* [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc
@ 2021-12-06 22:14 Khem Raj
2021-12-06 22:14 ` [PATCH 2/2] ffmpeg: Fix build on ppc32 Khem Raj
2021-12-06 22:51 ` [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Richard Purdie
0 siblings, 2 replies; 6+ messages in thread
From: Khem Raj @ 2021-12-06 22:14 UTC (permalink / raw)
To: openembedded-core; +Cc: Khem Raj
Signed-off-by: Khem Raj <raj.khem@gmail.com>
---
...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 <raj.khem@gmail.com>
-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 <edmar@freescale.com>
-Signed-off-by: Khem Raj <raj.khem@gmail.com>
----
- 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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <math.h>
-+#include <math_private.h>
-+#include <fenv_libc.h>
-+#include <inttypes.h>
-+
-+#include <sysdep.h>
-+#include <ldsodefs.h>
-+
-+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 <raj.khem@gmail.com>
-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 <raj.khem@gmail.com>
-
- * 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 <raj.khem@gmail.com>
----
- 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 <raj.khem@gmail.com>
-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 <raj.khem@gmail.com>
----
- 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 <raj.khem@gmail.com>
-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 <B40290@freescale.com>
-Signed-off-by: Khem Raj <raj.khem@gmail.com>
----
- 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 \
--
2.34.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 2/2] ffmpeg: Fix build on ppc32
2021-12-06 22:14 [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Khem Raj
@ 2021-12-06 22:14 ` Khem Raj
2021-12-06 22:51 ` [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Richard Purdie
1 sibling, 0 replies; 6+ messages in thread
From: Khem Raj @ 2021-12-06 22:14 UTC (permalink / raw)
To: openembedded-core; +Cc: Khem Raj
Fixes
| powerpc-yoe-linux/11.2.0/ld: libavformat/libavformat.so: undefined reference to `__atomic_fetch_sub_8'
Signed-off-by: Khem Raj <raj.khem@gmail.com>
---
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
--
2.34.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc
2021-12-06 22:14 [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Khem Raj
2021-12-06 22:14 ` [PATCH 2/2] ffmpeg: Fix build on ppc32 Khem Raj
@ 2021-12-06 22:51 ` Richard Purdie
2021-12-06 23:28 ` Khem Raj
1 sibling, 1 reply; 6+ messages in thread
From: Richard Purdie @ 2021-12-06 22:51 UTC (permalink / raw)
To: Khem Raj, openembedded-core
I did test such a patch today as well since we discussed it. I'd propose this
commit message since we need to say why:
"""
glibc: Drop ppc sqrt optimisations
OpenEmbedded isn't an upstream or a patch repository. These are optimisations
which for reasons unknown were never merged into upstream glibc. We have no
active ppc maintainers and these patches are at an evolutionary dead end.
Since they are optimisations, there may be small performance regressions by
removing them but OE can't carry patches like this indefinitely and these
appear to have no future.
"""
Cheers,
Richard
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc
2021-12-06 22:51 ` [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Richard Purdie
@ 2021-12-06 23:28 ` Khem Raj
2021-12-06 23:32 ` Richard Purdie
0 siblings, 1 reply; 6+ messages in thread
From: Khem Raj @ 2021-12-06 23:28 UTC (permalink / raw)
To: Richard Purdie; +Cc: openembedded-core
[-- Attachment #1: Type: text/plain, Size: 847 bytes --]
On Mon, Dec 6, 2021 at 2:51 PM Richard Purdie <
richard.purdie@linuxfoundation.org> wrote:
> I did test such a patch today as well since we discussed it. I'd propose
> this
> commit message since we need to say why:
>
> """
>
> glibc: Drop ppc sqrt optimisations
>
> OpenEmbedded isn't an upstream or a patch repository. These are
> optimisations
> which for reasons unknown were never merged into upstream glibc. We have no
> active ppc maintainers and these patches are at an evolutionary dead end.
> Since they are optimisations, there may be small performance regressions by
> removing them but OE can't carry patches like this indefinitely and these
> appear to have no future.
> """
This sounds good to me. I guess will add it to commit or do you want me to
send v2. This passed my local testing on qemuppc
>
>
> Cheers,
>
> Richard
>
>
[-- Attachment #2: Type: text/html, Size: 1397 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc
2021-12-06 23:28 ` Khem Raj
@ 2021-12-06 23:32 ` Richard Purdie
2021-12-07 2:09 ` Khem Raj
0 siblings, 1 reply; 6+ messages in thread
From: Richard Purdie @ 2021-12-06 23:32 UTC (permalink / raw)
To: Khem Raj; +Cc: openembedded-core
On Mon, 2021-12-06 at 15:28 -0800, Khem Raj wrote:
>
>
> On Mon, Dec 6, 2021 at 2:51 PM Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> > I did test such a patch today as well since we discussed it. I'd propose
> > this
> > commit message since we need to say why:
> >
> > """
> >
> > glibc: Drop ppc sqrt optimisations
> >
> > OpenEmbedded isn't an upstream or a patch repository. These are
> > optimisations
> > which for reasons unknown were never merged into upstream glibc. We have no
> > active ppc maintainers and these patches are at an evolutionary dead end.
> > Since they are optimisations, there may be small performance regressions by
> > removing them but OE can't carry patches like this indefinitely and these
> > appear to have no future.
> > """
> >
>
>
> This sounds good to me. I guess will add it to commit or do you want me to
> send v2. This passed my local testing on qemuppc
>
I can sort if you're ok with it, thanks.
Cheers,
Richard
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc
2021-12-06 23:32 ` Richard Purdie
@ 2021-12-07 2:09 ` Khem Raj
0 siblings, 0 replies; 6+ messages in thread
From: Khem Raj @ 2021-12-07 2:09 UTC (permalink / raw)
To: Richard Purdie; +Cc: openembedded-core
[-- Attachment #1: Type: text/plain, Size: 1164 bytes --]
On Mon, Dec 6, 2021 at 3:32 PM Richard Purdie <
richard.purdie@linuxfoundation.org> wrote:
> On Mon, 2021-12-06 at 15:28 -0800, Khem Raj wrote:
> >
> >
> > On Mon, Dec 6, 2021 at 2:51 PM Richard Purdie
> > <richard.purdie@linuxfoundation.org> wrote:
> > > I did test such a patch today as well since we discussed it. I'd
> propose
> > > this
> > > commit message since we need to say why:
> > >
> > > """
> > >
> > > glibc: Drop ppc sqrt optimisations
> > >
> > > OpenEmbedded isn't an upstream or a patch repository. These are
> > > optimisations
> > > which for reasons unknown were never merged into upstream glibc. We
> have no
> > > active ppc maintainers and these patches are at an evolutionary dead
> end.
> > > Since they are optimisations, there may be small performance
> regressions by
> > > removing them but OE can't carry patches like this indefinitely and
> these
> > > appear to have no future.
> > > """
> > >
> >
> >
> > This sounds good to me. I guess will add it to commit or do you want me
> to
> > send v2. This passed my local testing on qemuppc
> >
>
> I can sort if you're ok with it, thanks.
Sounds good
>
> Cheers,
>
> Richard
>
>
[-- Attachment #2: Type: text/html, Size: 2118 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-12-07 2:10 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-12-06 22:14 [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Khem Raj
2021-12-06 22:14 ` [PATCH 2/2] ffmpeg: Fix build on ppc32 Khem Raj
2021-12-06 22:51 ` [OE-core] [PATCH 1/2] glibc: Drop fsqrt implementation patches for fsl ppc Richard Purdie
2021-12-06 23:28 ` Khem Raj
2021-12-06 23:32 ` Richard Purdie
2021-12-07 2:09 ` Khem Raj
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox