From: Steven Munroe <munroesj@us.ibm.com>
To: Jakub Jelinek <jakub@redhat.com>,
GNU libc hacker <libc-hacker@sources.redhat.com>
Subject: [PATCH] Move powerpc-cpu math functions to trunc
Date: Fri, 01 Jun 2007 21:57:00 -0000 [thread overview]
Message-ID: <46609AB8.8050405@us.ibm.com> (raw)
In-Reply-To: <20070529213339.GJ4033@devserv.devel.redhat.com>
[-- Attachment #1: Type: text/plain, Size: 378 bytes --]
Jakub Jelinek wrote:
> Hi!
>
> Ulrich just asked me about your add-on, are you planning to submit it
> for inclusion into CVS (not as an add-on, but in normal sysdeps hierarchy)?
> He'd certainly prefer that over an add-on, ASAP.
>
> Jakub
>
Here are the math functions and supporting Implies files, ported to
the normal sysdeps hierarchy. Apply the strings patch first.
[-- Attachment #2: ppccpu-math-20070601.txt --]
[-- Type: text/plain, Size: 130003 bytes --]
2007-05-31 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/970/fpu/Implies: New file.
* sysdeps/powerpc/powerpc32/power5/fpu/Implies: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/Implies: New file.
* sysdeps/powerpc/powerpc32/power6/fpu/Implies: New file.
* sysdeps/powerpc/powerpc32/power6x/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/970/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/power5/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/power6/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/power6x/fpu/Implies: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/970/fpu/Implies: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/power4/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/power5/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/power5+/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/power6/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/power6x/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/970/fpu/Implies: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/power4/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/power5/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/power5+/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/power6/fpu/Implies:
New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/power6x/fpu/Implies:
New file.
2007-05-31 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llrint.S: Move.
* sysdeps/powerpc/powerpc32/power4/fpu/s_llrint.S: To here.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llrintf.S: Move.
* sysdeps/powerpc/powerpc32/power4/fpu/s_llrintf.S: To here.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llround.S: Move.
* sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S: To here.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llroundf.S: Move.
* sysdeps/powerpc/powerpc32/power4/fpu/s_llroundf.S: To here.
2007-05-22 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S
(LONG_DOUBLE_COMPAT): Specify correct version, GLIBC_2_1.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S
(LONG_DOUBLE_COMPAT): Specify correct version, GLIBC_2_1.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S
(LONG_DOUBLE_COMPAT): Specify correct version, GLIBC_2_1.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S
(LONG_DOUBLE_COMPAT): Specify correct version, GLIBC_2_1.
2007-05-21 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power4/fpu/slowexp.c: New file.
* sysdeps/powerpc/powerpc32/power4/fpu/w_sqrt.c: New file.
* sysdeps/powerpc/powerpc64/power4/fpu/slowexp.c: New file.
* sysdeps/powerpc/powerpc64/power4/fpu/w_sqrt.c: New file.
2007-03-15 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llrint.S
[LONG_DOUBLE_COMPAT]: Add compat_symbol for llrintl@@GLIBC_2_1.
2006-02-13 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power6/fpu/s_llrint.S: New File
* sysdeps/powerpc/powerpc32/power6/fpu/s_llrintf.S: New File
* sysdeps/powerpc/powerpc32/power6/fpu/s_llround.S: New File
* sysdeps/powerpc/powerpc32/power6/fpu/s_llroundf.S: New File
2006-10-20 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power4/fpu/slowpow.c: New file.
* sysdeps/powerpc/powerpc64/power4/fpu/slowpow.c: New file.
2006-10-03 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llround.S: New file.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llroundf.S: New file.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/Makefile: Moved.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/mpa.c: Likewise.
* sysdeps/powerpc/powerpc32/power4/fpu/Makefile: To here.
* sysdeps/powerpc/powerpc32/power4/fpu/mpa.c: Likewise.
2006-09-29 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power6x/fpu/s_lrint.S: New file.
* sysdeps/powerpc/powerpc32/power6x/fpu/s_lround.S: New file.
* sysdeps/powerpc/powerpc64/power6x/fpu/s_llrint.S: New file.
* sysdeps/powerpc/powerpc64/power6x/fpu/s_llround.S: New file.
2006-09-28 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power5+/fpu/s_llround.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_llroundf.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_lround.S: New file.
* sysdeps/powerpc/powerpc32/power6x/fpu/Implies: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_llround.S: New file.
* sysdeps/powerpc/powerpc64/power6x/fpu/Implies: New file.
2006-08-31 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/powerpc64/fpu/Makefile: New file.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/mpa.c: New file.
* sysdeps/powerpc/powerpc64/power4/fpu/Makefile: New file.
* sysdeps/powerpc/powerpc64/power4/fpu/mpa.c: New file.
2006-06-15 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/power5+/fpu/s_ceil.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_ceilf.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_floor.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_floorf.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_roundf.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S: New file.
* sysdeps/powerpc/powerpc32/power5+/fpu/s_truncf.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_ceil.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_ceilf.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_floor.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_floorf.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_roundf.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S: New file.
* sysdeps/powerpc/powerpc64/power5+/fpu/s_truncf.S: New file.
2006-03-20 Steven Munroe <sjmunroe@us.ibm.com>
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llrint.S: New file.
* sysdeps/powerpc/powerpc32/powerpc64/fpu/s_llrintf.S: New file.
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/970/fpu/Implies libc25/sysdeps/powerpc/powerpc32/970/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/970/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/970/fpu/Implies Thu May 31 08:19:49 2007
@@ -0,0 +1,1 @@
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/Makefile libc25/sysdeps/powerpc/powerpc32/power4/fpu/Makefile
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/Makefile Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/Makefile Tue Oct 10 09:58:56 2006
@@ -0,0 +1,5 @@
+# Makefile fragment for POWER4/5/5+ with FPU.
+
+ifeq ($(subdir),math)
+CFLAGS-mpa.c += --param max-unroll-times=4 -funroll-loops -fpeel-loops -ftree-loop-linear
+endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/mpa.c libc25/sysdeps/powerpc/powerpc32/power4/fpu/mpa.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/mpa.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/mpa.c Tue Oct 10 09:58:56 2006
@@ -0,0 +1,549 @@
+
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2006 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/************************************************************************/
+/* MODULE_NAME: mpa.c */
+/* */
+/* FUNCTIONS: */
+/* mcr */
+/* acr */
+/* cr */
+/* cpy */
+/* cpymn */
+/* norm */
+/* denorm */
+/* mp_dbl */
+/* dbl_mp */
+/* add_magnitudes */
+/* sub_magnitudes */
+/* add */
+/* sub */
+/* mul */
+/* inv */
+/* dvd */
+/* */
+/* Arithmetic functions for multiple precision numbers. */
+/* Relative errors are bounded */
+/************************************************************************/
+
+
+#include "endian.h"
+#include "mpa.h"
+#include "mpa2.h"
+#include <sys/param.h> /* For MIN() */
+/* mcr() compares the sizes of the mantissas of two multiple precision */
+/* numbers. Mantissas are compared regardless of the signs of the */
+/* numbers, even if x->d[0] or y->d[0] are zero. Exponents are also */
+/* disregarded. */
+static int mcr(const mp_no *x, const mp_no *y, int p) {
+ long i;
+ long p2 = p;
+ for (i=1; i<=p2; i++) {
+ if (X[i] == Y[i]) continue;
+ else if (X[i] > Y[i]) return 1;
+ else return -1; }
+ return 0;
+}
+
+
+
+/* acr() compares the absolute values of two multiple precision numbers */
+int __acr(const mp_no *x, const mp_no *y, int p) {
+ long i;
+
+ if (X[0] == ZERO) {
+ if (Y[0] == ZERO) i= 0;
+ else i=-1;
+ }
+ else if (Y[0] == ZERO) i= 1;
+ else {
+ if (EX > EY) i= 1;
+ else if (EX < EY) i=-1;
+ else i= mcr(x,y,p);
+ }
+
+ return i;
+}
+
+
+/* cr90 compares the values of two multiple precision numbers */
+int __cr(const mp_no *x, const mp_no *y, int p) {
+ int i;
+
+ if (X[0] > Y[0]) i= 1;
+ else if (X[0] < Y[0]) i=-1;
+ else if (X[0] < ZERO ) i= __acr(y,x,p);
+ else i= __acr(x,y,p);
+
+ return i;
+}
+
+
+/* Copy a multiple precision number. Set *y=*x. x=y is permissible. */
+void __cpy(const mp_no *x, mp_no *y, int p) {
+ long i;
+
+ EY = EX;
+ for (i=0; i <= p; i++) Y[i] = X[i];
+
+ return;
+}
+
+
+/* Copy a multiple precision number x of precision m into a */
+/* multiple precision number y of precision n. In case n>m, */
+/* the digits of y beyond the m'th are set to zero. In case */
+/* n<m, the digits of x beyond the n'th are ignored. */
+/* x=y is permissible. */
+
+void __cpymn(const mp_no *x, int m, mp_no *y, int n) {
+
+ long i,k;
+ long n2 = n;
+ long m2 = m;
+
+ EY = EX; k=MIN(m2,n2);
+ for (i=0; i <= k; i++) Y[i] = X[i];
+ for ( ; i <= n2; i++) Y[i] = ZERO;
+
+ return;
+}
+
+/* Convert a multiple precision number *x into a double precision */
+/* number *y, normalized case (|x| >= 2**(-1022))) */
+static void norm(const mp_no *x, double *y, int p)
+{
+ #define R radixi.d
+ long i;
+#if 0
+ int k;
+#endif
+ double a,c,u,v,z[5];
+ if (p<5) {
+ if (p==1) c = X[1];
+ else if (p==2) c = X[1] + R* X[2];
+ else if (p==3) c = X[1] + R*(X[2] + R* X[3]);
+ else if (p==4) c =(X[1] + R* X[2]) + R*R*(X[3] + R*X[4]);
+ }
+ else {
+ for (a=ONE, z[1]=X[1]; z[1] < TWO23; )
+ {a *= TWO; z[1] *= TWO; }
+
+ for (i=2; i<5; i++) {
+ z[i] = X[i]*a;
+ u = (z[i] + CUTTER)-CUTTER;
+ if (u > z[i]) u -= RADIX;
+ z[i] -= u;
+ z[i-1] += u*RADIXI;
+ }
+
+ u = (z[3] + TWO71) - TWO71;
+ if (u > z[3]) u -= TWO19;
+ v = z[3]-u;
+
+ if (v == TWO18) {
+ if (z[4] == ZERO) {
+ for (i=5; i <= p; i++) {
+ if (X[i] == ZERO) continue;
+ else {z[3] += ONE; break; }
+ }
+ }
+ else z[3] += ONE;
+ }
+
+ c = (z[1] + R *(z[2] + R * z[3]))/a;
+ }
+
+ c *= X[0];
+
+ for (i=1; i<EX; i++) c *= RADIX;
+ for (i=1; i>EX; i--) c *= RADIXI;
+
+ *y = c;
+ return;
+#undef R
+}
+
+/* Convert a multiple precision number *x into a double precision */
+/* number *y, denormalized case (|x| < 2**(-1022))) */
+static void denorm(const mp_no *x, double *y, int p)
+{
+ long i,k;
+ long p2 = p;
+ double c,u,z[5];
+#if 0
+ double a,v;
+#endif
+
+#define R radixi.d
+ if (EX<-44 || (EX==-44 && X[1]<TWO5))
+ { *y=ZERO; return; }
+
+ if (p2==1) {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=ZERO; z[3]=ZERO; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; z[3]=ZERO; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; z[3]=X[1]; k=1;}
+ }
+ else if (p2==2) {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=X[2]; z[3]=ZERO; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; z[3]=X[2]; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; z[3]=X[1]; k=1;}
+ }
+ else {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=X[2]; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; k=1;}
+ z[3] = X[k];
+ }
+
+ u = (z[3] + TWO57) - TWO57;
+ if (u > z[3]) u -= TWO5;
+
+ if (u==z[3]) {
+ for (i=k+1; i <= p2; i++) {
+ if (X[i] == ZERO) continue;
+ else {z[3] += ONE; break; }
+ }
+ }
+
+ c = X[0]*((z[1] + R*(z[2] + R*z[3])) - TWO10);
+
+ *y = c*TWOM1032;
+ return;
+
+#undef R
+}
+
+/* Convert a multiple precision number *x into a double precision number *y. */
+/* The result is correctly rounded to the nearest/even. *x is left unchanged */
+
+void __mp_dbl(const mp_no *x, double *y, int p) {
+#if 0
+ int i,k;
+ double a,c,u,v,z[5];
+#endif
+
+ if (X[0] == ZERO) {*y = ZERO; return; }
+
+ if (EX> -42) norm(x,y,p);
+ else if (EX==-42 && X[1]>=TWO10) norm(x,y,p);
+ else denorm(x,y,p);
+}
+
+
+/* dbl_mp() converts a double precision number x into a multiple precision */
+/* number *y. If the precision p is too small the result is truncated. x is */
+/* left unchanged. */
+
+void __dbl_mp(double x, mp_no *y, int p) {
+
+ long i,n;
+ long p2 = p;
+ double u;
+
+ /* Sign */
+ if (x == ZERO) {Y[0] = ZERO; return; }
+ else if (x > ZERO) Y[0] = ONE;
+ else {Y[0] = MONE; x=-x; }
+
+ /* Exponent */
+ for (EY=ONE; x >= RADIX; EY += ONE) x *= RADIXI;
+ for ( ; x < ONE; EY -= ONE) x *= RADIX;
+
+ /* Digits */
+ n=MIN(p2,4);
+ for (i=1; i<=n; i++) {
+ u = (x + TWO52) - TWO52;
+ if (u>x) u -= ONE;
+ Y[i] = u; x -= u; x *= RADIX; }
+ for ( ; i<=p2; i++) Y[i] = ZERO;
+ return;
+}
+
+
+/* add_magnitudes() adds the magnitudes of *x & *y assuming that */
+/* abs(*x) >= abs(*y) > 0. */
+/* The sign of the sum *z is undefined. x&y may overlap but not x&z or y&z. */
+/* No guard digit is used. The result equals the exact sum, truncated. */
+/* *x & *y are left unchanged. */
+
+static void add_magnitudes(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i,j,k;
+ long p2 = p;
+
+ EZ = EX;
+
+ i=p2; j=p2+ EY - EX; k=p2+1;
+
+ if (j<1)
+ {__cpy(x,z,p); return; }
+ else Z[k] = ZERO;
+
+ for (; j>0; i--,j--) {
+ Z[k] += X[i] + Y[j];
+ if (Z[k] >= RADIX) {
+ Z[k] -= RADIX;
+ Z[--k] = ONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (; i>0; i--) {
+ Z[k] += X[i];
+ if (Z[k] >= RADIX) {
+ Z[k] -= RADIX;
+ Z[--k] = ONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ if (Z[1] == ZERO) {
+ for (i=1; i<=p2; i++) Z[i] = Z[i+1]; }
+ else EZ += ONE;
+}
+
+
+/* sub_magnitudes() subtracts the magnitudes of *x & *y assuming that */
+/* abs(*x) > abs(*y) > 0. */
+/* The sign of the difference *z is undefined. x&y may overlap but not x&z */
+/* or y&z. One guard digit is used. The error is less than one ulp. */
+/* *x & *y are left unchanged. */
+
+static void sub_magnitudes(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i,j,k;
+ long p2 = p;
+
+ EZ = EX;
+
+ if (EX == EY) {
+ i=j=k=p2;
+ Z[k] = Z[k+1] = ZERO; }
+ else {
+ j= EX - EY;
+ if (j > p2) {__cpy(x,z,p); return; }
+ else {
+ i=p2; j=p2+1-j; k=p2;
+ if (Y[j] > ZERO) {
+ Z[k+1] = RADIX - Y[j--];
+ Z[k] = MONE; }
+ else {
+ Z[k+1] = ZERO;
+ Z[k] = ZERO; j--;}
+ }
+ }
+
+ for (; j>0; i--,j--) {
+ Z[k] += (X[i] - Y[j]);
+ if (Z[k] < ZERO) {
+ Z[k] += RADIX;
+ Z[--k] = MONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (; i>0; i--) {
+ Z[k] += X[i];
+ if (Z[k] < ZERO) {
+ Z[k] += RADIX;
+ Z[--k] = MONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (i=1; Z[i] == ZERO; i++) ;
+ EZ = EZ - i + 1;
+ for (k=1; i <= p2+1; )
+ Z[k++] = Z[i++];
+ for (; k <= p2; )
+ Z[k++] = ZERO;
+
+ return;
+}
+
+
+/* Add two multiple precision numbers. Set *z = *x + *y. x&y may overlap */
+/* but not x&z or y&z. One guard digit is used. The error is less than */
+/* one ulp. *x & *y are left unchanged. */
+
+void __add(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ int n;
+
+ if (X[0] == ZERO) {__cpy(y,z,p); return; }
+ else if (Y[0] == ZERO) {__cpy(x,z,p); return; }
+
+ if (X[0] == Y[0]) {
+ if (__acr(x,y,p) > 0) {add_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else {add_magnitudes(y,x,z,p); Z[0] = Y[0]; }
+ }
+ else {
+ if ((n=__acr(x,y,p)) == 1) {sub_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else if (n == -1) {sub_magnitudes(y,x,z,p); Z[0] = Y[0]; }
+ else Z[0] = ZERO;
+ }
+ return;
+}
+
+
+/* Subtract two multiple precision numbers. *z is set to *x - *y. x&y may */
+/* overlap but not x&z or y&z. One guard digit is used. The error is */
+/* less than one ulp. *x & *y are left unchanged. */
+
+void __sub(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ int n;
+
+ if (X[0] == ZERO) {__cpy(y,z,p); Z[0] = -Z[0]; return; }
+ else if (Y[0] == ZERO) {__cpy(x,z,p); return; }
+
+ if (X[0] != Y[0]) {
+ if (__acr(x,y,p) > 0) {add_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else {add_magnitudes(y,x,z,p); Z[0] = -Y[0]; }
+ }
+ else {
+ if ((n=__acr(x,y,p)) == 1) {sub_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else if (n == -1) {sub_magnitudes(y,x,z,p); Z[0] = -Y[0]; }
+ else Z[0] = ZERO;
+ }
+ return;
+}
+
+
+/* Multiply two multiple precision numbers. *z is set to *x * *y. x&y */
+/* may overlap but not x&z or y&z. In case p=1,2,3 the exact result is */
+/* truncated to p digits. In case p>3 the error is bounded by 1.001 ulp. */
+/* *x & *y are left unchanged. */
+
+void __mul(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i, i1, i2, j, k, k2;
+ long p2 = p;
+ double u, zk, zk2;
+
+ /* Is z=0? */
+ if (X[0]*Y[0]==ZERO)
+ { Z[0]=ZERO; return; }
+
+ /* Multiply, add and carry */
+ k2 = (p2<3) ? p2+p2 : p2+3;
+ zk = Z[k2]=ZERO;
+ for (k=k2; k>1; ) {
+ if (k > p2) {i1=k-p2; i2=p2+1; }
+ else {i1=1; i2=k; }
+#if 1
+ /* rearange this inner loop to allow the fmadd instructions to be
+ independent and execute in parallel on processors that have
+ dual symetrical FP pipelines. */
+ if (i1 < (i2-1))
+ {
+ /* make sure we have at least 2 iterations */
+ if (((i2 - i1) & 1L) == 1L)
+ {
+ /* Handle the odd iterations case. */
+ zk2 = x->d[i2-1]*y->d[i1];
+ }
+ else
+ zk2 = zero.d;
+ /* Do two multiply/adds per loop iteration, using independent
+ accumulators; zk and zk2. */
+ for (i=i1,j=i2-1; i<i2-1; i+=2,j-=2)
+ {
+ zk += x->d[i]*y->d[j];
+ zk2 += x->d[i+1]*y->d[j-1];
+ }
+ zk += zk2; /* final sum. */
+ }
+ else
+ {
+ /* Special case when iterations is 1. */
+ zk += x->d[i1]*y->d[i1];
+ }
+#else
+ /* The orginal code. */
+ for (i=i1,j=i2-1; i<i2; i++,j--) zk += X[i]*Y[j];
+#endif
+
+ u = (zk + CUTTER)-CUTTER;
+ if (u > zk) u -= RADIX;
+ Z[k] = zk - u;
+ zk = u*RADIXI;
+ --k;
+ }
+ Z[k] = zk;
+
+ /* Is there a carry beyond the most significant digit? */
+ if (Z[1] == ZERO) {
+ for (i=1; i<=p2; i++) Z[i]=Z[i+1];
+ EZ = EX + EY - 1; }
+ else
+ EZ = EX + EY;
+
+ Z[0] = X[0] * Y[0];
+ return;
+}
+
+
+/* Invert a multiple precision number. Set *y = 1 / *x. */
+/* Relative error bound = 1.001*r**(1-p) for p=2, 1.063*r**(1-p) for p=3, */
+/* 2.001*r**(1-p) for p>3. */
+/* *x=0 is not permissible. *x is left unchanged. */
+
+void __inv(const mp_no *x, mp_no *y, int p) {
+ long i;
+#if 0
+ int l;
+#endif
+ double t;
+ mp_no z,w;
+ static const int np1[] = {0,0,0,0,1,2,2,2,2,3,3,3,3,3,3,3,3,3,
+ 4,4,4,4,4,4,4,4,4,4,4,4,4,4,4};
+ const mp_no mptwo = {1,{1.0,2.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0}};
+
+ __cpy(x,&z,p); z.e=0; __mp_dbl(&z,&t,p);
+ t=ONE/t; __dbl_mp(t,y,p); EY -= EX;
+
+ for (i=0; i<np1[p]; i++) {
+ __cpy(y,&w,p);
+ __mul(x,&w,y,p);
+ __sub(&mptwo,y,&z,p);
+ __mul(&w,&z,y,p);
+ }
+ return;
+}
+
+
+/* Divide one multiple precision number by another.Set *z = *x / *y. *x & *y */
+/* are left unchanged. x&y may overlap but not x&z or y&z. */
+/* Relative error bound = 2.001*r**(1-p) for p=2, 2.063*r**(1-p) for p=3 */
+/* and 3.001*r**(1-p) for p>3. *y=0 is not permissible. */
+
+void __dvd(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ mp_no w;
+
+ if (X[0] == ZERO) Z[0] = ZERO;
+ else {__inv(y,&w,p); __mul(x,&w,z,p);}
+ return;
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llrint.S libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llrint.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llrint.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llrint.S Thu May 10 11:28:20 2007
@@ -0,0 +1,47 @@
+/* Round double to long int. PowerPC32 on PowerPC64 version.
+ Copyright (C) 2004, 2006 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., 51 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long long int[r3, r4] __llrint (double x[fp1]) */
+ENTRY (__llrint)
+ CALL_MCOUNT
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ fctid fp13,fp1
+ stfd fp13,8(r1)
+ nop /* Insure the following load is in a different dispatch group */
+ nop /* to avoid pipe stall on POWER4&5. */
+ nop
+ lwz r3,8(r1)
+ lwz r4,12(r1)
+ addi r1,r1,16
+ blr
+ END (__llrint)
+
+weak_alias (__llrint, llrint)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__llrint, __llrintl)
+weak_alias (__llrint, llrintl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llrint, llrintl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llrintf.S libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llrintf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llrintf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llrintf.S Thu Apr 06 10:50:35 2006
@@ -0,0 +1,39 @@
+/* Round float to long int. PowerPC32 on PowerPC64 version.
+ Copyright (C) 2004, 2006 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., 51 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+
+/* long long int[r3, r4] __llrintf (float x[fp1]) */
+ENTRY (__llrintf)
+ CALL_MCOUNT
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ fctid fp13,fp1
+ stfd fp13,8(r1)
+ nop /* Insure the following load is in a different dispatch group */
+ nop /* to avoid pipe stall on POWER4&5. */
+ nop
+ lwz r3,8(r1)
+ lwz r4,12(r1)
+ addi r1,r1,16
+ blr
+ END (__llrintf)
+
+weak_alias (__llrintf, llrintf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S Tue Oct 10 09:39:01 2006
@@ -0,0 +1,97 @@
+/* llround function. PowerPC32 on PowerPC64 version.
+ Copyright (C) 2004, 2006 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., 1 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .section .rodata.cst8,"aM",@progbits,8
+ .align 2
+.LC0: /* 0.0 */
+ .long 0x00000000
+.LC1: /* 0.5 */
+ .long 0x3f000000
+
+ .section ".text"
+
+/* long [r3] lround (float x [fp1])
+ IEEE 1003.1 lround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we can't use the PowerPC "round to Nearest" mode. Instead we set
+ "round toward Zero" mode and round by adding +-0.5 before rounding
+ to the integer value. */
+
+ENTRY (__llround)
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+#ifdef SHARED
+ mflr r11
+ cfi_register(lr,r11)
+# ifdef HAVE_ASM_PPC_REL16
+ bcl 20,31,1f
+1: mflr r9
+ addis r9,r9,.LC0-1b@ha
+ addi r9,r9,.LC0-1b@l
+# else
+ bl _GLOBAL_OFFSET_TABLE_@local-4
+ mflr r10
+ lwz r9,.LC0@got(10)
+# endif
+ mtlr r11
+ cfi_same_value (lr)
+ lfs fp12,0(r9)
+ lfs fp10,.LC1-.LC0(r9)
+#else
+ lis r9,.LC0@ha
+ lis r10,.LC1@ha
+ lfs fp12,.LC0@l(r9)
+ lfs fp10,.LC1@l(r10)
+#endif
+ fcmpu cr6,fp1,fp12 /* if (x > 0.0) */
+ ble- cr6,.L4
+ fadd fp1,fp1,fp10 /* x+= 0.5; */
+.L9:
+ fctidz fp2,fp1 /* Convert To Integer DW round toward 0. */
+ stfd fp2,8(r1)
+ nop /* Ensure the following load is in a different dispatch */
+ nop /* group to avoid pipe stall on POWER4&5. */
+ nop
+ lwz r4,12(r1)
+ lwz r3,8(r1)
+ addi r1,r1,16
+ blr
+.L4:
+ fsub fp1,fp1,fp10 /* x-= 0.5; */
+ b .L9
+ END (__llround)
+
+weak_alias (__llround, llround)
+
+strong_alias (__llround, __llroundf)
+weak_alias (__llround, llroundf)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__llround, llroundl)
+strong_alias (__llround, __llroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llround, llroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llroundf.S libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llroundf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/s_llroundf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/s_llroundf.S Tue Oct 10 09:39:01 2006
@@ -0,0 +1,1 @@
+/* __llroundf is in s_llround.S */
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/slowexp.c libc25/sysdeps/powerpc/powerpc32/power4/fpu/slowexp.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/slowexp.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/slowexp.c Mon May 21 11:46:25 2007
@@ -0,0 +1,66 @@
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2007 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/**************************************************************************/
+/* MODULE_NAME:slowexp.c */
+/* */
+/* FUNCTION:slowexp */
+/* */
+/* FILES NEEDED:mpa.h */
+/* mpa.c mpexp.c */
+/* */
+/*Converting from double precision to Multi-precision and calculating */
+/* e^x */
+/**************************************************************************/
+#include "math_private.h"
+
+#ifdef NO_LONG_DOUBLE
+#include "mpa.h"
+void __mpexp(mp_no *x, mp_no *y, int p);
+#endif
+
+/*Converting from double precision to Multi-precision and calculating e^x */
+double __slowexp(double x) {
+#ifdef NO_LONG_DOUBLE
+ double w,z,res,eps=3.0e-26;
+ int p;
+ mp_no mpx, mpy, mpz,mpw,mpeps,mpcor;
+
+ p=6;
+ __dbl_mp(x,&mpx,p); /* Convert a double precision number x */
+ /* into a multiple precision number mpx with prec. p. */
+ __mpexp(&mpx, &mpy, p); /* Multi-Precision exponential function */
+ __dbl_mp(eps,&mpeps,p);
+ __mul(&mpeps,&mpy,&mpcor,p);
+ __add(&mpy,&mpcor,&mpw,p);
+ __sub(&mpy,&mpcor,&mpz,p);
+ __mp_dbl(&mpw, &w, p);
+ __mp_dbl(&mpz, &z, p);
+ if (w == z) return w;
+ else { /* if calculating is not exactly */
+ p = 32;
+ __dbl_mp(x,&mpx,p);
+ __mpexp(&mpx, &mpy, p);
+ __mp_dbl(&mpy, &res, p);
+ return res;
+ }
+#else
+ return (double) __ieee754_expl((long double)x);
+#endif
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/slowpow.c libc25/sysdeps/powerpc/powerpc32/power4/fpu/slowpow.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/slowpow.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/slowpow.c Fri Oct 20 15:30:05 2006
@@ -0,0 +1,94 @@
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2006 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/*************************************************************************/
+/* MODULE_NAME:slowpow.c */
+/* */
+/* FUNCTION:slowpow */
+/* */
+/*FILES NEEDED:mpa.h */
+/* mpa.c mpexp.c mplog.c halfulp.c */
+/* */
+/* Given two IEEE double machine numbers y,x , routine computes the */
+/* correctly rounded (to nearest) value of x^y. Result calculated by */
+/* multiplication (in halfulp.c) or if result isn't accurate enough */
+/* then routine converts x and y into multi-precision doubles and */
+/* recompute. */
+/*************************************************************************/
+
+#include "mpa.h"
+#include "math_private.h"
+
+void __mpexp (mp_no * x, mp_no * y, int p);
+void __mplog (mp_no * x, mp_no * y, int p);
+double ulog (double);
+double __halfulp (double x, double y);
+
+double
+__slowpow (double x, double y, double z)
+{
+ double res, res1;
+ long double ldw, ldz, ldpp;
+ static const long double ldeps = 0x4.0p-96;
+
+ res = __halfulp (x, y); /* halfulp() returns -10 or x^y */
+ if (res >= 0)
+ return res; /* if result was really computed by halfulp */
+ /* else, if result was not really computed by halfulp */
+
+ /* Compute pow as long double, 106 bits */
+ ldz = __ieee754_logl ((long double) x);
+ ldw = (long double) y *ldz;
+ ldpp = __ieee754_expl (ldw);
+ res = (double) (ldpp + ldeps);
+ res1 = (double) (ldpp - ldeps);
+
+ if (res != res1) /* if result still not accurate enough */
+ { /* use mpa for higher persision. */
+ mp_no mpx, mpy, mpz, mpw, mpp, mpr, mpr1;
+ static const mp_no eps = { -3, {1.0, 4.0} };
+ int p;
+
+ p = 10; /* p=precision 240 bits */
+ __dbl_mp (x, &mpx, p);
+ __dbl_mp (y, &mpy, p);
+ __dbl_mp (z, &mpz, p);
+ __mplog (&mpx, &mpz, p); /* log(x) = z */
+ __mul (&mpy, &mpz, &mpw, p); /* y * z =w */
+ __mpexp (&mpw, &mpp, p); /* e^w =pp */
+ __add (&mpp, &eps, &mpr, p); /* pp+eps =r */
+ __mp_dbl (&mpr, &res, p);
+ __sub (&mpp, &eps, &mpr1, p); /* pp -eps =r1 */
+ __mp_dbl (&mpr1, &res1, p); /* converting into double precision */
+ if (res == res1)
+ return res;
+
+ /* if we get here result wasn't calculated exactly, continue for
+ more exact calculation using 768 bits. */
+ p = 32;
+ __dbl_mp (x, &mpx, p);
+ __dbl_mp (y, &mpy, p);
+ __dbl_mp (z, &mpz, p);
+ __mplog (&mpx, &mpz, p); /* log(c)=z */
+ __mul (&mpy, &mpz, &mpw, p); /* y*z =w */
+ __mpexp (&mpw, &mpp, p); /* e^w=pp */
+ __mp_dbl (&mpp, &res, p); /* converting into double precision */
+ }
+ return res;
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrt.c libc25/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrt.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrt.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrt.c Fri Jun 01 10:16:26 2007
@@ -0,0 +1,62 @@
+/* Double-precision floating point square root wrapper.
+ Copyright (C) 2004, 2007 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_ldbl_opt.h>
+#include "math.h"
+#include "math_private.h"
+#include <fenv_libc.h>
+
+#ifdef __STDC__
+double
+__sqrt (double x) /* wrapper sqrt */
+#else
+double
+__sqrt (x) /* wrapper sqrt */
+ double x;
+#endif
+{
+ double z;
+/* Power4 (ISA V2.0) and above implement sqrt in hardware. */
+ __asm __volatile (
+ " fsqrt %0,%1\n"
+ : "=f" (z)
+ : "f" (x));
+#ifdef _IEEE_LIBM
+ return z;
+#else
+ if (__builtin_expect (_LIB_VERSION == _IEEE_, 0))
+ return z;
+
+ if (__builtin_expect (x != x, 0))
+ return z;
+
+ if (__builtin_expect (x < 0.0, 0))
+ return __kernel_standard (x, x, 26); /* sqrt(negative) */
+ else
+ return z;
+#endif
+}
+
+weak_alias (__sqrt, sqrt)
+#ifdef NO_LONG_DOUBLE
+ strong_alias (__sqrt, __sqrtl) weak_alias (__sqrt, sqrtl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __sqrt, sqrtl, GLIBC_2_0);
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrtf.c libc25/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrtf.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrtf.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power4/fpu/w_sqrtf.c Fri Jun 01 10:23:29 2007
@@ -0,0 +1,60 @@
+/* Single-precision floating point square root wrapper.
+ Copyright (C) 2004, 2007 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 <sysdep.h>
+#include <ldsodefs.h>
+
+#ifdef __STDC__
+float
+__sqrtf (float x) /* wrapper sqrtf */
+#else
+float
+__sqrtf (x) /* wrapper sqrtf */
+ float x;
+#endif
+{
+#ifdef _IEEE_LIBM
+ return __ieee754_sqrtf (x);
+#else
+ float z;
+/* Power4 (ISA V2.0) and above implement sqrtf in hardware. */
+ __asm __volatile (
+ " fsqrts %0,%1\n"
+ : "=f" (z)
+ : "f" (x));
+
+ if (__builtin_expect (_LIB_VERSION == _IEEE_, 0))
+ return z;
+
+ if (__builtin_expect (x != x, 0))
+ return z;
+
+ if (__builtin_expect (x < 0.0, 0))
+ /* sqrtf(negative) */
+ return (float) __kernel_standard ((double) x, (double) x, 126);
+ else
+ return z;
+#endif
+}
+
+weak_alias (__sqrtf, sqrtf)
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/Implies libc25/sysdeps/powerpc/powerpc32/power5+/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/Implies Thu May 31 08:20:34 2007
@@ -0,0 +1,1 @@
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceil.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceil.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceil.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceil.S Thu Jun 15 13:29:18 2006
@@ -0,0 +1,37 @@
+/* ceil function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__ceil, 4, 0)
+ frip fp1, fp1
+ blr
+ END (__ceil)
+
+weak_alias (__ceil, ceil)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__ceil, ceill)
+strong_alias (__ceil, __ceill)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __ceil, ceill, GLIBC_2_0)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceilf.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceilf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceilf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_ceilf.S Thu Jun 15 13:29:08 2006
@@ -0,0 +1,30 @@
+/* ceilf function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__ceilf, 4, 0)
+ frip fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__ceilf)
+
+weak_alias (__ceilf, ceilf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_floor.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_floor.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_floor.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_floor.S Thu Jun 15 13:28:56 2006
@@ -0,0 +1,37 @@
+/* floor function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__floor, 4, 0)
+ frim fp1, fp1
+ blr
+ END (__floor)
+
+weak_alias (__floor, floor)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__floor, floorl)
+strong_alias (__floor, __floorl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __floor, floorl, GLIBC_2_0)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_floorf.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_floorf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_floorf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_floorf.S Thu Jun 15 13:28:47 2006
@@ -0,0 +1,30 @@
+/* floorf function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__floorf, 4, 0)
+ frim fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__floorf)
+
+weak_alias (__floorf, floorf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_llround.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_llround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_llround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_llround.S Mon Oct 09 14:00:22 2006
@@ -0,0 +1,60 @@
+/* lround function. POWER5+, PowerPC32 version.
+ Copyright (C) 2006 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., 1 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long [r3] llround (float x [fp1])
+ IEEE 1003.1 lround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use the Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power5"
+ENTRY (__llround)
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ frin fp2,fp1
+ fctidz fp3,fp2 /* Convert To Integer Word lround toward 0. */
+ stfd fp3,8(r1)
+ nop /* Ensure the following load is in a different dispatch */
+ nop /* group to avoid pipe stall on POWER4&5. */
+ nop
+ lwz r4,12(r1)
+ lwz r3,8(r1)
+ addi r1,r1,16
+ blr
+ END (__llround)
+
+weak_alias (__llround, llround)
+
+strong_alias (__llround, __llroundf)
+weak_alias (__llround, llroundf)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__llround, llroundl)
+strong_alias (__llround, __llroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llround, llroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_llroundf.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_llroundf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_llroundf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_llroundf.S Mon Oct 09 14:00:22 2006
@@ -0,0 +1,2 @@
+/* __llroundf is in s_llround.S */
+/* __llroundf is in s_llround.S */
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_lround.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_lround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_lround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_lround.S Mon Oct 09 14:00:22 2006
@@ -0,0 +1,58 @@
+/* lround function. POWER5+, PowerPC32 version.
+ Copyright (C) 2006 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., 1 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long [r3] lround (float x [fp1])
+ IEEE 1003.1 lround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use the Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power5"
+ENTRY (__lround)
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ frin fp2,fp1
+ fctiwz fp3,fp2 /* Convert To Integer Word lround toward 0. */
+ stfd fp3,8(r1)
+ nop /* Ensure the following load is in a different dispatch */
+ nop /* group to avoid pipe stall on POWER4&5. */
+ nop
+ lwz r3,12(r1)
+ addi r1,r1,16
+ blr
+ END (__lround)
+
+weak_alias (__lround, lround)
+
+strong_alias (__lround, __lroundf)
+weak_alias (__lround, lroundf)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__lround, lroundl)
+strong_alias (__lround, __lroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_round.S Tue May 22 17:57:07 2007
@@ -0,0 +1,37 @@
+/* round function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__round, 4, 0)
+ frin fp1, fp1
+ blr
+ END (__round)
+
+weak_alias (__round, round)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__round, roundl)
+strong_alias (__round, __roundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __round, roundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_roundf.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_roundf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_roundf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_roundf.S Thu Jun 15 13:29:47 2006
@@ -0,0 +1,30 @@
+/* roundf function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__roundf, 4, 0)
+ frin fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__roundf)
+
+weak_alias (__roundf, roundf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_trunc.S Tue May 22 17:58:40 2007
@@ -0,0 +1,37 @@
+/* trunc function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__trunc, 4, 0)
+ friz fp1, fp1
+ blr
+ END (__trunc)
+
+weak_alias (__trunc, trunc)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__trunc, truncl)
+strong_alias (__trunc, __truncl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __trunc, truncl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_truncf.S libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_truncf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5+/fpu/s_truncf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5+/fpu/s_truncf.S Thu Jun 15 13:30:33 2006
@@ -0,0 +1,30 @@
+/* truncf function. PowerPC32/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__truncf, 4, 0)
+ friz fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__truncf)
+
+weak_alias (__truncf, truncf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5/fpu/Implies libc25/sysdeps/powerpc/powerpc32/power5/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power5/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power5/fpu/Implies Thu May 31 08:20:16 2007
@@ -0,0 +1,1 @@
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/Implies libc25/sysdeps/powerpc/powerpc32/power6/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6/fpu/Implies Thu May 31 08:20:49 2007
@@ -0,0 +1,2 @@
+powerpc/powerpc32/power5+/fpu
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llrint.S libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llrint.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llrint.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llrint.S Tue May 22 17:43:36 2007
@@ -0,0 +1,46 @@
+/* Round double to long int. PowerPC32 on PowerPC64 version.
+ Copyright (C) 2004, 2006, 2007 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., 51 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+
+/* long long int[r3, r4] __llrint (double x[fp1]) */
+ENTRY (__llrint)
+ CALL_MCOUNT
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ fctid fp13,fp1
+ stfd fp13,8(r1)
+/* Insure the following load is in a different dispatch group by
+ inserting "group ending nop". */
+ ori r1,r1,0
+ lwz r3,8(r1)
+ lwz r4,12(r1)
+ addi r1,r1,16
+ blr
+ END (__llrint)
+
+weak_alias (__llrint, llrint)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__llrint, __llrintl)
+weak_alias (__llrint, llrintl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llrint, llrintl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llrintf.S libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llrintf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llrintf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llrintf.S Thu May 10 11:28:20 2007
@@ -0,0 +1,39 @@
+/* Round float to long int. PowerPC32 on PowerPC64 version.
+ Copyright (C) 2004, 2006, 2007 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., 51 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+
+/* long long int[r3, r4] __llrintf (float x[fp1]) */
+ENTRY (__llrintf)
+ CALL_MCOUNT
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ fctid fp13,fp1
+ stfd fp13,8(r1)
+/* Insure the following load is in a different dispatch group by
+ inserting "group ending nop". */
+ ori r1,r1,0
+ lwz r3,8(r1)
+ lwz r4,12(r1)
+ addi r1,r1,16
+ blr
+ END (__llrintf)
+
+weak_alias (__llrintf, llrintf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llround.S libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llround.S Thu May 10 11:28:20 2007
@@ -0,0 +1,60 @@
+/* lround function. POWER5+, PowerPC32 version.
+ Copyright (C) 2006, 2007 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., 1 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long [r3] llround (float x [fp1])
+ IEEE 1003.1 lround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use the Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power5"
+ENTRY (__llround)
+ stwu r1,-16(r1)
+ cfi_adjust_cfa_offset (16)
+ frin fp2,fp1
+ fctidz fp3,fp2 /* Convert To Integer Word lround toward 0. */
+ stfd fp3,8(r1)
+/* Insure the following load is in a different dispatch group by
+ inserting "group ending nop". */
+ ori r1,r1,0
+ lwz r4,12(r1)
+ lwz r3,8(r1)
+ addi r1,r1,16
+ blr
+ END (__llround)
+
+weak_alias (__llround, llround)
+
+strong_alias (__llround, __llroundf)
+weak_alias (__llround, llroundf)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__llround, llroundl)
+strong_alias (__llround, __llroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llround, llroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llroundf.S libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llroundf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6/fpu/s_llroundf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6/fpu/s_llroundf.S Thu May 10 11:28:20 2007
@@ -0,0 +1,2 @@
+/* __llroundf is in s_llround.S */
+/* __llroundf is in s_llround.S */
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/Implies libc25/sysdeps/powerpc/powerpc32/power6x/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6x/fpu/Implies Thu May 31 08:21:03 2007
@@ -0,0 +1,3 @@
+powerpc/powerpc32/power6/fpu
+powerpc/powerpc32/power5+/fpu
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/s_lrint.S libc25/sysdeps/powerpc/powerpc32/power6x/fpu/s_lrint.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/s_lrint.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6x/fpu/s_lrint.S Mon Oct 09 14:00:56 2006
@@ -0,0 +1,42 @@
+/* Round double to long int. POWER6x PowerPC32 version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power6"
+/* long int[r3] __lrint (double x[fp1]) */
+ENTRY (__lrint)
+ fctiw fp13,fp1
+ mftgpr r3,fp13
+ blr
+ END (__lrint)
+
+weak_alias (__lrint, lrint)
+
+strong_alias (__lrint, __lrintf)
+weak_alias (__lrint, lrintf)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__lrint, __lrintl)
+weak_alias (__lrint, lrintl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __lrint, lrintl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/s_lround.S libc25/sysdeps/powerpc/powerpc32/power6x/fpu/s_lround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc32/power6x/fpu/s_lround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc32/power6x/fpu/s_lround.S Mon Oct 09 14:00:56 2006
@@ -0,0 +1,52 @@
+/* lround function. POWER6x, PowerPC32 version.
+ Copyright (C) 2006 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., 1 Franklin Street, Fifth Floor, Boston MA
+ 02110-1301 USA. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long [r3] lround (float x [fp1])
+ IEEE 1003.1 lround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use the Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power6"
+ENTRY (__lround)
+ frin fp2,fp1 /* Pre-round +-0.5. */
+ fctiwz fp3,fp2 /* Convert To Integer Word lround toward 0. */
+ mftgpr r3,fp3 /* Transfer fpr3 to r3. */
+ blr
+ END (__lround)
+
+weak_alias (__lround, lround)
+
+strong_alias (__lround, __lroundf)
+weak_alias (__lround, lroundf)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__lround, lroundl)
+strong_alias (__lround, __lroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/970/fpu/Implies libc25/sysdeps/powerpc/powerpc64/970/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/970/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/970/fpu/Implies Sun Aug 20 09:06:20 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/Makefile libc25/sysdeps/powerpc/powerpc64/power4/fpu/Makefile
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/Makefile Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/Makefile Thu Aug 10 12:49:45 2006
@@ -0,0 +1,5 @@
+# Makefile fragment for POWER4/5/5+ platforms with FPU.
+
+ifeq ($(subdir),math)
+CFLAGS-mpa.c += --param max-unroll-times=4 -funroll-loops -fpeel-loops -ftree-loop-linear
+endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/mpa.c libc25/sysdeps/powerpc/powerpc64/power4/fpu/mpa.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/mpa.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/mpa.c Sun Aug 20 08:59:40 2006
@@ -0,0 +1,549 @@
+
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2006 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/************************************************************************/
+/* MODULE_NAME: mpa.c */
+/* */
+/* FUNCTIONS: */
+/* mcr */
+/* acr */
+/* cr */
+/* cpy */
+/* cpymn */
+/* norm */
+/* denorm */
+/* mp_dbl */
+/* dbl_mp */
+/* add_magnitudes */
+/* sub_magnitudes */
+/* add */
+/* sub */
+/* mul */
+/* inv */
+/* dvd */
+/* */
+/* Arithmetic functions for multiple precision numbers. */
+/* Relative errors are bounded */
+/************************************************************************/
+
+
+#include "endian.h"
+#include "mpa.h"
+#include "mpa2.h"
+#include <sys/param.h> /* For MIN() */
+/* mcr() compares the sizes of the mantissas of two multiple precision */
+/* numbers. Mantissas are compared regardless of the signs of the */
+/* numbers, even if x->d[0] or y->d[0] are zero. Exponents are also */
+/* disregarded. */
+static int mcr(const mp_no *x, const mp_no *y, int p) {
+ long i;
+ long p2 = p;
+ for (i=1; i<=p2; i++) {
+ if (X[i] == Y[i]) continue;
+ else if (X[i] > Y[i]) return 1;
+ else return -1; }
+ return 0;
+}
+
+
+
+/* acr() compares the absolute values of two multiple precision numbers */
+int __acr(const mp_no *x, const mp_no *y, int p) {
+ long i;
+
+ if (X[0] == ZERO) {
+ if (Y[0] == ZERO) i= 0;
+ else i=-1;
+ }
+ else if (Y[0] == ZERO) i= 1;
+ else {
+ if (EX > EY) i= 1;
+ else if (EX < EY) i=-1;
+ else i= mcr(x,y,p);
+ }
+
+ return i;
+}
+
+
+/* cr90 compares the values of two multiple precision numbers */
+int __cr(const mp_no *x, const mp_no *y, int p) {
+ int i;
+
+ if (X[0] > Y[0]) i= 1;
+ else if (X[0] < Y[0]) i=-1;
+ else if (X[0] < ZERO ) i= __acr(y,x,p);
+ else i= __acr(x,y,p);
+
+ return i;
+}
+
+
+/* Copy a multiple precision number. Set *y=*x. x=y is permissible. */
+void __cpy(const mp_no *x, mp_no *y, int p) {
+ long i;
+
+ EY = EX;
+ for (i=0; i <= p; i++) Y[i] = X[i];
+
+ return;
+}
+
+
+/* Copy a multiple precision number x of precision m into a */
+/* multiple precision number y of precision n. In case n>m, */
+/* the digits of y beyond the m'th are set to zero. In case */
+/* n<m, the digits of x beyond the n'th are ignored. */
+/* x=y is permissible. */
+
+void __cpymn(const mp_no *x, int m, mp_no *y, int n) {
+
+ long i,k;
+ long n2 = n;
+ long m2 = m;
+
+ EY = EX; k=MIN(m2,n2);
+ for (i=0; i <= k; i++) Y[i] = X[i];
+ for ( ; i <= n2; i++) Y[i] = ZERO;
+
+ return;
+}
+
+/* Convert a multiple precision number *x into a double precision */
+/* number *y, normalized case (|x| >= 2**(-1022))) */
+static void norm(const mp_no *x, double *y, int p)
+{
+ #define R radixi.d
+ long i;
+#if 0
+ int k;
+#endif
+ double a,c,u,v,z[5];
+ if (p<5) {
+ if (p==1) c = X[1];
+ else if (p==2) c = X[1] + R* X[2];
+ else if (p==3) c = X[1] + R*(X[2] + R* X[3]);
+ else if (p==4) c =(X[1] + R* X[2]) + R*R*(X[3] + R*X[4]);
+ }
+ else {
+ for (a=ONE, z[1]=X[1]; z[1] < TWO23; )
+ {a *= TWO; z[1] *= TWO; }
+
+ for (i=2; i<5; i++) {
+ z[i] = X[i]*a;
+ u = (z[i] + CUTTER)-CUTTER;
+ if (u > z[i]) u -= RADIX;
+ z[i] -= u;
+ z[i-1] += u*RADIXI;
+ }
+
+ u = (z[3] + TWO71) - TWO71;
+ if (u > z[3]) u -= TWO19;
+ v = z[3]-u;
+
+ if (v == TWO18) {
+ if (z[4] == ZERO) {
+ for (i=5; i <= p; i++) {
+ if (X[i] == ZERO) continue;
+ else {z[3] += ONE; break; }
+ }
+ }
+ else z[3] += ONE;
+ }
+
+ c = (z[1] + R *(z[2] + R * z[3]))/a;
+ }
+
+ c *= X[0];
+
+ for (i=1; i<EX; i++) c *= RADIX;
+ for (i=1; i>EX; i--) c *= RADIXI;
+
+ *y = c;
+ return;
+#undef R
+}
+
+/* Convert a multiple precision number *x into a double precision */
+/* number *y, denormalized case (|x| < 2**(-1022))) */
+static void denorm(const mp_no *x, double *y, int p)
+{
+ long i,k;
+ long p2 = p;
+ double c,u,z[5];
+#if 0
+ double a,v;
+#endif
+
+#define R radixi.d
+ if (EX<-44 || (EX==-44 && X[1]<TWO5))
+ { *y=ZERO; return; }
+
+ if (p2==1) {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=ZERO; z[3]=ZERO; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; z[3]=ZERO; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; z[3]=X[1]; k=1;}
+ }
+ else if (p2==2) {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=X[2]; z[3]=ZERO; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; z[3]=X[2]; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; z[3]=X[1]; k=1;}
+ }
+ else {
+ if (EX==-42) {z[1]=X[1]+TWO10; z[2]=X[2]; k=3;}
+ else if (EX==-43) {z[1]= TWO10; z[2]=X[1]; k=2;}
+ else {z[1]= TWO10; z[2]=ZERO; k=1;}
+ z[3] = X[k];
+ }
+
+ u = (z[3] + TWO57) - TWO57;
+ if (u > z[3]) u -= TWO5;
+
+ if (u==z[3]) {
+ for (i=k+1; i <= p2; i++) {
+ if (X[i] == ZERO) continue;
+ else {z[3] += ONE; break; }
+ }
+ }
+
+ c = X[0]*((z[1] + R*(z[2] + R*z[3])) - TWO10);
+
+ *y = c*TWOM1032;
+ return;
+
+#undef R
+}
+
+/* Convert a multiple precision number *x into a double precision number *y. */
+/* The result is correctly rounded to the nearest/even. *x is left unchanged */
+
+void __mp_dbl(const mp_no *x, double *y, int p) {
+#if 0
+ int i,k;
+ double a,c,u,v,z[5];
+#endif
+
+ if (X[0] == ZERO) {*y = ZERO; return; }
+
+ if (EX> -42) norm(x,y,p);
+ else if (EX==-42 && X[1]>=TWO10) norm(x,y,p);
+ else denorm(x,y,p);
+}
+
+
+/* dbl_mp() converts a double precision number x into a multiple precision */
+/* number *y. If the precision p is too small the result is truncated. x is */
+/* left unchanged. */
+
+void __dbl_mp(double x, mp_no *y, int p) {
+
+ long i,n;
+ long p2 = p;
+ double u;
+
+ /* Sign */
+ if (x == ZERO) {Y[0] = ZERO; return; }
+ else if (x > ZERO) Y[0] = ONE;
+ else {Y[0] = MONE; x=-x; }
+
+ /* Exponent */
+ for (EY=ONE; x >= RADIX; EY += ONE) x *= RADIXI;
+ for ( ; x < ONE; EY -= ONE) x *= RADIX;
+
+ /* Digits */
+ n=MIN(p2,4);
+ for (i=1; i<=n; i++) {
+ u = (x + TWO52) - TWO52;
+ if (u>x) u -= ONE;
+ Y[i] = u; x -= u; x *= RADIX; }
+ for ( ; i<=p2; i++) Y[i] = ZERO;
+ return;
+}
+
+
+/* add_magnitudes() adds the magnitudes of *x & *y assuming that */
+/* abs(*x) >= abs(*y) > 0. */
+/* The sign of the sum *z is undefined. x&y may overlap but not x&z or y&z. */
+/* No guard digit is used. The result equals the exact sum, truncated. */
+/* *x & *y are left unchanged. */
+
+static void add_magnitudes(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i,j,k;
+ long p2 = p;
+
+ EZ = EX;
+
+ i=p2; j=p2+ EY - EX; k=p2+1;
+
+ if (j<1)
+ {__cpy(x,z,p); return; }
+ else Z[k] = ZERO;
+
+ for (; j>0; i--,j--) {
+ Z[k] += X[i] + Y[j];
+ if (Z[k] >= RADIX) {
+ Z[k] -= RADIX;
+ Z[--k] = ONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (; i>0; i--) {
+ Z[k] += X[i];
+ if (Z[k] >= RADIX) {
+ Z[k] -= RADIX;
+ Z[--k] = ONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ if (Z[1] == ZERO) {
+ for (i=1; i<=p2; i++) Z[i] = Z[i+1]; }
+ else EZ += ONE;
+}
+
+
+/* sub_magnitudes() subtracts the magnitudes of *x & *y assuming that */
+/* abs(*x) > abs(*y) > 0. */
+/* The sign of the difference *z is undefined. x&y may overlap but not x&z */
+/* or y&z. One guard digit is used. The error is less than one ulp. */
+/* *x & *y are left unchanged. */
+
+static void sub_magnitudes(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i,j,k;
+ long p2 = p;
+
+ EZ = EX;
+
+ if (EX == EY) {
+ i=j=k=p2;
+ Z[k] = Z[k+1] = ZERO; }
+ else {
+ j= EX - EY;
+ if (j > p2) {__cpy(x,z,p); return; }
+ else {
+ i=p2; j=p2+1-j; k=p2;
+ if (Y[j] > ZERO) {
+ Z[k+1] = RADIX - Y[j--];
+ Z[k] = MONE; }
+ else {
+ Z[k+1] = ZERO;
+ Z[k] = ZERO; j--;}
+ }
+ }
+
+ for (; j>0; i--,j--) {
+ Z[k] += (X[i] - Y[j]);
+ if (Z[k] < ZERO) {
+ Z[k] += RADIX;
+ Z[--k] = MONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (; i>0; i--) {
+ Z[k] += X[i];
+ if (Z[k] < ZERO) {
+ Z[k] += RADIX;
+ Z[--k] = MONE; }
+ else
+ Z[--k] = ZERO;
+ }
+
+ for (i=1; Z[i] == ZERO; i++) ;
+ EZ = EZ - i + 1;
+ for (k=1; i <= p2+1; )
+ Z[k++] = Z[i++];
+ for (; k <= p2; )
+ Z[k++] = ZERO;
+
+ return;
+}
+
+
+/* Add two multiple precision numbers. Set *z = *x + *y. x&y may overlap */
+/* but not x&z or y&z. One guard digit is used. The error is less than */
+/* one ulp. *x & *y are left unchanged. */
+
+void __add(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ int n;
+
+ if (X[0] == ZERO) {__cpy(y,z,p); return; }
+ else if (Y[0] == ZERO) {__cpy(x,z,p); return; }
+
+ if (X[0] == Y[0]) {
+ if (__acr(x,y,p) > 0) {add_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else {add_magnitudes(y,x,z,p); Z[0] = Y[0]; }
+ }
+ else {
+ if ((n=__acr(x,y,p)) == 1) {sub_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else if (n == -1) {sub_magnitudes(y,x,z,p); Z[0] = Y[0]; }
+ else Z[0] = ZERO;
+ }
+ return;
+}
+
+
+/* Subtract two multiple precision numbers. *z is set to *x - *y. x&y may */
+/* overlap but not x&z or y&z. One guard digit is used. The error is */
+/* less than one ulp. *x & *y are left unchanged. */
+
+void __sub(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ int n;
+
+ if (X[0] == ZERO) {__cpy(y,z,p); Z[0] = -Z[0]; return; }
+ else if (Y[0] == ZERO) {__cpy(x,z,p); return; }
+
+ if (X[0] != Y[0]) {
+ if (__acr(x,y,p) > 0) {add_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else {add_magnitudes(y,x,z,p); Z[0] = -Y[0]; }
+ }
+ else {
+ if ((n=__acr(x,y,p)) == 1) {sub_magnitudes(x,y,z,p); Z[0] = X[0]; }
+ else if (n == -1) {sub_magnitudes(y,x,z,p); Z[0] = -Y[0]; }
+ else Z[0] = ZERO;
+ }
+ return;
+}
+
+
+/* Multiply two multiple precision numbers. *z is set to *x * *y. x&y */
+/* may overlap but not x&z or y&z. In case p=1,2,3 the exact result is */
+/* truncated to p digits. In case p>3 the error is bounded by 1.001 ulp. */
+/* *x & *y are left unchanged. */
+
+void __mul(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ long i, i1, i2, j, k, k2;
+ long p2 = p;
+ double u, zk, zk2;
+
+ /* Is z=0? */
+ if (X[0]*Y[0]==ZERO)
+ { Z[0]=ZERO; return; }
+
+ /* Multiply, add and carry */
+ k2 = (p2<3) ? p2+p2 : p2+3;
+ zk = Z[k2]=ZERO;
+ for (k=k2; k>1; ) {
+ if (k > p2) {i1=k-p2; i2=p2+1; }
+ else {i1=1; i2=k; }
+#if 1
+ /* rearange this inner loop to allow the fmadd instructions to be
+ independent and execute in parallel on processors that have
+ dual symetrical FP pipelines. */
+ if (i1 < (i2-1))
+ {
+ /* make sure we have at least 2 iterations */
+ if (((i2 - i1) & 1L) == 1L)
+ {
+ /* Handle the odd iterations case. */
+ zk2 = x->d[i2-1]*y->d[i1];
+ }
+ else
+ zk2 = zero.d;
+ /* Do two multiply/adds per loop iteration, using independent
+ accumulators; zk and zk2. */
+ for (i=i1,j=i2-1; i<i2-1; i+=2,j-=2)
+ {
+ zk += x->d[i]*y->d[j];
+ zk2 += x->d[i+1]*y->d[j-1];
+ }
+ zk += zk2; /* final sum. */
+ }
+ else
+ {
+ /* Special case when iterations is 1. */
+ zk += x->d[i1]*y->d[i1];
+ }
+#else
+ /* The orginal code. */
+ for (i=i1,j=i2-1; i<i2; i++,j--) zk += X[i]*Y[j];
+#endif
+
+ u = (zk + CUTTER)-CUTTER;
+ if (u > zk) u -= RADIX;
+ Z[k] = zk - u;
+ zk = u*RADIXI;
+ --k;
+ }
+ Z[k] = zk;
+
+ /* Is there a carry beyond the most significant digit? */
+ if (Z[1] == ZERO) {
+ for (i=1; i<=p2; i++) Z[i]=Z[i+1];
+ EZ = EX + EY - 1; }
+ else
+ EZ = EX + EY;
+
+ Z[0] = X[0] * Y[0];
+ return;
+}
+
+
+/* Invert a multiple precision number. Set *y = 1 / *x. */
+/* Relative error bound = 1.001*r**(1-p) for p=2, 1.063*r**(1-p) for p=3, */
+/* 2.001*r**(1-p) for p>3. */
+/* *x=0 is not permissible. *x is left unchanged. */
+
+void __inv(const mp_no *x, mp_no *y, int p) {
+ long i;
+#if 0
+ int l;
+#endif
+ double t;
+ mp_no z,w;
+ static const int np1[] = {0,0,0,0,1,2,2,2,2,3,3,3,3,3,3,3,3,3,
+ 4,4,4,4,4,4,4,4,4,4,4,4,4,4,4};
+ const mp_no mptwo = {1,{1.0,2.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,
+ 0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0}};
+
+ __cpy(x,&z,p); z.e=0; __mp_dbl(&z,&t,p);
+ t=ONE/t; __dbl_mp(t,y,p); EY -= EX;
+
+ for (i=0; i<np1[p]; i++) {
+ __cpy(y,&w,p);
+ __mul(x,&w,y,p);
+ __sub(&mptwo,y,&z,p);
+ __mul(&w,&z,y,p);
+ }
+ return;
+}
+
+
+/* Divide one multiple precision number by another.Set *z = *x / *y. *x & *y */
+/* are left unchanged. x&y may overlap but not x&z or y&z. */
+/* Relative error bound = 2.001*r**(1-p) for p=2, 2.063*r**(1-p) for p=3 */
+/* and 3.001*r**(1-p) for p>3. *y=0 is not permissible. */
+
+void __dvd(const mp_no *x, const mp_no *y, mp_no *z, int p) {
+
+ mp_no w;
+
+ if (X[0] == ZERO) Z[0] = ZERO;
+ else {__inv(y,&w,p); __mul(x,&w,z,p);}
+ return;
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/slowexp.c libc25/sysdeps/powerpc/powerpc64/power4/fpu/slowexp.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/slowexp.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/slowexp.c Mon May 21 11:44:09 2007
@@ -0,0 +1,66 @@
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2007 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/**************************************************************************/
+/* MODULE_NAME:slowexp.c */
+/* */
+/* FUNCTION:slowexp */
+/* */
+/* FILES NEEDED:mpa.h */
+/* mpa.c mpexp.c */
+/* */
+/*Converting from double precision to Multi-precision and calculating */
+/* e^x */
+/**************************************************************************/
+#include "math_private.h"
+
+#ifdef NO_LONG_DOUBLE
+#include "mpa.h"
+void __mpexp(mp_no *x, mp_no *y, int p);
+#endif
+
+/*Converting from double precision to Multi-precision and calculating e^x */
+double __slowexp(double x) {
+#ifdef NO_LONG_DOUBLE
+ double w,z,res,eps=3.0e-26;
+ int p;
+ mp_no mpx, mpy, mpz,mpw,mpeps,mpcor;
+
+ p=6;
+ __dbl_mp(x,&mpx,p); /* Convert a double precision number x */
+ /* into a multiple precision number mpx with prec. p. */
+ __mpexp(&mpx, &mpy, p); /* Multi-Precision exponential function */
+ __dbl_mp(eps,&mpeps,p);
+ __mul(&mpeps,&mpy,&mpcor,p);
+ __add(&mpy,&mpcor,&mpw,p);
+ __sub(&mpy,&mpcor,&mpz,p);
+ __mp_dbl(&mpw, &w, p);
+ __mp_dbl(&mpz, &z, p);
+ if (w == z) return w;
+ else { /* if calculating is not exactly */
+ p = 32;
+ __dbl_mp(x,&mpx,p);
+ __mpexp(&mpx, &mpy, p);
+ __mp_dbl(&mpy, &res, p);
+ return res;
+ }
+#else
+ return (double) __ieee754_expl((long double)x);
+#endif
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/slowpow.c libc25/sysdeps/powerpc/powerpc64/power4/fpu/slowpow.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/slowpow.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/slowpow.c Fri Oct 20 15:24:56 2006
@@ -0,0 +1,94 @@
+/*
+ * IBM Accurate Mathematical Library
+ * written by International Business Machines Corp.
+ * Copyright (C) 2001, 2006 Free Software Foundation
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+/*************************************************************************/
+/* MODULE_NAME:slowpow.c */
+/* */
+/* FUNCTION:slowpow */
+/* */
+/*FILES NEEDED:mpa.h */
+/* mpa.c mpexp.c mplog.c halfulp.c */
+/* */
+/* Given two IEEE double machine numbers y,x , routine computes the */
+/* correctly rounded (to nearest) value of x^y. Result calculated by */
+/* multiplication (in halfulp.c) or if result isn't accurate enough */
+/* then routine converts x and y into multi-precision doubles and */
+/* recompute. */
+/*************************************************************************/
+
+#include "mpa.h"
+#include "math_private.h"
+
+void __mpexp (mp_no * x, mp_no * y, int p);
+void __mplog (mp_no * x, mp_no * y, int p);
+double ulog (double);
+double __halfulp (double x, double y);
+
+double
+__slowpow (double x, double y, double z)
+{
+ double res, res1;
+ long double ldw, ldz, ldpp;
+ static const long double ldeps = 0x4.0p-96;
+
+ res = __halfulp (x, y); /* halfulp() returns -10 or x^y */
+ if (res >= 0)
+ return res; /* if result was really computed by halfulp */
+ /* else, if result was not really computed by halfulp */
+
+ /* Compute pow as long double, 106 bits */
+ ldz = __ieee754_logl ((long double) x);
+ ldw = (long double) y *ldz;
+ ldpp = __ieee754_expl (ldw);
+ res = (double) (ldpp + ldeps);
+ res1 = (double) (ldpp - ldeps);
+
+ if (res != res1) /* if result still not accurate enough */
+ { /* use mpa for higher persision. */
+ mp_no mpx, mpy, mpz, mpw, mpp, mpr, mpr1;
+ static const mp_no eps = { -3, {1.0, 4.0} };
+ int p;
+
+ p = 10; /* p=precision 240 bits */
+ __dbl_mp (x, &mpx, p);
+ __dbl_mp (y, &mpy, p);
+ __dbl_mp (z, &mpz, p);
+ __mplog (&mpx, &mpz, p); /* log(x) = z */
+ __mul (&mpy, &mpz, &mpw, p); /* y * z =w */
+ __mpexp (&mpw, &mpp, p); /* e^w =pp */
+ __add (&mpp, &eps, &mpr, p); /* pp+eps =r */
+ __mp_dbl (&mpr, &res, p);
+ __sub (&mpp, &eps, &mpr1, p); /* pp -eps =r1 */
+ __mp_dbl (&mpr1, &res1, p); /* converting into double precision */
+ if (res == res1)
+ return res;
+
+ /* if we get here result wasn't calculated exactly, continue for
+ more exact calculation using 768 bits. */
+ p = 32;
+ __dbl_mp (x, &mpx, p);
+ __dbl_mp (y, &mpy, p);
+ __dbl_mp (z, &mpz, p);
+ __mplog (&mpx, &mpz, p); /* log(c)=z */
+ __mul (&mpy, &mpz, &mpw, p); /* y*z =w */
+ __mpexp (&mpw, &mpp, p); /* e^w=pp */
+ __mp_dbl (&mpp, &res, p); /* converting into double precision */
+ }
+ return res;
+}
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrt.c libc25/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrt.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrt.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrt.c Fri Jun 01 10:18:46 2007
@@ -0,0 +1,62 @@
+/* Double-precision floating point square root wrapper.
+ Copyright (C) 2004, 2007 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_ldbl_opt.h>
+#include "math.h"
+#include "math_private.h"
+#include <fenv_libc.h>
+
+#ifdef __STDC__
+double
+__sqrt (double x) /* wrapper sqrt */
+#else
+double
+__sqrt (x) /* wrapper sqrt */
+ double x;
+#endif
+{
+ double z;
+/* Power4 (ISA V2.0) and above implement sqrt in hardware. */
+ __asm __volatile (
+ " fsqrt %0,%1\n"
+ : "=f" (z)
+ : "f" (x));
+#ifdef _IEEE_LIBM
+ return z;
+#else
+ if (__builtin_expect (_LIB_VERSION == _IEEE_, 0))
+ return z;
+
+ if (__builtin_expect (x != x, 0))
+ return z;
+
+ if (__builtin_expect (x < 0.0, 0))
+ return __kernel_standard (x, x, 26); /* sqrt(negative) */
+ else
+ return z;
+#endif
+}
+
+weak_alias (__sqrt, sqrt)
+#ifdef NO_LONG_DOUBLE
+ strong_alias (__sqrt, __sqrtl) weak_alias (__sqrt, sqrtl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __sqrt, sqrtl, GLIBC_2_0);
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrtf.c libc25/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrtf.c
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrtf.c Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power4/fpu/w_sqrtf.c Fri Jun 01 10:23:54 2007
@@ -0,0 +1,60 @@
+/* Single-precision floating point square root wrapper.
+ Copyright (C) 2004, 2007 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 <sysdep.h>
+#include <ldsodefs.h>
+
+#ifdef __STDC__
+float
+__sqrtf (float x) /* wrapper sqrtf */
+#else
+float
+__sqrtf (x) /* wrapper sqrtf */
+ float x;
+#endif
+{
+#ifdef _IEEE_LIBM
+ return __ieee754_sqrtf (x);
+#else
+ float z;
+/* Power4 (ISA V2.0) and above implement sqrtf in hardware. */
+ __asm __volatile (
+ " fsqrts %0,%1\n"
+ : "=f" (z)
+ : "f" (x));
+
+ if (__builtin_expect (_LIB_VERSION == _IEEE_, 0))
+ return z;
+
+ if (__builtin_expect (x != x, 0))
+ return z;
+
+ if (__builtin_expect (x < 0.0, 0))
+ /* sqrtf(negative) */
+ return (float) __kernel_standard ((double) x, (double) x, 126);
+ else
+ return z;
+#endif
+}
+
+weak_alias (__sqrtf, sqrtf)
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/Implies libc25/sysdeps/powerpc/powerpc64/power5+/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/Implies Sun Aug 20 09:05:08 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceil.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceil.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceil.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceil.S Thu Jun 15 11:45:17 2006
@@ -0,0 +1,38 @@
+/* ceil function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__ceil, 4, 0)
+ CALL_MCOUNT 0
+ frip fp1, fp1
+ blr
+ END (__ceil)
+
+weak_alias (__ceil, ceil)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__ceil, ceill)
+strong_alias (__ceil, __ceill)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __ceil, ceill, GLIBC_2_0)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceilf.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceilf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceilf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_ceilf.S Thu Jun 15 11:45:38 2006
@@ -0,0 +1,31 @@
+/* ceilf function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__ceilf, 4, 0)
+ CALL_MCOUNT 0
+ frip fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__ceilf)
+
+weak_alias (__ceilf, ceilf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_floor.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_floor.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_floor.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_floor.S Thu Jun 15 11:45:07 2006
@@ -0,0 +1,38 @@
+/* floor function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__floor, 4, 0)
+ CALL_MCOUNT 0
+ frim fp1, fp1
+ blr
+ END (__floor)
+
+weak_alias (__floor, floor)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__floor, floorl)
+strong_alias (__floor, __floorl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_0)
+compat_symbol (libm, __floor, floorl, GLIBC_2_0)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_floorf.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_floorf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_floorf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_floorf.S Thu Jun 15 11:44:52 2006
@@ -0,0 +1,31 @@
+/* floorf function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__floorf, 4, 0)
+ CALL_MCOUNT 0
+ frim fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__floorf)
+
+weak_alias (__floorf, floorf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_llround.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_llround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_llround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_llround.S Mon Oct 09 14:00:22 2006
@@ -0,0 +1,59 @@
+/* llround function. POWER5+, PowerPC64 version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long long [r3] llround (float x [fp1])
+ IEEE 1003.1 llround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power5"
+EALIGN (__llround, 4, 0)
+ CALL_MCOUNT 0
+ frin fp2, fp1 /* Round to nearest +-0.5. */
+ fctidz fp3, fp2 /* Convert To Integer DW round toward 0. */
+ stfd fp3, -16(r1)
+ nop /* Insure the following load is in a different dispatch group */
+ nop /* to avoid pipe stall on POWER4&5. */
+ nop
+ ld r3, -16(r1)
+ blr
+ END (__llround)
+
+strong_alias (__llround, __lround)
+weak_alias (__llround, llround)
+weak_alias (__lround, lround)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__llround, llroundl)
+strong_alias (__llround, __llroundl)
+weak_alias (__lround, lroundl)
+strong_alias (__lround, __lroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llround, llroundl, GLIBC_2_1)
+compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_round.S Tue May 22 17:56:44 2007
@@ -0,0 +1,38 @@
+/* round function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__round, 4, 0)
+ CALL_MCOUNT 0
+ frin fp1, fp1
+ blr
+ END (__round)
+
+weak_alias (__round, round)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__round, roundl)
+strong_alias (__round, __roundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __round, roundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_roundf.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_roundf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_roundf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_roundf.S Thu Jun 15 11:43:56 2006
@@ -0,0 +1,31 @@
+/* roundf function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__roundf, 4, 0)
+ CALL_MCOUNT 0
+ frin fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__roundf)
+
+weak_alias (__roundf, roundf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_trunc.S Tue May 22 17:57:49 2007
@@ -0,0 +1,38 @@
+/* trunc function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power5"
+EALIGN (__trunc, 4, 0)
+ CALL_MCOUNT 0
+ friz fp1, fp1
+ blr
+ END (__trunc)
+
+weak_alias (__trunc, trunc)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__trunc, truncl)
+strong_alias (__trunc, __truncl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __trunc, truncl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_truncf.S libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_truncf.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5+/fpu/s_truncf.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5+/fpu/s_truncf.S Thu Jun 15 11:36:15 2006
@@ -0,0 +1,31 @@
+/* truncf function. PowerPC64/power5+ version.
+ Copyright (C) 2006 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 <sysdep.h>
+
+ .machine "power5"
+EALIGN (__truncf, 4, 0)
+ CALL_MCOUNT 0
+ friz fp1, fp1 /* The rounding instructions are double. */
+ frsp fp1, fp1 /* But we need to set ooverflow for float. */
+ blr
+ END (__truncf)
+
+weak_alias (__truncf, truncf)
+
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5/fpu/Implies libc25/sysdeps/powerpc/powerpc64/power5/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power5/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power5/fpu/Implies Thu Nov 09 14:12:01 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6/fpu/Implies libc25/sysdeps/powerpc/powerpc64/power6/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power6/fpu/Implies Sun Aug 20 09:06:40 2006
@@ -0,0 +1,2 @@
+powerpc/powerpc64/power5+/fpu
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/Implies libc25/sysdeps/powerpc/powerpc64/power6x/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power6x/fpu/Implies Mon Oct 09 14:00:22 2006
@@ -0,0 +1,3 @@
+powerpc/powerpc64/power6/fpu
+powerpc/powerpc64/power5+/fpu
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/s_llrint.S libc25/sysdeps/powerpc/powerpc64/power6x/fpu/s_llrint.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/s_llrint.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power6x/fpu/s_llrint.S Mon Oct 09 14:00:56 2006
@@ -0,0 +1,45 @@
+/* Round double to long int. POWER6x PowerPC64 version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+ .machine "power6"
+/* long long int[r3] __llrint (double x[fp1]) */
+ENTRY (__llrint)
+ CALL_MCOUNT 0
+ fctid fp13,fp1
+ mftgpr r3,fp13
+ blr
+ END (__llrint)
+
+strong_alias (__llrint, __lrint)
+weak_alias (__llrint, llrint)
+weak_alias (__lrint, lrint)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__llrint, __llrintl)
+weak_alias (__llrint, llrintl)
+strong_alias (__lrint, __lrintl)
+weak_alias (__lrint, lrintl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llrint, llrintl, GLIBC_2_1)
+compat_symbol (libm, __lrint, lrintl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/s_llround.S libc25/sysdeps/powerpc/powerpc64/power6x/fpu/s_llround.S
--- libc25-cvstip-20070510/sysdeps/powerpc/powerpc64/power6x/fpu/s_llround.S Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/powerpc/powerpc64/power6x/fpu/s_llround.S Mon Oct 09 14:00:56 2006
@@ -0,0 +1,55 @@
+/* llround function. POWER6x PowerPC64 version.
+ Copyright (C) 2006 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 <sysdep.h>
+#include <math_ldbl_opt.h>
+
+/* long long [r3] llround (float x [fp1])
+ IEEE 1003.1 llround function. IEEE specifies "round to the nearest
+ integer value, rounding halfway cases away from zero, regardless of
+ the current rounding mode." However PowerPC Architecture defines
+ "round to Nearest" as "Choose the best approximation. In case of a
+ tie, choose the one that is even (least significant bit o).".
+ So we pre-round using the V2.02 Floating Round to Integer Nearest
+ instruction before we use Floating Convert to Integer Word with
+ round to zero instruction. */
+
+ .machine "power6"
+ENTRY (__llround)
+ CALL_MCOUNT 0
+ frin fp2,fp1 /* Round to nearest +-0.5. */
+ fctidz fp3,fp2 /* Convert To Integer DW round toward 0. */
+ mftgpr r3,fp3 /* Transfer integer to R3. */
+ blr
+ END (__llround)
+
+strong_alias (__llround, __lround)
+weak_alias (__llround, llround)
+weak_alias (__lround, lround)
+
+#ifdef NO_LONG_DOUBLE
+weak_alias (__llround, llroundl)
+strong_alias (__llround, __llroundl)
+weak_alias (__lround, lroundl)
+strong_alias (__lround, __lroundl)
+#endif
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __llround, llroundl, GLIBC_2_1)
+compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
+#endif
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/970/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/970/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/970/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/970/fpu/Implies Thu Apr 06 10:50:35 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/970/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power4/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power4/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power4/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power4/fpu/Implies Thu Apr 06 10:50:35 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5+/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5+/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5+/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5+/fpu/Implies Thu Apr 06 10:50:35 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/power5+/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power5/fpu/Implies Thu Apr 06 10:50:35 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/power5/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6/fpu/Implies Mon Jun 19 15:25:23 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/power6/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6x/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6x/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6x/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc32/power6x/fpu/Implies Mon Oct 09 14:00:22 2006
@@ -0,0 +1,4 @@
+# Make sure this comes before the powerpc/powerpc32/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc32/fpu/Implies.
+powerpc/powerpc32/power6x/fpu
+powerpc/powerpc32/power6/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/970/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/970/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/970/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/970/fpu/Implies Sun Aug 20 09:13:50 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power4/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power4/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power4/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power4/fpu/Implies Sun Aug 20 09:14:01 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5+/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5+/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5+/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5+/fpu/Implies Thu Jun 15 08:44:40 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc64/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc64/fpu/Implies.
+powerpc/powerpc64/power5+/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power5/fpu/Implies Sun Aug 20 09:14:05 2006
@@ -0,0 +1,1 @@
+powerpc/powerpc64/power4/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6/fpu/Implies Mon Jun 19 14:51:14 2006
@@ -0,0 +1,3 @@
+# Make sure this comes before the powerpc/powerpc64/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc64/fpu/Implies.
+powerpc/powerpc64/power6/fpu
diff -urN libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6x/fpu/Implies libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6x/fpu/Implies
--- libc25-cvstip-20070510/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6x/fpu/Implies Wed Dec 31 18:00:00 1969
+++ libc25/sysdeps/unix/sysv/linux/powerpc/powerpc64/power6x/fpu/Implies Mon Oct 09 14:00:22 2006
@@ -0,0 +1,4 @@
+# Make sure this comes before the powerpc/powerpc64/fpu that's
+# listed in unix/sysv/linux/powerpc/powerpc64/fpu/Implies.
+powerpc/powerpc64/power6x/fpu
+powerpc/powerpc64/power6/fpu
prev parent reply other threads:[~2007-06-01 21:57 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <4653610A.1060101@us.ibm.com>
[not found] ` <20070522214043.GN4033@devserv.devel.redhat.com>
[not found] ` <46560D3F.80809@us.ibm.com>
[not found] ` <20070529213339.GJ4033@devserv.devel.redhat.com>
2007-06-01 21:55 ` [PATCH] Move powerpc-cpu string " Steven Munroe
2007-06-01 21:57 ` Steven Munroe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=46609AB8.8050405@us.ibm.com \
--to=munroesj@us.ibm.com \
--cc=jakub@redhat.com \
--cc=libc-hacker@sources.redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).