Khem Raj -Date: Wed, 18 Mar 2015 00:16:38 +0000 -Subject: [PATCH] __ieee754_sqrt{,f} are now inline functions and call out - __slow versions - -Upstream-Status: Pending - -Signed-off-by: Khem Raj ---- - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c | 12 ++++++++++-- - sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c | 8 +++++++- - sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c | 14 +++++++++++--- - sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c | 12 ++++++++++-- - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c | 14 +++++++++++--- - sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c | 12 ++++++++++-- - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c | 8 ++++++++ - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c | 8 ++++++++ - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c | 12 ++++++++++-- - sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c | 9 ++++++++- - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c | 14 +++++++++++--- - sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c | 12 ++++++++++-- - 12 files changed, 114 insertions(+), 21 deletions(-) - -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -index 1795fd6c3e..daa83f3fe8 100644 ---- a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrt.c -@@ -40,7 +40,7 @@ static const float half = 0.5; - simultaneously. */ - - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - { - if (__builtin_expect (b > 0, 1)) - { -@@ -77,7 +77,7 @@ __ieee754_sqrt (double b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -@@ -126,4 +126,12 @@ __ieee754_sqrt (double b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -index a917f313ab..b812cf1705 100644 ---- a/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/603e/fpu/e_sqrtf.c -@@ -38,7 +38,7 @@ static const float threehalf = 1.5; - square root. */ - - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - { - if (__builtin_expect (b > 0, 1)) - { -@@ -93,4 +93,10 @@ __ieee754_sqrtf (float b) - } - return f_washf (b); - } -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} - strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -index fc4a74990e..7038a70b47 100644 ---- a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrt.c -@@ -41,10 +41,10 @@ static const float half = 0.5; - - #ifdef __STDC__ - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - #else - double --__ieee754_sqrt (b) -+__slow_ieee754_sqrt (b) - double b; - #endif - { -@@ -83,7 +83,7 @@ __ieee754_sqrt (b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -@@ -132,4 +132,12 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -index 9d175122a8..10de1f0cc3 100644 ---- a/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e500mc/fpu/e_sqrtf.c -@@ -39,10 +39,10 @@ static const float threehalf = 1.5; - - #ifdef __STDC__ - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - #else - float --__ieee754_sqrtf (b) -+__slow_ieee754_sqrtf (b) - float b; - #endif - { -@@ -99,4 +99,12 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+ -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} -+ - strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -index fc4a74990e..7038a70b47 100644 ---- a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrt.c -@@ -41,10 +41,10 @@ static const float half = 0.5; - - #ifdef __STDC__ - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - #else - double --__ieee754_sqrt (b) -+__slow_ieee754_sqrt (b) - double b; - #endif - { -@@ -83,7 +83,7 @@ __ieee754_sqrt (b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -@@ -132,4 +132,12 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -index 9d175122a8..10de1f0cc3 100644 ---- a/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e5500/fpu/e_sqrtf.c -@@ -39,10 +39,10 @@ static const float threehalf = 1.5; - - #ifdef __STDC__ - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - #else - float --__ieee754_sqrtf (b) -+__slow_ieee754_sqrtf (b) - float b; - #endif - { -@@ -99,4 +99,12 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+ -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} -+ - strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -index fc4a74990e..1c34244bd8 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -@@ -132,4 +132,12 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -index 9d175122a8..812653558f 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -@@ -99,4 +99,12 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+ -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} -+ - strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -index 1795fd6c3e..13a81973e3 100644 ---- a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrt.c -@@ -40,7 +40,7 @@ static const float half = 0.5; - simultaneously. */ - - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - { - if (__builtin_expect (b > 0, 1)) - { -@@ -77,7 +77,7 @@ __ieee754_sqrt (double b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -@@ -126,4 +126,12 @@ __ieee754_sqrt (double b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -index a917f313ab..fae2d81210 100644 ---- a/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc64/e5500/fpu/e_sqrtf.c -@@ -38,7 +38,7 @@ static const float threehalf = 1.5; - square root. */ - - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - { - if (__builtin_expect (b > 0, 1)) - { -@@ -93,4 +93,11 @@ __ieee754_sqrtf (float b) - } - return f_washf (b); - } -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} -+ - strong_alias (__ieee754_sqrtf, __sqrtf_finite) -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -index fc4a74990e..7038a70b47 100644 ---- a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrt.c -@@ -41,10 +41,10 @@ static const float half = 0.5; - - #ifdef __STDC__ - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - #else - double --__ieee754_sqrt (b) -+__slow_ieee754_sqrt (b) - double b; - #endif - { -@@ -83,7 +83,7 @@ __ieee754_sqrt (b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -@@ -132,4 +132,12 @@ __ieee754_sqrt (b) - } - return f_wash (b); - } -+ -+#undef __ieee754_sqrt -+double -+__ieee754_sqrt (double x) -+{ -+ return __slow_ieee754_sqrt (x); -+} -+ - strong_alias (__ieee754_sqrt, __sqrt_finite) -diff --git a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -index 9d175122a8..10de1f0cc3 100644 ---- a/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc64/e6500/fpu/e_sqrtf.c -@@ -39,10 +39,10 @@ static const float threehalf = 1.5; - - #ifdef __STDC__ - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - #else - float --__ieee754_sqrtf (b) -+__slow_ieee754_sqrtf (b) - float b; - #endif - { -@@ -99,4 +99,12 @@ __ieee754_sqrtf (b) - } - return f_washf (b); - } -+ -+#undef __ieee754_sqrtf -+float -+__ieee754_sqrtf (float x) -+{ -+ return __slow_ieee754_sqrtf (x); -+} -+ - strong_alias (__ieee754_sqrtf, __sqrtf_finite) diff --git a/meta/recipes-core/glibc/glibc/0014-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch b/meta/recipes-core/glibc/glibc/0014-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch deleted file mode 100644 index 088b810530d..00000000000 --- a/meta/recipes-core/glibc/glibc/0014-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch +++ /dev/null @@ -1,58 +0,0 @@ -From 133870f12ba36686dd8df1311fac32a4c5b28579 Mon Sep 17 00:00:00 2001 -From: Khem Raj -Date: Wed, 18 Mar 2015 00:24:46 +0000 -Subject: [PATCH] __ieee754_sqrt{,f} are now inline functions and call out - __slow versions - -Upstream-Status: Pending - -Signed-off-by: chunrong guo -Signed-off-by: Khem Raj ---- - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c | 6 +++--- - sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c | 4 ++-- - 2 files changed, 5 insertions(+), 5 deletions(-) - -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -index 1c34244bd8..7038a70b47 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrt.c -@@ -41,10 +41,10 @@ static const float half = 0.5; - - #ifdef __STDC__ - double --__ieee754_sqrt (double b) -+__slow_ieee754_sqrt (double b) - #else - double --__ieee754_sqrt (b) -+__slow_ieee754_sqrt (b) - double b; - #endif - { -@@ -83,7 +83,7 @@ __ieee754_sqrt (b) - - /* Handle small numbers by scaling. */ - if (__builtin_expect ((u.parts.msw & 0x7ff00000) <= 0x02000000, 0)) -- return __ieee754_sqrt (b * two108) * twom54; -+ return __slow_ieee754_sqrt (b * two108) * twom54; - - #define FMADD(a_, c_, b_) \ - ({ double __r; \ -diff --git a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -index 812653558f..10de1f0cc3 100644 ---- a/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -+++ b/sysdeps/powerpc/powerpc32/e6500/fpu/e_sqrtf.c -@@ -39,10 +39,10 @@ static const float threehalf = 1.5; - - #ifdef __STDC__ - float --__ieee754_sqrtf (float b) -+__slow_ieee754_sqrtf (float b) - #else - float --__ieee754_sqrtf (b) -+__slow_ieee754_sqrtf (b) - float b; - #endif - { diff --git a/meta/recipes-core/glibc/glibc_2.34.bb b/meta/recipes-core/glibc/glibc_2.34.bb index 1826eba1756..c9322124334 100644 --- a/meta/recipes-core/glibc/glibc_2.34.bb +++ b/meta/recipes-core/glibc/glibc_2.34.bb @@ -30,11 +30,7 @@ NATIVESDKFIXES:class-nativesdk = "\ SRC_URI = "${GLIBC_GIT_URI};branch=${SRCBRANCH};name=glibc \ ${NATIVESDKFIXES} \ - file://0009-fsl-e500-e5500-e6500-603e-fsqrt-implementation.patch \ - file://0010-ppc-sqrt-Fix-undefined-reference-to-__sqrt_finite.patch \ - file://0011-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch \ file://0013-eglibc-run-libm-err-tab.pl-with-specific-dirs-in-S.patch \ - file://0014-__ieee754_sqrt-f-are-now-inline-functions-and-call-o.patch \ file://0015-sysdeps-gnu-configure.ac-handle-correctly-libc_cv_ro.patch \ file://0016-yes-within-the-path-sets-wrong-config-variables.patch \ file://0017-timezone-re-written-tzselect-as-posix-sh.patch \ From patchwork Mon Dec 6 22:14:18 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Khem Raj X-Patchwork-Id: 689 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from aws-us-west-2-korg-lkml-1.web.codeaurora.org (localhost.localdomain [127.0.0.1]) by smtp.lore.kernel.org (Postfix) with ESMTP id 2EA97C433EF for ; Mon, 6 Dec 2021 22:14:22 +0000 (UTC) Received: from mail-pl1-f181.google.com (mail-pl1-f181.google.com [209.85.214.181]) by mx.groups.io with SMTP id smtpd.web09.58627.1638828861746336542 for ; Mon, 06 Dec 2021 14:14:21 -0800 Authentication-Results: mx.groups.io; dkim=pass header.i=@gmail.com header.s=20210112 header.b=ot8dyk1Q; spf=pass (domain: gmail.com, ip: 209.85.214.181, mailfrom: raj.khem@gmail.com) Received: by mail-pl1-f181.google.com with SMTP id m24so7990462pls.10 for ; Mon, 06 Dec 2021 14:14:21 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=YRuRodE2RmcYPUsiqo4CL/LRzqqJvdAdG3Yff/9vpOQ=; b=ot8dyk1Q/J7Qg0CjzTOWSTLeGNg6D4zQcCpv45A9e6RhFBqEf8nIfIdx5n0F3rxdMM +iPBy0JkVe1jvyArMUi9xl1Zx50AhNZc2wVYoU8gQsdUOWdRI3UbAA0aIGC5kdAwGMhf 4BnwiTy9efUuTt7ZINkz8WhQ/u5dLopxKnoUYeGExqI7DvLtMTn3mi66DMsp7Tv3LC1r gHLmoom6DbaeuedlzTf9GssAId/7vFWz82T34KZai9zzgwGn6KBwzH9g/vDl3TRasnG+ uauu6s4S0jyrUW5UMEVQy1EYA8atp+Iltzv3NSHVZv3mYXMF/ZnjLl01gss+fMoOXPZb DGfg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=YRuRodE2RmcYPUsiqo4CL/LRzqqJvdAdG3Yff/9vpOQ=; b=V53l6AQJQUPlKFJ1hTzCoR0kDsolVjq8h7LJg/lny45meHhRr7S8vAbjYqUQWuYBaF HbkOKEEFG99aS/mogXMiuYwuk8NqSHM4Zs2Zm6jOUPi+ZuI1xBvWFY/t0oni4mReZc7C gbHx7pgBlQfTJKRizjubbZKOhtv7/7jNJhZJcpv1kP1airut7nD1AnrnF68iDJNhjR8m jmxekaL2ybrdUAHmr/UE7i7LKFLfdH0B0oJVyLlOnQhllIx5s55GKnSJrpk7aq3ETM3+ bzhkLW6bZjeAJgjnzAH3fe1gLAVKA9tmMwLGqOMO0PjDET95YqXjXS91TbJAZcKKPgCd 4ScA== X-Gm-Message-State: AOAM531EMAkZOt5vzFYq3eJpTRlEajk21sZ21zG3yPKBFI0fqDKGV2bv tETkKN8ifwFFkaXF/fzIcYUOjSebqlY/Qw== X-Google-Smtp-Source: ABdhPJw3FboGawLEu9h6OP602GCpGS6+2A04eNHzzLxiI8biexgMSSboPCWspgyN+ahdT/HIBK2ffQ== X-Received: by 2002:a17:902:728e:b0:143:a388:868b with SMTP id d14-20020a170902728e00b00143a388868bmr46989576pll.33.1638828861004; Mon, 06 Dec 2021 14:14:21 -0800 (PST) Received: from apollo.hsd1.ca.comcast.net ([2601:646:9200:a0f0::ce68]) by smtp.gmail.com with ESMTPSA id il7sm318687pjb.54.2021.12.06.14.14.20 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 06 Dec 2021 14:14:20 -0800 (PST) From: Khem Raj To: openembedded-core@lists.openembedded.org Cc: Khem Raj Subject: [PATCH 2/2] ffmpeg: Fix build on ppc32 Date: Mon, 6 Dec 2021 14:14:18 -0800 Message-Id: <20211206221418.2281162-2-raj.khem@gmail.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20211206221418.2281162-1-raj.khem@gmail.com> References: <20211206221418.2281162-1-raj.khem@gmail.com> MIME-Version: 1.0 List-Id: X-Webhook-Received: from li982-79.members.linode.com [45.33.32.79] by aws-us-west-2-korg-lkml-1.web.codeaurora.org with HTTPS for ; Mon, 06 Dec 2021 22:14:22 -0000 X-Groupsio-URL: https://lists.openembedded.org/g/openembedded-core/message/159270 Fixes | powerpc-yoe-linux/11.2.0/ld: libavformat/libavformat.so: undefined reference to `__atomic_fetch_sub_8' Signed-off-by: Khem Raj --- meta/recipes-multimedia/ffmpeg/ffmpeg_4.4.1.bb | 1 + 1 file changed, 1 insertion(+) diff --git a/meta/recipes-multimedia/ffmpeg/ffmpeg_4.4.1.bb b/meta/recipes-multimedia/ffmpeg/ffmpeg_4.4.1.bb index 685214a751b..3ba07c31d6b 100644 --- a/meta/recipes-multimedia/ffmpeg/ffmpeg_4.4.1.bb +++ b/meta/recipes-multimedia/ffmpeg/ffmpeg_4.4.1.bb @@ -131,6 +131,7 @@ EXTRA_OECONF:append:mips = " --extra-libs=-latomic --disable-mips32r5 --disable- --disable-loongson2 --disable-loongson3 --disable-mmi --disable-msa --disable-msa2" EXTRA_OECONF:append:riscv32 = " --extra-libs=-latomic" EXTRA_OECONF:append:armv5 = " --extra-libs=-latomic" +EXTRA_OECONF:append:powerpc = " --extra-libs=-latomic" # gold crashes on x86, another solution is to --disable-asm but thats more hacky # ld.gold: internal error in relocate_section, at ../../gold/i386.cc:3684