nptl/Makefile
Unbreak building of tst-key5
(stanshebs, google-local)
+
+elf/elf.h
+math/libm-test.inc
+string/*
+sysdeps/ieee754/*
+sysdeps/powerpc/*
+ For b/18010034, backport IBM optimizations for Power8.
+ https://sourceware.org/git/?p=glibc.git;a=shortlog;h=refs/heads/ibm/2.19/master
+ (stanshebs, backport)
#define DT_PPC64_OPD (DT_LOPROC + 1)
#define DT_PPC64_OPDSZ (DT_LOPROC + 2)
#define DT_PPC64_OPT (DT_LOPROC + 3)
-#define DT_PPC64_NUM 3
+#define DT_PPC64_NUM 4
/* PowerPC64 specific values for the DT_PPC64_OPT Dyn entry. */
#define PPC64_OPT_TLS 1
TEST_f_f (ceil, -72057594037927936.75L, -72057594037927936.0L),
TEST_f_f (ceil, -72057594037927937.5L, -72057594037927937.0L),
+ /* Check cases where first double is a exact integer higher than 2^52 and
+ the precision is determined by second long double for IBM long double. */
+ TEST_f_f (ceil, 34503599627370498.515625L, 34503599627370499.0L),
+ TEST_f_f (ceil, -34503599627370498.515625L, -34503599627370498.0L),
+# if LDBL_MANT_DIG >= 106
+ TEST_f_f (ceil, 1192568192774434123539907640624.484375L, 1192568192774434123539907640625.0L),
+ TEST_f_f (ceil, -1192568192774434123539907640624.484375L, -1192568192774434123539907640624.0L),
+# endif
+
TEST_f_f (ceil, 10141204801825835211973625643007.5L, 10141204801825835211973625643008.0L),
TEST_f_f (ceil, 10141204801825835211973625643008.25L, 10141204801825835211973625643009.0L),
TEST_f_f (ceil, 10141204801825835211973625643008.5L, 10141204801825835211973625643009.0L),
TEST_fI_f1 (frexp, 12.8L, 0.8L, 4, NO_INEXACT_EXCEPTION),
TEST_fI_f1 (frexp, -27.34L, -0.854375L, 5, NO_INEXACT_EXCEPTION),
+
+#if defined TEST_LDOUBLE && LDBL_MANT_DIG >= 106
+ TEST_fI_f1 (frexp, 1.0L-0x1p-106L, 1.0L-0x1p-106L, 0, NO_INEXACT_EXCEPTION),
+ TEST_fI_f1 (frexp, 1.0L, 0.5L, 1, NO_INEXACT_EXCEPTION),
+ TEST_fI_f1 (frexp, 1.0L+0x1p-105L, 0.5L+0x1p-106L, 1, NO_INEXACT_EXCEPTION),
+ TEST_fI_f1 (frexp, -1.0L+0x1p-106L, -1.0L+0x1p-106L, 0, NO_INEXACT_EXCEPTION),
+ TEST_fI_f1 (frexp, -1.0L, -0.5L, 1, NO_INEXACT_EXCEPTION),
+ TEST_fI_f1 (frexp, -1.0L-0x1p-105L, -0.5L-0x1p-106L, 1, NO_INEXACT_EXCEPTION),
+#endif
};
static void
TEST_f_f (nearbyint, -281474976710656.75, -281474976710657.0, NO_INEXACT_EXCEPTION),
TEST_f_f (nearbyint, -562949953421312.75, -562949953421313.0, NO_INEXACT_EXCEPTION),
TEST_f_f (nearbyint, -1125899906842624.75, -1125899906842625.0, NO_INEXACT_EXCEPTION),
+#endif
+#ifdef TEST_LDOUBLE
+ /* Check cases where first double is a exact integer higher than 2^52 and
+ the precision is determined by second long double for IBM long double. */
+ TEST_f_f (nearbyint, 34503599627370498.515625L, 34503599627370499.0L),
+ TEST_f_f (nearbyint, -34503599627370498.515625L, -34503599627370499.0L),
+# if LDBL_MANT_DIG >= 106
+ TEST_f_f (nearbyint, 1024.5000000000001L, 1025.0L, NO_INEXACT_EXCEPTION),
+ TEST_f_f (nearbyint, 1025.5000000000001L, 1026.0L, NO_INEXACT_EXCEPTION),
+ TEST_f_f (nearbyint, -1024.5000000000001L, -1025.0L, NO_INEXACT_EXCEPTION),
+ TEST_f_f (nearbyint, -1025.5000000000001L, -1026.0L, NO_INEXACT_EXCEPTION),
+ TEST_f_f (nearbyint, 1192568192774434123539907640624.484375L, 1192568192774434123539907640624.0L),
+ TEST_f_f (nearbyint, -1192568192774434123539907640624.484375L, -1192568192774434123539907640624.0L),
+# endif
#endif
};
// XXX Enable once gcc is fixed.
//TEST_ff_f (nextafter, 0x0.00000040000000000000p-16385L, -0.1L, 0x0.0000003ffffffff00000p-16385L),
#endif
+#if defined TEST_LDOUBLE && LDBL_MANT_DIG == 106
+ TEST_ff_f (nextafter, 1.0L, -10.0L, 1.0L-0x1p-106L, NO_EXCEPTION),
+ TEST_ff_f (nextafter, 1.0L, 10.0L, 1.0L+0x1p-105L, NO_EXCEPTION),
+ TEST_ff_f (nextafter, 1.0L-0x1p-106L, 10.0L, 1.0L, NO_EXCEPTION),
+ TEST_ff_f (nextafter, -1.0L, -10.0L, -1.0L-0x1p-105L, NO_EXCEPTION),
+ TEST_ff_f (nextafter, -1.0L, 10.0L, -1.0L+0x1p-106L, NO_EXCEPTION),
+ TEST_ff_f (nextafter, -1.0L+0x1p-106L, -10.0L, -1.0L, NO_EXCEPTION),
+#endif
/* XXX We need the hexadecimal FP number representation here for further
tests. */
TEST_f_f (rint, 4503599627370497.5L, 4503599627370498.0L, INEXACT_EXCEPTION),
# if LDBL_MANT_DIG > 100
+ TEST_f_f (rint, 1024.5000000000001L, 1025.0L, INEXACT_EXCEPTION),
+ TEST_f_f (rint, 1025.5000000000001L, 1026.0L, INEXACT_EXCEPTION),
+ TEST_f_f (rint, -1024.5000000000001L, -1025.0L, INEXACT_EXCEPTION),
+ TEST_f_f (rint, -1025.5000000000001L, -1026.0L, INEXACT_EXCEPTION),
TEST_f_f (rint, 4503599627370494.5000000000001L, 4503599627370495.0L, INEXACT_EXCEPTION),
TEST_f_f (rint, 4503599627370495.5000000000001L, 4503599627370496.0L, INEXACT_EXCEPTION),
TEST_f_f (rint, 4503599627370496.5000000000001L, 4503599627370497.0L, INEXACT_EXCEPTION),
TEST_f_f (round, -72057594037927936.75L, -72057594037927937.0L),
TEST_f_f (round, -72057594037927937.5L, -72057594037927938.0L),
+ /* Check cases where first double is a exact integer higher than 2^52 and
+ the precision is determined by second long double for IBM long double. */
+ TEST_f_f (round, 34503599627370498.515625L, 34503599627370499.0L),
+ TEST_f_f (round, -34503599627370498.515625L, -34503599627370499.0L),
+# if LDBL_MANT_DIG >= 106
+ TEST_f_f (round, 1192568192774434123539907640624.484375L, 1192568192774434123539907640624.0L),
+ TEST_f_f (round, -1192568192774434123539907640624.484375L, -1192568192774434123539907640624.0L),
+# endif
+
TEST_f_f (round, 10141204801825835211973625643007.5L, 10141204801825835211973625643008.0L),
TEST_f_f (round, 10141204801825835211973625643008.25L, 10141204801825835211973625643008.0L),
TEST_f_f (round, 10141204801825835211973625643008.5L, 10141204801825835211973625643009.0L),
#define a2 dest
#define a2const
-#include <memmove.c>
+#include <string/memmove.c>
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
-#if HAVE_CONFIG_H
-# include <config.h>
-#endif
-
-#if defined _LIBC || HAVE_STRING_H
-# include <string.h>
-#else
-# include <strings.h>
-# ifndef strchr
-# define strchr index
-# endif
-#endif
+#include <string.h>
#undef strcspn
+#ifndef STRCSPN
+# define STRCSPN strcspn
+#endif
+
/* Return the length of the maximum initial segment of S
which contains no characters from REJECT. */
size_t
-strcspn (s, reject)
- const char *s;
- const char *reject;
+STRCSPN (const char *s, const char *reject)
{
size_t count = 0;
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
-#ifdef HAVE_CONFIG_H
-# include <config.h>
-#endif
-
-#if defined _LIBC || defined HAVE_CONFIG_H
-# include <string.h>
-#endif
+#include <string.h>
#undef strpbrk
+#ifndef STRPBRK
+#define STRPBRK strpbrk
+#endif
+
/* Find the first occurrence in S of any character in ACCEPT. */
char *
-strpbrk (s, accept)
- const char *s;
- const char *accept;
+STRPBRK (const char *s, const char *accept)
{
while (*s != '\0')
{
#undef strrchr
+#ifndef STRRCHR
+# define STRRCHR strrchr
+#endif
+
/* Find the last occurrence of C in S. */
char *
-strrchr (const char *s, int c)
+STRRCHR (const char *s, int c)
{
const char *found, *p;
#include <string.h>
#undef strspn
+#ifndef STRSPN
+#define STRSPN strspn
+#endif
/* Return the length of the maximum initial segment
of S which contains only characters in ACCEPT. */
size_t
-strspn (s, accept)
- const char *s;
- const char *accept;
+STRSPN (const char *s, const char *accept)
{
const char *p;
const char *a;
# branred depends on precise IEEE double rounding
CFLAGS-branred.c = $(config-cflags-nofma)
CFLAGS-e_sqrt.c = $(config-cflags-nofma)
+CFLAGS-e_pow.c = $(config-cflags-nofma)
endif
#include <math_private.h>
#include <math_ldbl_opt.h>
-static const long double
-two107 = 162259276829213363391578010288128.0; /* 0x4670000000000000, 0 */
-
long double __frexpl(long double x, int *eptr)
{
- uint64_t hx, lx, ix, ixl;
- int64_t explo;
- double xhi, xlo;
+ uint64_t hx, lx, ix, ixl;
+ int64_t explo, expon;
+ double xhi, xlo;
+
+ ldbl_unpack (x, &xhi, &xlo);
+ EXTRACT_WORDS64 (hx, xhi);
+ EXTRACT_WORDS64 (lx, xlo);
+ ixl = 0x7fffffffffffffffULL & lx;
+ ix = 0x7fffffffffffffffULL & hx;
+ expon = 0;
+ if (ix >= 0x7ff0000000000000ULL || ix == 0)
+ {
+ /* 0,inf,nan. */
+ *eptr = expon;
+ return x;
+ }
+ expon = ix >> 52;
+ if (expon == 0)
+ {
+ /* Denormal high double, the low double must be 0.0. */
+ int cnt;
+
+ /* Normalize. */
+ if (sizeof (ix) == sizeof (long))
+ cnt = __builtin_clzl (ix);
+ else if ((ix >> 32) != 0)
+ cnt = __builtin_clzl ((long) (ix >> 32));
+ else
+ cnt = __builtin_clzl ((long) ix) + 32;
+ cnt = cnt - 12;
+ expon -= cnt;
+ ix <<= cnt + 1;
+ }
+ expon -= 1022;
+ ix &= 0x000fffffffffffffULL;
+ hx &= 0x8000000000000000ULL;
+ hx |= (1022LL << 52) | ix;
- ldbl_unpack (x, &xhi, &xlo);
- EXTRACT_WORDS64 (hx, xhi);
- EXTRACT_WORDS64 (lx, xlo);
- ixl = 0x7fffffffffffffffULL&lx;
- ix = 0x7fffffffffffffffULL&hx;
- *eptr = 0;
- if(ix>=0x7ff0000000000000ULL||ix==0) return x; /* 0,inf,nan */
- if (ix<0x0010000000000000ULL) { /* subnormal */
- x *= two107;
- xhi = ldbl_high (x);
- EXTRACT_WORDS64 (hx, xhi);
- ix = hx&0x7fffffffffffffffULL;
- *eptr = -107;
+ if (ixl != 0)
+ {
+ /* If the high double is an exact power of two and the low
+ double has the opposite sign, then the exponent calculated
+ from the high double is one too big. */
+ if (ix == 0
+ && (int64_t) (hx ^ lx) < 0)
+ {
+ hx += 1L << 52;
+ expon -= 1;
}
- *eptr += (ix>>52)-1022;
- if (ixl != 0ULL) {
- explo = (ixl>>52) - (ix>>52) + 0x3fe;
- if ((ixl&0x7ff0000000000000ULL) == 0LL) {
- /* the lower double is a denormal so we need to correct its
- mantissa and perhaps its exponent. */
- int cnt;
+ explo = ixl >> 52;
+ if (explo == 0)
+ {
+ /* The low double started out as a denormal. Normalize its
+ mantissa and adjust the exponent. */
+ int cnt;
- if (sizeof (ixl) == sizeof (long))
- cnt = __builtin_clzl (ixl);
- else if ((ixl >> 32) != 0)
- cnt = __builtin_clzl ((long) (ixl >> 32));
- else
- cnt = __builtin_clzl ((long) ixl) + 32;
- cnt = cnt - 12;
- lx = (lx&0x8000000000000000ULL) | ((explo-cnt)<<52)
- | ((ixl<<(cnt+1))&0x000fffffffffffffULL);
- } else
- lx = (lx&0x800fffffffffffffULL) | (explo<<52);
- } else
- lx = 0ULL;
+ if (sizeof (ixl) == sizeof (long))
+ cnt = __builtin_clzl (ixl);
+ else if ((ixl >> 32) != 0)
+ cnt = __builtin_clzl ((long) (ixl >> 32));
+ else
+ cnt = __builtin_clzl ((long) ixl) + 32;
+ cnt = cnt - 12;
+ explo -= cnt;
+ ixl <<= cnt + 1;
+ }
+
+ /* With variable precision we can't assume much about the
+ magnitude of the returned low double. It may even be a
+ denormal. */
+ explo -= expon;
+ ixl &= 0x000fffffffffffffULL;
+ lx &= 0x8000000000000000ULL;
+ if (explo <= 0)
+ {
+ /* Handle denormal low double. */
+ if (explo > -52)
+ {
+ ixl |= 1LL << 52;
+ ixl >>= 1 - explo;
+ }
+ else
+ {
+ ixl = 0;
+ lx = 0;
+ if ((hx & 0x7ff0000000000000ULL) == (1023LL << 52))
+ {
+ /* Oops, the adjustment we made above for values a
+ little smaller than powers of two turned out to
+ be wrong since the returned low double will be
+ zero. This can happen if the input was
+ something weird like 0x1p1000 - 0x1p-1000. */
+ hx -= 1L << 52;
+ expon += 1;
+ }
+ }
+ explo = 0;
+ }
+ lx |= (explo << 52) | ixl;
+ }
- hx = (hx&0x800fffffffffffffULL) | 0x3fe0000000000000ULL;
- INSERT_WORDS64 (xhi, hx);
- INSERT_WORDS64 (xlo, lx);
- x = ldbl_pack (xhi, xlo);
- return x;
+ INSERT_WORDS64 (xhi, hx);
+ INSERT_WORDS64 (xlo, lx);
+ x = ldbl_pack (xhi, xlo);
+ *eptr = expon;
+ return x;
}
#ifdef IS_IN_libm
long_double_symbol (libm, __frexpl, frexpl);
if (fabs (u.d[0].d) < TWO52)
{
+ double xh = u.d[0].d;
double high = u.d[0].d;
feholdexcept (&env);
if (high > 0.0)
high += TWO52;
if (high == 0.0) high = -0.0;
}
+ if (u.d[1].d > 0.0 && (xh - high == 0.5))
+ high += 1.0;
+ else if (u.d[1].d < 0.0 && (-(xh - high) == 0.5))
+ high -= 1.0;
u.d[0].d = high;
u.d[1].d = 0.0;
math_force_eval (u.d[0]);
long double __nextafterl(long double x, long double y)
{
- int64_t hx,hy,ihx,ihy;
- uint64_t lx;
+ int64_t hx, hy, ihx, ihy, lx;
double xhi, xlo, yhi;
ldbl_unpack (x, &xhi, &xlo);
u = math_opt_barrier (x);
x -= __LDBL_DENORM_MIN__;
if (ihx < 0x0360000000000000LL
- || (hx > 0 && (int64_t) lx <= 0)
- || (hx < 0 && (int64_t) lx > 1)) {
+ || (hx > 0 && lx <= 0)
+ || (hx < 0 && lx > 1)) {
u = u * u;
math_force_eval (u); /* raise underflow flag */
}
return x;
}
- if (ihx < 0x06a0000000000000LL) { /* ulp will denormal */
- INSERT_WORDS64 (yhi, hx & (0x7ffLL<<52));
- u = yhi;
- u *= 0x1.0000000000000p-105L;
+ /* If the high double is an exact power of two and the low
+ double is the opposite sign, then 1ulp is one less than
+ what we might determine from the high double. Similarly
+ if X is an exact power of two, and positive, because
+ making it a little smaller will result in the exponent
+ decreasing by one and normalisation of the mantissa. */
+ if ((hx & 0x000fffffffffffffLL) == 0
+ && ((lx != 0 && (hx ^ lx) < 0)
+ || (lx == 0 && hx >= 0)))
+ ihx -= 1LL << 52;
+ if (ihx < (106LL << 52)) { /* ulp will denormal */
+ INSERT_WORDS64 (yhi, ihx & (0x7ffLL<<52));
+ u = yhi * 0x1p-105;
} else {
- INSERT_WORDS64 (yhi, (hx & (0x7ffLL<<52))-(0x069LL<<52));
+ INSERT_WORDS64 (yhi, (ihx & (0x7ffLL<<52))-(105LL<<52));
u = yhi;
}
return x - u;
u = math_opt_barrier (x);
x += __LDBL_DENORM_MIN__;
if (ihx < 0x0360000000000000LL
- || (hx > 0 && (int64_t) lx < 0 && lx != 0x8000000000000001LL)
- || (hx < 0 && (int64_t) lx >= 0)) {
+ || (hx > 0 && lx < 0 && lx != 0x8000000000000001LL)
+ || (hx < 0 && lx >= 0)) {
u = u * u;
math_force_eval (u); /* raise underflow flag */
}
x = -0.0L;
return x;
}
- if (ihx < 0x06a0000000000000LL) { /* ulp will denormal */
- INSERT_WORDS64 (yhi, hx & (0x7ffLL<<52));
- u = yhi;
- u *= 0x1.0000000000000p-105L;
+ /* If the high double is an exact power of two and the low
+ double is the opposite sign, then 1ulp is one less than
+ what we might determine from the high double. Similarly
+ if X is an exact power of two, and negative, because
+ making it a little larger will result in the exponent
+ decreasing by one and normalisation of the mantissa. */
+ if ((hx & 0x000fffffffffffffLL) == 0
+ && ((lx != 0 && (hx ^ lx) < 0)
+ || (lx == 0 && hx < 0)))
+ ihx -= 1LL << 52;
+ if (ihx < (106LL << 52)) { /* ulp will denormal */
+ INSERT_WORDS64 (yhi, ihx & (0x7ffLL<<52));
+ u = yhi * 0x1p-105;
} else {
- INSERT_WORDS64 (yhi, (hx & (0x7ffLL<<52))-(0x069LL<<52));
+ INSERT_WORDS64 (yhi, (ihx & (0x7ffLL<<52))-(105LL<<52));
u = yhi;
}
return x + u;
#define PPC_FEATURE2_HAS_EBB 0x10000000 /* Event Base Branching */
#define PPC_FEATURE2_HAS_ISEL 0x08000000 /* Integer Select */
#define PPC_FEATURE2_HAS_TAR 0x04000000 /* Target Address Register */
+#define PPC_FEATURE2_HAS_VEC_CRYPTO 0x02000000 /* Target supports vector
+ instruction. */
+++ /dev/null
-/* Copy memory to memory until the specified number of bytes
- has been copied. Overlap is handled correctly.
- Copyright (C) 1991-2014 Free Software Foundation, Inc.
- This file is part of the GNU C Library.
- Contributed by Torbjorn Granlund (tege@sics.se).
-
- 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; see the file COPYING.LIB. If
- not, see <http://www.gnu.org/licenses/>. */
-
-#include <string.h>
-#include <memcopy.h>
-#include <pagecopy.h>
-
-/* All this is so that bcopy.c can #include
- this file after defining some things. */
-#ifndef a1
-#define a1 dest /* First arg is DEST. */
-#define a1const
-#define a2 src /* Second arg is SRC. */
-#define a2const const
-#undef memmove
-#endif
-#if !defined(RETURN) || !defined(rettype)
-#define RETURN(s) return (s) /* Return DEST. */
-#define rettype void *
-#endif
-
-#ifndef MEMMOVE
-#define MEMMOVE memmove
-#endif
-
-rettype
-MEMMOVE (a1, a2, len)
- a1const void *a1;
- a2const void *a2;
- size_t len;
-{
- unsigned long int dstp = (long int) dest;
- unsigned long int srcp = (long int) src;
-
- /* If there is no overlap between ranges, call the builtin memcpy. */
- if (dstp >= srcp + len || srcp > dstp + len)
- __builtin_memcpy (dest, src, len);
-
- /* This test makes the forward copying code be used whenever possible.
- Reduces the working set. */
- else if (dstp - srcp >= len) /* *Unsigned* compare! */
- {
- /* Copy from the beginning to the end. */
-
- /* If there not too few bytes to copy, use word copy. */
- if (len >= OP_T_THRES)
- {
- /* Copy just a few bytes to make DSTP aligned. */
- len -= (-dstp) % OPSIZ;
- BYTE_COPY_FWD (dstp, srcp, (-dstp) % OPSIZ);
-
- /* Copy whole pages from SRCP to DSTP by virtual address
- manipulation, as much as possible. */
-
- PAGE_COPY_FWD_MAYBE (dstp, srcp, len, len);
-
- /* Copy from SRCP to DSTP taking advantage of the known
- alignment of DSTP. Number of bytes remaining is put
- in the third argument, i.e. in LEN. This number may
- vary from machine to machine. */
-
- WORD_COPY_FWD (dstp, srcp, len, len);
-
- /* Fall out and copy the tail. */
- }
-
- /* There are just a few bytes to copy. Use byte memory operations. */
- BYTE_COPY_FWD (dstp, srcp, len);
- }
- else
- {
- /* Copy from the end to the beginning. */
- srcp += len;
- dstp += len;
-
- /* If there not too few bytes to copy, use word copy. */
- if (len >= OP_T_THRES)
- {
- /* Copy just a few bytes to make DSTP aligned. */
- len -= dstp % OPSIZ;
- BYTE_COPY_BWD (dstp, srcp, dstp % OPSIZ);
-
- /* Copy from SRCP to DSTP taking advantage of the known
- alignment of DSTP. Number of bytes remaining is put
- in the third argument, i.e. in LEN. This number may
- vary from machine to machine. */
-
- WORD_COPY_BWD (dstp, srcp, len, len);
-
- /* Fall out and copy the tail. */
- }
-
- /* There are just a few bytes to copy. Use byte memory operations. */
- BYTE_COPY_BWD (dstp, srcp, len);
- }
-
- RETURN (dest);
-}
-#ifndef memmove
-libc_hidden_builtin_def (memmove)
-#endif
fcmpu cr7,fp1,fp12 /* if (x > 0.0 */
ble cr7,L(lessthanzero)
mtfsb0 4*cr7+lt /* Disable FE_INEXACT exception */
- fadd fp0,fp1,fp13 /* x += TWO52 */
- fsub fp1,fp0,fp13 /* x -= TWO52 */
+ fadd fp1,fp1,fp13 /* x += TWO52 */
+ fsub fp1,fp1,fp13 /* x -= TWO52 */
fabs fp1,fp1 /* if (x == 0.0 */
mtfsb0 4*cr1+eq /* Clear any FE_INEXACT exception */
blr
L(lessthanzero):
bgelr cr7
mtfsb0 4*cr7+lt /* Disable FE_INEXACT exception */
- fsub fp0,fp13,fp1 /* x -= TWO52 */
- fsub fp0,fp0,fp13 /* x += TWO52 */
- fneg fp1,fp0 /* if (x == 0.0) */
+ fsub fp1,fp1,fp13 /* x -= TWO52 */
+ fadd fp1,fp1,fp13 /* x += TWO52 */
+ fnabs fp1,fp1 /* if (x == 0.0) */
mtfsb0 4*cr1+eq /* Clear any FE_INEXACT exception */
blr
END (__nearbyint)
fcmpu cr7,fp1,fp12 /* if (x > 0.0 */
ble cr7,L(lessthanzero)
mtfsb0 4*cr7+lt /* Disable FE_INEXACT exception */
- fadds fp0,fp1,fp13 /* x += TWO23 */
- fsubs fp1,fp0,fp13 /* x -= TWO23 */
+ fadds fp1,fp1,fp13 /* x += TWO23 */
+ fsubs fp1,fp1,fp13 /* x -= TWO23 */
+ fabs fp1,fp1 /* if (x == 0.0) */
mtfsb0 4*cr1+eq /* Clear any FE_INEXACT exception */
blr
L(lessthanzero):
bgelr cr7
mtfsb0 4*cr7+lt /* Disable FE_INEXACT exception */
- fsubs fp0,fp13,fp1 /* x -= TWO23 */
- fsubs fp0,fp0,fp13 /* x += TWO23 */
- fneg fp1,fp0 /* if (x == 0.0) */
+ fsubs fp1,fp1,fp13 /* x -= TWO23 */
+ fadds fp1,fp1,fp13 /* x += TWO23 */
+ fnabs fp1,fp1 /* if (x == 0.0) */
mtfsb0 4*cr1+eq /* Clear any FE_INEXACT exception */
blr
END (__nearbyintf)
((byte *) dst_ep)[0] = __x; \
} \
} while (0)
+
+/* The powerpc memcpy implementation is safe to use for memmove. */
+#undef MEMCPY_OK_FOR_FWD_MEMMOVE
+#define MEMCPY_OK_FOR_FWD_MEMMOVE 1
/* Align to word boundary. */
cmplwi cr5, rLEN, 31
- insrdi rCHR, rCHR, 8, 48 /* Replicate byte to halfword. */
+ insrwi rCHR, rCHR, 8, 16 /* Replicate byte to halfword. */
beq+ L(aligned)
mtcrf 0x01, rMEMP0
subfic rALIGN, rALIGN, 4
/* Handle the case of size < 31. */
L(aligned):
mtcrf 0x01, rLEN
- insrdi rCHR, rCHR, 16, 32 /* Replicate halfword to word. */
+ insrwi rCHR, rCHR, 16, 0 /* Replicate halfword to word. */
ble cr5, L(medium)
/* Align to 32-byte boundary. */
andi. rALIGN, rMEMP, 0x1C
strchr-power7 strchr-ppc32 wcschr-power7 wcschr-power6 \
wcschr-ppc32 wcsrchr-power7 wcsrchr-power6 wcsrchr-ppc32 \
wcscpy-power7 wcscpy-power6 wcscpy-ppc32 wordcopy-power7 \
- wordcopy-power6 wordcopy-ppc32
+ wordcopy-power6 wordcopy-ppc32 memmove-power7 memmove-ppc
CFLAGS-strncase-power7.c += -mcpu=power7 -funroll-loops
CFLAGS-strncase_l-power7.c += -mcpu=power7 -funroll-loops
#include <sysdep.h>
-/* memset ifunc selector is not built for static and memset@local
- for shared builds makes the linker point the call to the ifunc
- selector. */
-#ifdef SHARED
-# define MEMSET __memset_ppc
-#else
-# define MEMSET memset
-#endif
-
ENTRY (__bzero_ppc)
mr r5,r4
li r4,0
- b MEMSET@local
+ b __memset_ppc@local
END (__bzero_ppc)
<http://www.gnu.org/licenses/>. */
/* Define multiple versions only for definition in libc. */
-#ifndef NOT_IN_libc
+#if defined SHARED && !defined NOT_IN_libc
# include <string.h>
# include <strings.h>
# include "init-arch.h"
__memcpy_cell)
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_ppc))
+ /* Support sysdeps/powerpc/powerpc32/power4/multiarch/memmove.c. */
+ IFUNC_IMPL (i, name, memmove,
+ IFUNC_IMPL_ADD (array, i, memmove, hwcap & PPC_FEATURE_HAS_VSX,
+ __memmove_power7)
+ IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_ppc))
+
/* Support sysdeps/powerpc/powerpc32/power4/multiarch/memset.c. */
IFUNC_IMPL (i, name, memset,
IFUNC_IMPL_ADD (array, i, memset, hwcap & PPC_FEATURE_HAS_VSX,
and fills the previous ones. */
#define INIT_ARCH() \
unsigned long int hwcap = __GLRO(dl_hwcap); \
+ unsigned long int __attribute__((unused)) hwcap2 = __GLRO(dl_hwcap2); \
if (hwcap & PPC_FEATURE_ARCH_2_06) \
hwcap |= PPC_FEATURE_ARCH_2_05 | \
PPC_FEATURE_POWER5_PLUS | \
#ifdef SHARED
# undef libc_hidden_builtin_def
-# define libc_hidden_builtin_def(name) \
- __hidden_ver1 (__memchr_ppc, __GI_memchr, __memchr_ppc);
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1(__memchr_ppc, __GI_memchr, __memchr_ppc);
#endif
extern __typeof (memchr) __memchr_ppc attribute_hidden;
<http://www.gnu.org/licenses/>. */
#ifndef NOT_IN_libc
+# undef memcpy
+/* Redefine memchr so that the compiler won't make the weak_alias point
+ to internal hidden definition (__GI_memchr), since PPC32 does not
+ support local IFUNC calls. */
+# define memchr __redirect_memchr
# include <string.h>
-# include <shlib-compat.h>
# include "init-arch.h"
-extern __typeof (__memchr) __memchr_ppc attribute_hidden;
-extern __typeof (__memchr) __memchr_power7 attribute_hidden;
+extern __typeof (__redirect_memchr) __memchr_ppc attribute_hidden;
+extern __typeof (__redirect_memchr) __memchr_power7 attribute_hidden;
-/* Avoid DWARF definition DIE on ifunc symbol so that GDB can handle
- ifunc symbol properly. */
-libc_ifunc (__memchr,
+extern __typeof (__redirect_memchr) __libc_memchr;
+
+libc_ifunc (__libc_memchr,
(hwcap & PPC_FEATURE_HAS_VSX)
? __memchr_power7
: __memchr_ppc);
-
-weak_alias (__memchr, memchr)
-libc_hidden_builtin_def (memchr)
+#undef memchr
+weak_alias (__libc_memchr, memchr)
#else
#include <string/memchr.c>
#endif
--- /dev/null
+/* Power7 multiarch memmove.
+ Copyright (C) 2014 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; see the file COPYING.LIB. If
+ not, see <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+extern __typeof (memcpy) __memcpy_power7;
+#define memcpy __memcpy_power7
+
+extern __typeof (memmove) __memmove_power7;
+#define MEMMOVE __memmove_power7
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <string/memmove.c>
--- /dev/null
+/* Power7 multiarch memmove.
+ Copyright (C) 2014 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; see the file COPYING.LIB. If
+ not, see <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+extern __typeof (memcpy) __memcpy_ppc;
+#define memcpy __memcpy_ppc
+
+extern __typeof (memmove) __memmove_ppc;
+#define MEMMOVE __memmove_ppc
+
+#if defined SHARED
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1 (__memmove_ppc, __GI_memmove, __memmove_ppc);
+#endif
+
+#include <string/memmove.c>
--- /dev/null
+/* Multiple versions of memmove. PowerPC32 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#if defined SHARED && !defined NOT_IN_libc
+/* Redefine memmove so that the compiler won't complain about the type
+ mismatch with the IFUNC selector in strong_alias, below. */
+# include <string.h>
+# include "init-arch.h"
+
+extern __typeof (memmove) __memmove_ppc attribute_hidden;
+extern __typeof (memmove) __memmove_power7 attribute_hidden;
+
+libc_ifunc (memmove,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __memmove_power7
+ : __memmove_ppc);
+#else
+# include <string/memmove.c>
+#endif
cfi_endproc; \
ASM_SIZE_DIRECTIVE(__strcasecmp_power7)
-#undef weak_alias(name, alias)
+#undef weak_alias
#define weak_alias(name, alias)
#undef libc_hidden_builtin_def
cfi_endproc; \
ASM_SIZE_DIRECTIVE(__strcasecmp_l_power7)
-#undef weak_alias(name, alias)
+#undef weak_alias
#define weak_alias(name, alias)
#undef libc_hidden_builtin_def
ble- cr1, L(small)
/* Align to word boundary. */
cmplwi cr5, rLEN, 31
- insrdi rCHR, rCHR, 8, 48 /* Replicate byte to halfword. */
+ insrwi rCHR, rCHR, 8, 16 /* Replicate byte to halfword. */
beq+ L(aligned)
mtcrf 0x01, rMEMP0
subfic rALIGN, rALIGN, 4
/* Handle the case of size < 31. */
L(aligned):
mtcrf 0x01, rLEN
- insrdi rCHR, rCHR, 16, 32 /* Replicate halfword to word. */
+ insrwi rCHR, rCHR, 16, 0 /* Replicate halfword to word. */
ble cr5, L(medium)
/* Align to 32-byte boundary. */
andi. rALIGN, rMEMP, 0x1C
--- /dev/null
+powerpc/powerpc32/power6/fpu
CALL_MCOUNT
dcbt 0,r3
clrrwi r8,r3,2
- insrdi r4,r4,8,48
+ insrwi r4,r4,8,16 /* Replicate byte to word. */
add r7,r3,r5 /* Calculate the last acceptable address. */
- insrdi r4,r4,16,32
+ insrwi r4,r4,16,0
cmplwi r5,16
li r9, -1
rlwinm r6,r3,3,27,28 /* Calculate padding. */
ble cr1, L(copy_LT_32) /* If move < 32 bytes use short move
code. */
- andi. 11,3,7 /* Check alignment of DST. */
- clrlwi 10,4,29 /* Check alignment of SRC. */
+ andi. 11,3,15 /* Check alignment of DST. */
+ clrlwi 10,4,28 /* Check alignment of SRC. */
cmplw cr6,10,11 /* SRC and DST alignments match? */
mr 12,4
mr 31,5
dcbt r9,r6,16 /* Stream hint, decreasing addresses. */
/* Replicate BYTE to word. */
- rldimi r4,r4,8,48
- rldimi r4,r4,16,32
+ insrwi r4,r4,8,16
+ insrwi r4,r4,16,0
li r6,-4
li r9,-1
rlwinm r0,r0,3,27,28 /* Calculate padding. */
cfi_offset(31,-8)
/* Replicate byte to word. */
- insrdi 4,4,8,48
- insrdi 4,4,16,32
+ insrwi 4,4,8,16
+ insrwi 4,4,16,0
ble cr6,L(small) /* If length <= 8, use short copy code. */
clrrwi r8,r3,2 /* Align the address to word boundary. */
/* Replicate byte to word. */
- rldimi r4,r4,8,48
- rldimi r4,r4,16,32
+ insrwi r4,r4,8,16
+ insrwi r4,r4,16,0
/* Now r4 has a word of c bytes. */
beq cr7,L(null_match)
/* Replicate byte to word. */
- insrdi r4,r4,8,48
- insrdi r4,r4,16,32
+ insrwi r4,r4,8,16
+ insrwi r4,r4,16,0
/* Now r4 has a word of c bytes and r0 has
a word of null bytes. */
clrrwi r8,r3,2 /* Align the address to word boundary. */
/* Replicate byte to word. */
- insrdi r4,r4,8,48
- insrdi r4,r4,16,32
+ insrwi r4,r4,8,16
+ insrwi r4,r4,16,0
rlwinm r6,r3,3,27,28 /* Calculate padding. */
lwz r12,0(r8) /* Load word from memory. */
#define ENTRY_POINT _start
+#if _CALL_ELF != 2
/* We have to provide a special declaration. */
#define ENTRY_POINT_DECL(class) class void _start (void);
#define TEXT_START \
({ extern unsigned long int _start_as_data[] asm ("_start"); \
_start_as_data[0]; })
+#endif
s_copysign-ppc64 s_finite-power7 s_finite-ppc64 \
s_finitef-ppc64 s_isinff-ppc64 s_isinf-power7 \
s_isinf-ppc64 s_modf-power5+ s_modf-ppc64 \
- s_modff-power5+ s_modff-ppc64
+ s_modff-power5+ s_modff-ppc64 s_isnan-power8 \
+ s_isinf-power8 s_finite-power8
libm-sysdep_routines += s_isnan-power7 s_isnan-power6x s_isnan-power6 \
s_isnan-power5 s_isnan-ppc64 s_llround-power6x \
s_logbl-power7 s_logb-ppc64 s_logbf-ppc64 \
s_logbl-ppc64 s_modf-power5+ s_modf-ppc64 \
s_modff-power5+ s_modff-ppc64 e_hypot-ppc64 \
- e_hypot-power7 e_hypotf-ppc64 e_hypotf-power7
+ e_hypot-power7 e_hypotf-ppc64 e_hypotf-power7 \
+ s_isnan-power8 s_isinf-power8 s_finite-power8 \
+ s_llrint-power8 s_llround-power8
CFLAGS-s_logbf-power7.c = -mcpu=power7
CFLAGS-s_logbl-power7.c = -mcpu=power7
--- /dev/null
+/* isnan(). PowerPC64/POWER7 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+#undef hidden_def
+#define hidden_def(name)
+#undef weak_alias
+#define weak_alias(name, alias)
+#undef strong_alias
+#define strong_alias(name, alias)
+#undef compat_symbol
+#define compat_symbol(lib, name, symbol, ver)
+
+#define __finite __finite_power8
+
+#include <sysdeps/powerpc/powerpc64/power8/fpu/s_finite.S>
extern __typeof (__finite) __finite_ppc64 attribute_hidden;
extern __typeof (__finite) __finite_power7 attribute_hidden;
+extern __typeof (__finite) __finite_power8 attribute_hidden;
libc_ifunc (__finite,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __finite_power7
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __finite_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __finite_power7
: __finite_ppc64);
weak_alias (__finite, finite)
extern __typeof (__finitef) __finitef_ppc64 attribute_hidden;
/* The double-precision version also works for single-precision. */
extern __typeof (__finitef) __finite_power7 attribute_hidden;
+extern __typeof (__finitef) __finite_power8 attribute_hidden;
libc_ifunc (__finitef,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __finite_power7
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __finite_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __finite_power7
: __finitef_ppc64);
weak_alias (__finitef, finitef)
--- /dev/null
+/* isinf(). PowerPC64/POWER8 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+#undef hidden_def
+#define hidden_def(name)
+#undef weak_alias
+#define weak_alias(name, alias)
+#undef strong_alias
+#define strong_alias(name, alias)
+#undef compat_symbol
+#define compat_symbol(lib, name, alias, ver)
+
+#define __isinf __isinf_power8
+
+#include <sysdeps/powerpc/powerpc64/power8/fpu/s_isinf.S>
extern __typeof (__isinf) __isinf_ppc64 attribute_hidden;
extern __typeof (__isinf) __isinf_power7 attribute_hidden;
+extern __typeof (__isinf) __isinf_power8 attribute_hidden;
libc_ifunc (__isinf,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __isinf_power7
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __isinf_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __isinf_power7
: __isinf_ppc64);
weak_alias (__isinf, isinf)
extern __typeof (__isinff) __isinff_ppc64 attribute_hidden;
/* The double-precision version also works for single-precision. */
extern __typeof (__isinff) __isinf_power7 attribute_hidden;
+extern __typeof (__isinff) __isinf_power8 attribute_hidden;
libc_ifunc (__isinff,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __isinf_power7
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __isinf_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __isinf_power7
: __isinff_ppc64);
weak_alias (__isinff, isinff)
--- /dev/null
+/* isnan(). PowerPC64/POWER7 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+#undef hidden_def
+#define hidden_def(name)
+#undef weak_alias
+#define weak_alias(name, alias)
+#undef strong_alias
+#define strong_alias(name, alias)
+#undef compat_symbol
+#define compat_symbol(lib, name, symbol, ver)
+
+#define __isnan __isnan_power8
+
+#include <sysdeps/powerpc/powerpc64/power8/fpu/s_isnan.S>
extern __typeof (__isnan) __isnan_power6 attribute_hidden;
extern __typeof (__isnan) __isnan_power6x attribute_hidden;
extern __typeof (__isnan) __isnan_power7 attribute_hidden;
+extern __typeof (__isnan) __isnan_power8 attribute_hidden;
libc_ifunc (__isnan,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __isnan_power7 :
- (hwcap & PPC_FEATURE_POWER6_EXT)
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __isnan_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __isnan_power7 :
+ (hwcap & PPC_FEATURE_POWER6_EXT)
? __isnan_power6x :
- (hwcap & PPC_FEATURE_ARCH_2_05)
- ? __isnan_power6 :
- (hwcap & PPC_FEATURE_POWER5)
- ? __isnan_power5
+ (hwcap & PPC_FEATURE_ARCH_2_05)
+ ? __isnan_power6 :
+ (hwcap & PPC_FEATURE_POWER5)
+ ? __isnan_power5
: __isnan_ppc64);
weak_alias (__isnan, isnan)
extern __typeof (__isnanf) __isnan_power6 attribute_hidden;
extern __typeof (__isnanf) __isnan_power6x attribute_hidden;
extern __typeof (__isnanf) __isnan_power7 attribute_hidden;
+extern __typeof (__isnanf) __isnan_power8 attribute_hidden;
libc_ifunc (__isnanf,
- (hwcap & PPC_FEATURE_ARCH_2_06)
- ? __isnan_power7 :
- (hwcap & PPC_FEATURE_POWER6_EXT)
- ? __isnan_power6x :
- (hwcap & PPC_FEATURE_ARCH_2_05)
- ? __isnan_power6 :
- (hwcap & PPC_FEATURE_POWER5)
- ? __isnan_power5
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __isnan_power8 :
+ (hwcap & PPC_FEATURE_ARCH_2_06)
+ ? __isnan_power7 :
+ (hwcap & PPC_FEATURE_POWER6_EXT)
+ ? __isnan_power6x :
+ (hwcap & PPC_FEATURE_ARCH_2_05)
+ ? __isnan_power6 :
+ (hwcap & PPC_FEATURE_POWER5)
+ ? __isnan_power5
: __isnan_ppc64);
weak_alias (__isnanf, isnanf)
--- /dev/null
+/* Round double to long int. PowerPC64/POWER6X default version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+#undef weak_alias
+#define weak_alias(a,b)
+#undef strong_alias
+#define strong_alias(a,b)
+#undef compat_symbol
+#define compat_symbol(a,b,c,d)
+
+#define __llrint __llrint_power8
+
+#include <sysdeps/powerpc/powerpc64/power8/fpu/s_llrint.S>
extern __typeof (__llrint) __llrint_ppc64 attribute_hidden;
extern __typeof (__llrint) __llrint_power6x attribute_hidden;
+extern __typeof (__llrint) __llrint_power8 attribute_hidden;
libc_ifunc (__llrint,
- (hwcap & PPC_FEATURE_POWER6_EXT)
- ? __llrint_power6x
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __llrint_power8 :
+ (hwcap & PPC_FEATURE_POWER6_EXT)
+ ? __llrint_power6x
: __llrint_ppc64);
weak_alias (__llrint, llrint)
--- /dev/null
+/* llround(). PowerPC64 default version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+#undef weak_alias
+#define weak_alias(name, alias)
+#undef strong_alias
+#define strong_alias(name, alias)
+#undef compat_symbol
+#define compat_symbol(lib, name, alias, ver)
+
+#define __llround __llround_power8
+
+#include <sysdeps/powerpc/powerpc64/power8/fpu/s_llround.S>
extern __typeof (__llround) __llround_ppc64 attribute_hidden;
extern __typeof (__llround) __llround_power5plus attribute_hidden;
extern __typeof (__llround) __llround_power6x attribute_hidden;
+extern __typeof (__llround) __llround_power8 attribute_hidden;
libc_ifunc (__llround,
- (hwcap & PPC_FEATURE_POWER6_EXT)
- ? __llround_power6x :
- (hwcap & PPC_FEATURE_POWER5_PLUS)
- ? __llround_power5plus
+ (hwcap2 & PPC_FEATURE2_ARCH_2_07)
+ ? __llround_power8 :
+ (hwcap & PPC_FEATURE_POWER6_EXT)
+ ? __llround_power6x :
+ (hwcap & PPC_FEATURE_POWER5_PLUS)
+ ? __llround_power5plus
: __llround_ppc64);
weak_alias (__llround, llround)
+++ /dev/null
-/* s_ceill.S IBM extended format long double version.
- Copyright (C) 2004-2014 Free Software Foundation, Inc.
- This file is part of the GNU C Library.
-
- The GNU C Library is free software; you can redistribute it and/or
- modify it under the terms of the GNU Lesser General Public
- License as published by the Free Software Foundation; either
- version 2.1 of the License, or (at your option) any later version.
-
- The GNU C Library is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- Lesser General Public License for more details.
-
- You should have received a copy of the GNU Lesser General Public
- License along with the GNU C Library; if not, see
- <http://www.gnu.org/licenses/>. */
-
-#include <sysdep.h>
-#include <math_ldbl_opt.h>
-
- .section ".toc","aw"
-.LC0: /* 2**52 */
- .tc FD_43300000_0[TC],0x4330000000000000
-
- .section ".text"
-
-/* long double [fp1,fp2] ceill (long double x [fp1,fp2])
- IEEE 1003.1 ceil function.
-
- PowerPC64 long double uses the IBM extended format which is
- represented two 64-floating point double values. The values are
- non-overlapping giving an effective precision of 106 bits. The first
- double contains the high order bits of mantissa and is always ceiled
- to represent a normal ceiling of long double to double. Since the
- long double value is sum of the high and low values, the low double
- normally has the opposite sign to compensate for the this ceiling.
-
- For long double there are two cases:
- 1) |x| < 2**52, all the integer bits are in the high double.
- ceil the high double and set the low double to -0.0.
- 2) |x| >= 2**52, ceiling involves both doubles.
- See the comment before label .L2 for details.
- */
-
-ENTRY (__ceill)
- mffs fp11 /* Save current FPU rounding mode. */
- lfd fp13,.LC0@toc(2)
- fabs fp0,fp1
- fabs fp9,fp2
- fsub fp12,fp13,fp13 /* generate 0.0 */
- fcmpu cr7,fp0,fp13 /* if (fabs(x) > TWO52) */
- fcmpu cr6,fp1,fp12 /* if (x > 0.0) */
- bnl- cr7,.L2
- mtfsfi 7,2 /* Set rounding mode toward +inf. */
- fneg fp2,fp12
- ble- cr6,.L1
- fadd fp1,fp1,fp13 /* x+= TWO52; */
- fsub fp1,fp1,fp13 /* x-= TWO52; */
- fabs fp1,fp1 /* if (x == 0.0) */
-.L0:
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- blr /* x = 0.0; */
-.L1:
- bge- cr6,.L0 /* if (x < 0.0) */
- fsub fp1,fp1,fp13 /* x-= TWO52; */
- fadd fp1,fp1,fp13 /* x+= TWO52; */
- fcmpu cr5,fp1,fp12 /* if (x > 0.0) */
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- fnabs fp1,fp1 /* if (x == 0.0) */
- blr /* x = -0.0; */
-
-/* The high double is > TWO52 so we need to round the low double and
- perhaps the high double. In this case we have to round the low
- double and handle any adjustment to the high double that may be
- caused by rounding (up). This is complicated by the fact that the
- high double may already be rounded and the low double may have the
- opposite sign to compensate.This gets a bit tricky so we use the
- following algorithm:
-
- tau = floor(x_high/TWO52);
- x0 = x_high - tau;
- x1 = x_low + tau;
- r1 = rint(x1);
- y_high = x0 + r1;
- y_low = x0 - y_high + r1;
- return y; */
-.L2:
- fcmpu cr7,fp9,fp13 /* if (|x_low| > TWO52) */
- fcmpu cr0,fp9,fp12 /* || (|x_low| == 0.0) */
- fcmpu cr5,fp2,fp12 /* if (x_low > 0.0) */
- bgelr- cr7 /* return x; */
- beqlr- cr0
- mtfsfi 7,2 /* Set rounding mode toward +inf. */
- fdiv fp8,fp1,fp13 /* x_high/TWO52 */
-
- bng- cr6,.L6 /* if (x > 0.0) */
- fctidz fp0,fp8
- fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */
- bng cr5,.L4 /* if (x_low > 0.0) */
- fmr fp3,fp1
- fmr fp4,fp2
- b .L5
-.L4: /* if (x_low < 0.0) */
- fsub fp3,fp1,fp8 /* x0 = x_high - tau; */
- fadd fp4,fp2,fp8 /* x1 = x_low + tau; */
-.L5:
- fadd fp5,fp4,fp13 /* r1 = r1 + TWO52; */
- fsub fp5,fp5,fp13 /* r1 = r1 - TWO52; */
- b .L9
-.L6: /* if (x < 0.0) */
- fctidz fp0,fp8
- fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */
- bnl cr5,.L7 /* if (x_low < 0.0) */
- fmr fp3,fp1
- fmr fp4,fp2
- b .L8
-.L7: /* if (x_low > 0.0) */
- fsub fp3,fp1,fp8 /* x0 = x_high - tau; */
- fadd fp4,fp2,fp8 /* x1 = x_low + tau; */
-.L8:
- fsub fp5,fp4,fp13 /* r1-= TWO52; */
- fadd fp5,fp5,fp13 /* r1+= TWO52; */
-.L9:
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- fadd fp1,fp3,fp5 /* y_high = x0 + r1; */
- fsub fp2,fp3,fp1 /* y_low = x0 - y_high + r1; */
- fadd fp2,fp2,fp5
- blr
-END (__ceill)
-
-long_double_symbol (libm, __ceill, ceill)
+++ /dev/null
-/* nearbyint long double.
- IBM extended format long double version.
- Copyright (C) 2004-2014 Free Software Foundation, Inc.
- This file is part of the GNU C Library.
-
- The GNU C Library is free software; you can redistribute it and/or
- modify it under the terms of the GNU Lesser General Public
- License as published by the Free Software Foundation; either
- version 2.1 of the License, or (at your option) any later version.
-
- The GNU C Library is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- Lesser General Public License for more details.
-
- You should have received a copy of the GNU Lesser General Public
- License along with the GNU C Library; if not, see
- <http://www.gnu.org/licenses/>. */
-
-#include <sysdep.h>
-#include <math_ldbl_opt.h>
-
- .section ".toc","aw"
-.LC0: /* 2**52 */
- .tc FD_43300000_0[TC],0x4330000000000000
- .section ".text"
-
-/* long double [fp1,fp2] nearbyintl (long double x [fp1,fp2])
- IEEE 1003.1 nearbyintl function. nearbyintl is similar to the rintl
- but does raise the "inexact" exception. This implementation is
- based on rintl but explicitly masks the inexact exception on entry
- and clears any pending inexact before restoring the exception mask
- on exit.
-
- PowerPC64 long double uses the IBM extended format which is
- represented two 64-floating point double values. The values are
- non-overlapping giving an effective precision of 106 bits. The first
- double contains the high order bits of mantissa and is always rounded
- to represent a normal rounding of long double to double. Since the
- long double value is sum of the high and low values, the low double
- normally has the opposite sign to compensate for the this rounding.
-
- For long double there are two cases:
- 1) |x| < 2**52, all the integer bits are in the high double.
- floor the high double and set the low double to -0.0.
- 2) |x| >= 2**52, Rounding involves both doubles.
- See the comment before label .L2 for details.
- */
-ENTRY (__nearbyintl)
- mffs fp11 /* Save current FPSCR. */
- lfd fp13,.LC0@toc(2)
- fabs fp0,fp1
- mtfsb0 28 /* Disable "inexact" exceptions. */
- fsub fp12,fp13,fp13 /* generate 0.0 */
- fabs fp9,fp2
- fcmpu cr7,fp0,fp13 /* if (fabs(x) > TWO52) */
- fcmpu cr6,fp1,fp12 /* if (x > 0.0) */
- bnl- cr7,.L2
- fmr fp2,fp12
- bng- cr6,.L4
- fadd fp1,fp1,fp13 /* x+= TWO52; */
- fsub fp1,fp1,fp13 /* x-= TWO52; */
- b .L9
-.L4:
- bnl- cr6,.L9 /* if (x < 0.0) */
- fsub fp1,fp13,fp1 /* x = TWO52 - x; */
- fsub fp0,fp1,fp13 /* x = - (x - TWO52); */
- fneg fp1,fp0
-.L9:
- mtfsb0 6 /* Clear any pending "inexact" exceptions. */
- mtfsf 0x01,fp11 /* restore exception mask. */
- blr
-
-/* The high double is > TWO52 so we need to round the low double and
- perhaps the high double. This gets a bit tricky so we use the
- following algorithm:
-
- tau = floor(x_high/TWO52);
- x0 = x_high - tau;
- x1 = x_low + tau;
- r1 = nearbyint(x1);
- y_high = x0 + r1;
- y_low = r1 - tau;
- return y; */
-.L2:
- fcmpu cr7,fp9,fp13 /* if (|x_low| > TWO52) */
- fcmpu cr0,fp9,fp12 /* || (|x_low| == 0.0) */
- bge- cr7,.L9 /* return x; */
- beq- cr0,.L9
- fdiv fp8,fp1,fp13 /* x_high/TWO52 */
- fctidz fp0,fp8
- fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */
- fsub fp3,fp1,fp8 /* x0 = x_high - tau; */
- fadd fp4,fp2,fp8 /* x1 = x_low + tau; */
-
- fcmpu cr6,fp4,fp12 /* if (x1 > 0.0) */
- bng- cr6,.L8
- fadd fp5,fp4,fp13 /* r1 = x1 + TWO52; */
- fsub fp5,fp5,fp13 /* r1 = r1 - TWO52; */
- b .L6
-.L8:
- fmr fp5,fp4
- bge- cr6,.L6 /* if (x1 < 0.0) */
- fsub fp5,fp13,fp4 /* r1 = TWO52 - x1; */
- fsub fp0,fp5,fp13 /* r1 = - (r1 - TWO52); */
- fneg fp5,fp0
-.L6:
- fadd fp1,fp3,fp5 /* y_high = x0 + r1; */
- fsub fp2,fp5,fp8 /* y_low = r1 - tau; */
- b .L9
-END (__nearbyintl)
-
-long_double_symbol (libm, __nearbyintl, nearbyintl)
+++ /dev/null
-/* long double round function.
- IBM extended format long double version.
- Copyright (C) 2004-2014 Free Software Foundation, Inc.
- This file is part of the GNU C Library.
-
- The GNU C Library is free software; you can redistribute it and/or
- modify it under the terms of the GNU Lesser General Public
- License as published by the Free Software Foundation; either
- version 2.1 of the License, or (at your option) any later version.
-
- The GNU C Library is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- Lesser General Public License for more details.
-
- You should have received a copy of the GNU Lesser General Public
- License along with the GNU C Library; if not, see
- <http://www.gnu.org/licenses/>. */
-
-#include <sysdep.h>
-#include <math_ldbl_opt.h>
-
- .section ".toc","aw"
-.LC0: /* 2**52 */
- .tc FD_43300000_0[TC],0x4330000000000000
-.LC1: /* 0.5 */
- .tc FD_3fe00000_0[TC],0x3fe0000000000000
- .section ".text"
-
-/* long double [fp1,fp2] roundl (long double x [fp1,fp2])
- IEEE 1003.1 round 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 (__roundl)
- mffs fp11 /* Save current FPU rounding mode. */
- lfd fp13,.LC0@toc(2)
- fabs fp0,fp1
- fabs fp9,fp2
- fsub fp12,fp13,fp13 /* generate 0.0 */
- fcmpu cr7,fp0,fp13 /* if (fabs(x) > TWO52) */
- fcmpu cr6,fp1,fp12 /* if (x > 0.0) */
- bnl- cr7,.L2
- mtfsfi 7,1 /* Set rounding mode toward 0. */
- lfd fp10,.LC1@toc(2)
- ble- cr6,.L1
- fneg fp2,fp12
- fadd fp1,fp1,fp10 /* x+= 0.5; */
- fadd fp1,fp1,fp13 /* x+= TWO52; */
- fsub fp1,fp1,fp13 /* x-= TWO52; */
- fabs fp1,fp1 /* if (x == 0.0) x = 0.0; */
-.L0:
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- blr
-.L1:
- fsub fp9,fp1,fp10 /* x-= 0.5; */
- fneg fp2,fp12
- bge- cr6,.L0 /* if (x < 0.0) */
- fsub fp1,fp9,fp13 /* x-= TWO52; */
- fadd fp1,fp1,fp13 /* x+= TWO52; */
- fnabs fp1,fp1 /* if (x == 0.0) x = -0.0; */
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- blr
-
-/* The high double is > TWO52 so we need to round the low double and
- perhaps the high double. In this case we have to round the low
- double and handle any adjustment to the high double that may be
- caused by rounding (up). This is complicated by the fact that the
- high double may already be rounded and the low double may have the
- opposite sign to compensate.This gets a bit tricky so we use the
- following algorithm:
-
- tau = floor(x_high/TWO52);
- x0 = x_high - tau;
- x1 = x_low + tau;
- r1 = rint(x1);
- y_high = x0 + r1;
- y_low = x0 - y_high + r1;
- return y; */
-.L2:
- fcmpu cr7,fp9,fp13 /* if (|x_low| > TWO52) */
- fcmpu cr0,fp9,fp12 /* || (|x_low| == 0.0) */
- fcmpu cr5,fp2,fp12 /* if (x_low > 0.0) */
- lfd fp10,.LC1@toc(2)
- bgelr- cr7 /* return x; */
- beqlr- cr0
- mtfsfi 7,1 /* Set rounding mode toward 0. */
- fdiv fp8,fp1,fp13 /* x_high/TWO52 */
-
- bng- cr6,.L6 /* if (x > 0.0) */
- fctidz fp0,fp8
- fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */
- bng cr5,.L4 /* if (x_low > 0.0) */
- fmr fp3,fp1
- fmr fp4,fp2
- b .L5
-.L4: /* if (x_low < 0.0) */
- fsub fp3,fp1,fp8 /* x0 = x_high - tau; */
- fadd fp4,fp2,fp8 /* x1 = x_low + tau; */
-.L5:
- fadd fp5,fp4,fp10 /* r1 = x1 + 0.5; */
- fadd fp5,fp5,fp13 /* r1 = r1 + TWO52; */
- fsub fp5,fp5,fp13 /* r1 = r1 - TWO52; */
- b .L9
-.L6: /* if (x < 0.0) */
- fctidz fp0,fp8
- fcfid fp8,fp0 /* tau = floor(x_high/TWO52); */
- bnl cr5,.L7 /* if (x_low < 0.0) */
- fmr fp3,fp1
- fmr fp4,fp2
- b .L8
-.L7: /* if (x_low > 0.0) */
- fsub fp3,fp1,fp8 /* x0 = x_high - tau; */
- fadd fp4,fp2,fp8 /* x1 = x_low + tau; */
-.L8:
- fsub fp5,fp4,fp10 /* r1 = x1 - 0.5; */
- fsub fp5,fp5,fp13 /* r1-= TWO52; */
- fadd fp5,fp5,fp13 /* r1+= TWO52; */
-.L9:
- mtfsf 0x01,fp11 /* restore previous rounding mode. */
- fadd fp1,fp3,fp5 /* y_high = x0 + r1; */
- fsub fp2,fp3,fp1 /* y_low = x0 - y_high + r1; */
- fadd fp2,fp2,fp5
- blr
-END (__roundl)
-
-long_double_symbol (libm, __roundl, roundl)
wcschr-power6 wcschr-ppc64 wcsrchr-power7 wcsrchr-power6 \
wcsrchr-ppc64 wcscpy-power7 wcscpy-power6 wcscpy-ppc64 \
wordcopy-power7 wordcopy-power6 wordcopy-ppc64 \
- strcpy-power7 strcpy-ppc64 stpcpy-power7 stpcpy-ppc64
+ strcpy-power7 strcpy-ppc64 stpcpy-power7 stpcpy-ppc64 \
+ strrchr-power7 strrchr-ppc64 strncat-power7 strncat-ppc64 \
+ strspn-power7 strspn-ppc64 strcspn-power7 strcspn-ppc64 \
+ strpbrk-power7 strpbrk-ppc64 strncpy-power7 strncpy-ppc64 \
+ stpncpy-power7 stpncpy-ppc64 strcmp-power7 strcmp-ppc64 \
+ strcat-power7 strcat-ppc64 memmove-power7 memmove-ppc64 \
+ bcopy-ppc64
CFLAGS-strncase-power7.c += -mcpu=power7 -funroll-loops
CFLAGS-strncase_l-power7.c += -mcpu=power7 -funroll-loops
--- /dev/null
+/* PowerPC64 default bcopy.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+extern __typeof (bcopy) __bcopy_ppc attribute_hidden;
+
+#define bcopy __bcopy_ppc
+
+#include <string/bcopy.c>
--- /dev/null
+/* PowerPC64 multiarch bcopy.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+#include "init-arch.h"
+
+extern __typeof (bcopy) __bcopy_ppc attribute_hidden;
+/* __bcopy_power7 symbol is implemented at memmove-power7.S */
+extern __typeof (bcopy) __bcopy_power7 attribute_hidden;
+
+libc_ifunc (bcopy,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __bcopy_power7
+ : __bcopy_ppc);
__memcpy_power4)
IFUNC_IMPL_ADD (array, i, memcpy, 1, __memcpy_ppc))
+ /* Support sysdeps/powerpc/powerpc64/multiarch/memmove.c. */
+ IFUNC_IMPL (i, name, memmove,
+ IFUNC_IMPL_ADD (array, i, memmove, hwcap & PPC_FEATURE_HAS_VSX,
+ __memmove_power7)
+ IFUNC_IMPL_ADD (array, i, memmove, 1, __memmove_ppc))
+
/* Support sysdeps/powerpc/powerpc64/multiarch/memset.c. */
IFUNC_IMPL (i, name, memset,
IFUNC_IMPL_ADD (array, i, memset, hwcap & PPC_FEATURE_HAS_VSX,
__bzero_power4)
IFUNC_IMPL_ADD (array, i, bzero, 1, __bzero_ppc))
+ /* Support sysdeps/powerpc/powerpc64/multiarch/bcopy.c. */
+ IFUNC_IMPL (i, name, bcopy,
+ IFUNC_IMPL_ADD (array, i, bcopy, hwcap & PPC_FEATURE_HAS_VSX,
+ __bcopy_power7)
+ IFUNC_IMPL_ADD (array, i, bcopy, 1, __bcopy_ppc))
+
/* Support sysdeps/powerpc/powerpc64/multiarch/mempcpy.c. */
IFUNC_IMPL (i, name, mempcpy,
IFUNC_IMPL_ADD (array, i, mempcpy,
IFUNC_IMPL_ADD (array, i, wcscpy, 1,
__wcscpy_ppc))
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strrchr.c. */
+ IFUNC_IMPL (i, name, strrchr,
+ IFUNC_IMPL_ADD (array, i, strrchr,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strrchr_power7)
+ IFUNC_IMPL_ADD (array, i, strrchr, 1,
+ __strrchr_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strncat.c. */
+ IFUNC_IMPL (i, name, strncat,
+ IFUNC_IMPL_ADD (array, i, strncat,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strncat_power7)
+ IFUNC_IMPL_ADD (array, i, strncat, 1,
+ __strncat_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strspn.c. */
+ IFUNC_IMPL (i, name, strspn,
+ IFUNC_IMPL_ADD (array, i, strspn,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strspn_power7)
+ IFUNC_IMPL_ADD (array, i, strspn, 1,
+ __strspn_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strcspn.c. */
+ IFUNC_IMPL (i, name, strcspn,
+ IFUNC_IMPL_ADD (array, i, strcspn,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strcspn_power7)
+ IFUNC_IMPL_ADD (array, i, strcspn, 1,
+ __strcspn_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strpbrk.c. */
+ IFUNC_IMPL (i, name, strpbrk,
+ IFUNC_IMPL_ADD (array, i, strpbrk,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strpbrk_power7)
+ IFUNC_IMPL_ADD (array, i, strpbrk, 1,
+ __strpbrk_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strncpy.c. */
+ IFUNC_IMPL (i, name, strncpy,
+ IFUNC_IMPL_ADD (array, i, strncpy,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strncpy_power7)
+ IFUNC_IMPL_ADD (array, i, strncpy, 1,
+ __strncpy_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/stpncpy.c. */
+ IFUNC_IMPL (i, name, stpncpy,
+ IFUNC_IMPL_ADD (array, i, stpncpy,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __stpncpy_power7)
+ IFUNC_IMPL_ADD (array, i, stpncpy, 1,
+ __stpncpy_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strcmp.c. */
+ IFUNC_IMPL (i, name, strcmp,
+ IFUNC_IMPL_ADD (array, i, strcmp,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strcmp_power7)
+ IFUNC_IMPL_ADD (array, i, strcmp, 1,
+ __strcmp_ppc))
+
+ /* Support sysdeps/powerpc/powerpc64/multiarch/strcat.c. */
+ IFUNC_IMPL (i, name, strcat,
+ IFUNC_IMPL_ADD (array, i, strcat,
+ hwcap & PPC_FEATURE_HAS_VSX,
+ __strcat_power7)
+ IFUNC_IMPL_ADD (array, i, strcat, 1,
+ __strcat_ppc))
+
return i;
}
END_2(__memcpy_ppc)
# undef libc_hidden_builtin_def
-# define libc_hidden_builtin_def(name) \
- .globl __GI_memcpy; __GI_memcpy = __memcpy_ppc
+# define libc_hidden_builtin_def(name)
#endif
#include <sysdeps/powerpc/powerpc64/memcpy.S>
DSO. In static binaries we need memcpy before the initialization
happened. */
#if defined SHARED && !defined NOT_IN_libc
+/* Redefine memcpy so that the compiler won't complain about the type
+ mismatch with the IFUNC selector in strong_alias, below. */
+# undef memcpy
+# define memcpy __redirect_memcpy
# include <string.h>
-# include <shlib-compat.h>
# include "init-arch.h"
-extern __typeof (memcpy) __memcpy_ppc attribute_hidden;
-extern __typeof (memcpy) __memcpy_power4 attribute_hidden;
-extern __typeof (memcpy) __memcpy_cell attribute_hidden;
-extern __typeof (memcpy) __memcpy_power6 attribute_hidden;
-extern __typeof (memcpy) __memcpy_a2 attribute_hidden;
-extern __typeof (memcpy) __memcpy_power7 attribute_hidden;
+extern __typeof (__redirect_memcpy) __libc_memcpy;
-/* Avoid DWARF definition DIE on ifunc symbol so that GDB can handle
- ifunc symbol properly. */
-libc_ifunc (memcpy,
+extern __typeof (__redirect_memcpy) __memcpy_ppc attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_power4 attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_cell attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_power6 attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_a2 attribute_hidden;
+extern __typeof (__redirect_memcpy) __memcpy_power7 attribute_hidden;
+
+libc_ifunc (__libc_memcpy,
(hwcap & PPC_FEATURE_HAS_VSX)
? __memcpy_power7 :
(hwcap & PPC_FEATURE_ARCH_2_06)
(hwcap & PPC_FEATURE_POWER4)
? __memcpy_power4
: __memcpy_ppc);
+
+#undef memcpy
+strong_alias (__libc_memcpy, memcpy);
+libc_hidden_ver (__libc_memcpy, memcpy);
#endif
--- /dev/null
+/* Optimized memmove implementation for PowerPC64/POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__memmove_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__memmove_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__memmove_power7)
+
+#undef END_GEN_TB
+#define END_GEN_TB(name, mask) \
+ cfi_endproc; \
+ TRACEBACK_MASK(__memmove_power7,mask) \
+ END_2(__memmove_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#undef bcopy
+#define bcopy __bcopy_power7
+
+#include <sysdeps/powerpc/powerpc64/power7/memmove.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define MEMMOVE __memmove_ppc
+#if !defined(NOT_IN_libc) && defined(SHARED)
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name)
+#endif
+
+extern __typeof (memmove) __memmove_ppc attribute_hidden;
+
+#include <string/memmove.c>
--- /dev/null
+/* Multiple versions of memmove. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* Define multiple versions only for the definition in lib and for
+ DSO. In static binaries we need memmove before the initialization
+ happened. */
+#if defined SHARED && !defined NOT_IN_libc
+/* Redefine memmove so that the compiler won't complain about the type
+ mismatch with the IFUNC selector in strong_alias, below. */
+# undef memmove
+# define memmove __redirect_memmove
+# include <string.h>
+# include "init-arch.h"
+
+extern __typeof (__redirect_memmove) __libc_memmove;
+
+extern __typeof (__redirect_memmove) __memmove_ppc attribute_hidden;
+extern __typeof (__redirect_memmove) __memmove_power7 attribute_hidden;
+
+libc_ifunc (__libc_memmove,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __memmove_power7
+ : __memmove_ppc);
+
+#undef memmove
+strong_alias (__libc_memmove, memmove);
+libc_hidden_ver (__libc_memmove, memmove);
+#else
+# include <string/memmove.c>
+#endif
END_2(__memset_ppc)
# undef libc_hidden_builtin_def
-# define libc_hidden_builtin_def(name) \
- .globl __GI_memset; __GI_memset = __memset_ppc
+# define libc_hidden_builtin_def(name)
#endif
/* Do not implement __bzero at powerpc64/memset.S. */
/* Define multiple versions only for definition in libc. */
#if defined SHARED && !defined NOT_IN_libc
+/* Redefine memset so that the compiler won't complain about the type
+ mismatch with the IFUNC selector in strong_alias, below. */
+# undef memset
+# define memset __redirect_memset
# include <string.h>
# include <shlib-compat.h>
# include "init-arch.h"
-extern __typeof (memset) __memset_ppc attribute_hidden;
-extern __typeof (memset) __memset_power4 attribute_hidden;
-extern __typeof (memset) __memset_power6 attribute_hidden;
-extern __typeof (memset) __memset_power7 attribute_hidden;
+extern __typeof (__redirect_memset) __libc_memset;
+
+extern __typeof (__redirect_memset) __memset_ppc attribute_hidden;
+extern __typeof (__redirect_memset) __memset_power4 attribute_hidden;
+extern __typeof (__redirect_memset) __memset_power6 attribute_hidden;
+extern __typeof (__redirect_memset) __memset_power7 attribute_hidden;
/* Avoid DWARF definition DIE on ifunc symbol so that GDB can handle
ifunc symbol properly. */
-libc_ifunc (memset,
+libc_ifunc (__libc_memset,
(hwcap & PPC_FEATURE_HAS_VSX)
? __memset_power7 :
(hwcap & PPC_FEATURE_ARCH_2_05)
(hwcap & PPC_FEATURE_POWER4)
? __memset_power4
: __memset_ppc);
+
+#undef memset
+strong_alias (__libc_memset, memset);
+libc_hidden_ver (__libc_memset, memset);
#endif
--- /dev/null
+/* Optimized stpncpy implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#define USE_AS_STPNCPY
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__stpncpy_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__stpncpy_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__stpncpy_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__stpncpy_power7) \
+ END_2(__stpncpy_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#define MEMSET __memset_power7
+
+#include <sysdeps/powerpc/powerpc64/power7/stpncpy.S>
--- /dev/null
+/* Default stpncpy implementation for PowerPC64.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#define STPNCPY __stpncpy_ppc
+#ifdef SHARED
+#undef libc_hidden_def
+#define libc_hidden_def(name) \
+ __hidden_ver1 (__stpncpy_ppc, __GI___stpncpy, __stpncpy_ppc);
+#endif
+
+#include <string/stpncpy.c>
--- /dev/null
+/* Multiple versions of stpncpy. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (__stpncpy) __stpncpy_ppc attribute_hidden;
+extern __typeof (__stpncpy) __stpncpy_power7 attribute_hidden;
+
+libc_ifunc (__stpncpy,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __stpncpy_power7
+ : __stpncpy_ppc);
+
+weak_alias (__stpncpy, stpncpy)
+#endif
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/ >. */
+
+#include <string.h>
+
+#define STRCAT __strcat_power7
+
+#undef libc_hidden_def
+#define libc_hidden_def(name)
+
+extern typeof (strcpy) __strcpy_power7;
+extern typeof (strlen) __strlen_power7;
+
+#define strcpy __strcpy_power7
+#define strlen __strlen_power7
+#include <sysdeps/powerpc/strcat.c>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/ >. */
+
+#include <string.h>
+
+#define STRCAT __strcat_ppc
+#ifdef SHARED
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1 (__strcat_ppc, __GI_strcat, __strcat_ppc);
+#endif
+
+extern __typeof (strcat) __strcat_ppc attribute_hidden;
+
+#include <sysdeps/powerpc/strcat.c>
--- /dev/null
+/* Multiple versions of strcat. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strcat) __strcat_ppc attribute_hidden;
+extern __typeof (strcat) __strcat_power7 attribute_hidden;
+
+libc_ifunc (strcat,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strcat_power7
+ : __strcat_ppc);
+#endif
--- /dev/null
+/* Optimized strcmp implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strcmp_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strcmp_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strcmp_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strcmp_power7) \
+ END_2(__strcmp_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <sysdeps/powerpc/powerpc64/power7/strcmp.S>
--- /dev/null
+/* Default strcmp implementation for PowerPC64.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#if defined SHARED && !defined NOT_IN_libc
+# undef EALIGN
+# define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strcmp_ppc) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strcmp_ppc): \
+ cfi_startproc; \
+ LOCALENTRY(__strcmp_ppc)
+
+# undef END
+# define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strcmp_ppc) \
+ END_2(__strcmp_ppc)
+
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ .globl __GI_strcmp; __GI_strcmp = __strcmp_ppc
+#endif
+
+#include <sysdeps/powerpc/powerpc64/strcmp.S>
--- /dev/null
+/* Multiple versions of strcmp. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#if defined SHARED && !defined NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strcmp) __strcmp_ppc attribute_hidden;
+extern __typeof (strcmp) __strcmp_power7 attribute_hidden;
+
+libc_ifunc (strcmp,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strcmp_power7
+ : __strcmp_ppc);
+#endif
--- /dev/null
+/* Optimized strcspn implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strcspn_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strcspn_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strcspn_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strcspn_power7) \
+ END_2(__strcspn_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <sysdeps/powerpc/powerpc64/power7/strcspn.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define STRCSPN __strcspn_ppc
+#ifdef SHARED
+
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1 (__strcspn_ppc, __GI_strcspn, __strcspn_ppc);
+#endif
+
+extern __typeof (strcspn) __strcspn_ppc attribute_hidden;
+
+#include <string/strcspn.c>
--- /dev/null
+/* Multiple versions of strcspn. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strcspn) __strcspn_ppc attribute_hidden;
+extern __typeof (strcspn) __strcspn_power7 attribute_hidden;
+
+libc_ifunc (strcspn,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strcspn_power7
+ : __strcspn_ppc);
+#endif
END_2(__strlen_ppc)
# undef libc_hidden_builtin_def
-# define libc_hidden_builtin_def(name) \
- .globl __GI_strlen; __GI_strlen = __strlen_ppc
+# define libc_hidden_builtin_def(name)
#endif
#include <sysdeps/powerpc/powerpc64/strlen.S>
<http://www.gnu.org/licenses/>. */
#if defined SHARED && !defined NOT_IN_libc
+/* Redefine strlen so that the compiler won't complain about the type
+ mismatch with the IFUNC selector in strong_alias, below. */
+# undef strlen
+# define strlen __redirect_strlen
# include <string.h>
# include <shlib-compat.h>
# include "init-arch.h"
-extern __typeof (strlen) __strlen_ppc attribute_hidden;
-extern __typeof (strlen) __strlen_power7 attribute_hidden;
+extern __typeof (__redirect_strlen) __libc_strlen;
-libc_ifunc (strlen,
+extern __typeof (__redirect_strlen) __strlen_ppc attribute_hidden;
+extern __typeof (__redirect_strlen) __strlen_power7 attribute_hidden;
+
+libc_ifunc (__libc_strlen,
(hwcap & PPC_FEATURE_HAS_VSX)
? __strlen_power7
: __strlen_ppc);
+
+#undef strlen
+strong_alias (__libc_strlen, strlen)
+libc_hidden_ver (__libc_strlen, strlen)
#endif
--- /dev/null
+/* Optimized strncat implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strncat_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strncat_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strncat_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strncat_power7) \
+ END_2(__strncat_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#define STRLEN __strlen_power7
+
+#include <sysdeps/powerpc/powerpc64/power7/strncat.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/ >. */
+
+#include <string.h>
+
+#define STRNCAT __strncat_ppc
+#ifdef SHARED
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1 (__strncat_ppc, __GI_strncat, __strncat_ppc);
+#endif
+
+extern __typeof (strncat) __strncat_ppc attribute_hidden;
+
+#include <string/strncat.c>
--- /dev/null
+/* Multiple versions of strncat. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strncat) __strncat_ppc attribute_hidden;
+extern __typeof (strncat) __strncat_power7 attribute_hidden;
+
+libc_ifunc (strncat,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strncat_power7
+ : __strncat_ppc);
+#endif
--- /dev/null
+/* Optimized strncpy implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strncpy_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strncpy_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strncpy_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strncpy_power7) \
+ END_2(__strncpy_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#define MEMSET __memset_power7
+
+#include <sysdeps/powerpc/powerpc64/power7/strncpy.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define STRNCPY __strncpy_ppc
+#undef weak_alias
+#define weak_alias(name, aliasname) \
+ extern __typeof (__strncpy_ppc) aliasname \
+ __attribute__ ((weak, alias ("__strncpy_ppc")));
+#if !defined(NOT_IN_libc) && defined(SHARED)
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1(__strncpy_ppc, __GI_strncpy, __strncpy_ppc);
+#endif
+
+extern __typeof (strncpy) __strncpy_ppc attribute_hidden;
+
+#include <string/strncpy.c>
--- /dev/null
+/* Multiple versions of strncpy.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/ >. */
+
+/* Define multiple versions only for definition in libc. */
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strncpy) __strncpy_ppc attribute_hidden;
+extern __typeof (strncpy) __strncpy_power7 attribute_hidden;
+
+/* Avoid DWARF definition DIE on ifunc symbol so that GDB can handle
+ ifunc symbol properly. */
+libc_ifunc (strncpy,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strncpy_power7
+ : __strncpy_ppc);
+
+#endif
--- /dev/null
+/* Optimized strpbrk implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strpbrk_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strpbrk_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strpbrk_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strpbrk_power7) \
+ END_2(__strpbrk_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <sysdeps/powerpc/powerpc64/power7/strpbrk.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define STRPBRK __strpbrk_ppc
+#ifdef SHARED
+
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1 (__strpbrk_ppc, __GI_strpbrk, __strpbrk_ppc);
+#endif
+
+extern __typeof (strpbrk) __strpbrk_ppc attribute_hidden;
+
+#include <string/strpbrk.c>
--- /dev/null
+/* Multiple versions of strpbrk. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strpbrk) __strpbrk_ppc attribute_hidden;
+extern __typeof (strpbrk) __strpbrk_power7 attribute_hidden;
+
+libc_ifunc (strpbrk,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strpbrk_power7
+ : __strpbrk_ppc);
+#endif
--- /dev/null
+/* Optimized strrchr implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef ENTRY
+#define ENTRY(name) \
+ .section ".text"; \
+ ENTRY_2(__strrchr_power7) \
+ .align ALIGNARG(2); \
+ BODY_LABEL(__strrchr_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strrchr_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strrchr_power7) \
+ END_2(__strrchr_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <sysdeps/powerpc/powerpc64/power7/strrchr.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define STRRCHR __strrchr_ppc
+#undef weak_alias
+#define weak_alias(name, aliasname) \
+ extern __typeof (__strrchr_ppc) aliasname \
+ __attribute__ ((weak, alias ("__strrchr_ppc")));
+#if !defined(NOT_IN_libc) && defined(SHARED)
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1(__strrchr_ppc, __GI_strrchr, __strrchr_ppc);
+#endif
+
+extern __typeof (strrchr) __strrchr_ppc attribute_hidden;
+
+#include <string/strrchr.c>
--- /dev/null
+/* Multiple versions of strrchr. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* Define multiple versions only for definition in libc. */
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strrchr) __strrchr_ppc attribute_hidden;
+extern __typeof (strrchr) __strrchr_power7 attribute_hidden;
+
+/* Avoid DWARF definition DIE on ifunc symbol so that GDB can handle
+ ifunc symbol properly. */
+libc_ifunc (strrchr,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strrchr_power7
+ : __strrchr_ppc);
+weak_alias (strrchr, rindex)
+#endif
--- /dev/null
+/* Optimized strspn implementation for POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+#undef EALIGN
+#define EALIGN(name, alignt, words) \
+ .section ".text"; \
+ ENTRY_2(__strspn_power7) \
+ .align ALIGNARG(alignt); \
+ EALIGN_W_##words; \
+ BODY_LABEL(__strspn_power7): \
+ cfi_startproc; \
+ LOCALENTRY(__strspn_power7)
+
+#undef END
+#define END(name) \
+ cfi_endproc; \
+ TRACEBACK(__strspn_power7) \
+ END_2(__strspn_power7)
+
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name)
+
+#include <sysdeps/powerpc/powerpc64/power7/strspn.S>
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <string.h>
+
+#define STRSPN __strspn_ppc
+#undef weak_alias
+#define weak_alias(name, aliasname) \
+ extern __typeof (__strspn_ppc) aliasname \
+ __attribute__ ((weak, alias ("__strspn_ppc")));
+#if !defined(NOT_IN_libc) && defined(SHARED)
+# undef libc_hidden_builtin_def
+# define libc_hidden_builtin_def(name) \
+ __hidden_ver1(__strspn_ppc, __GI_strspn, __strspn_ppc);
+#endif
+
+extern __typeof (strspn) __strspn_ppc attribute_hidden;
+
+#include <string/strspn.c>
--- /dev/null
+/* Multiple versions of strspn. PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef NOT_IN_libc
+# include <string.h>
+# include <shlib-compat.h>
+# include "init-arch.h"
+
+extern __typeof (strspn) __strspn_ppc attribute_hidden;
+extern __typeof (strspn) __strspn_power7 attribute_hidden;
+
+libc_ifunc (strspn,
+ (hwcap & PPC_FEATURE_HAS_VSX)
+ ? __strspn_power7
+ : __strspn_ppc);
+#endif
-powerpc/powerpc64/power5/fpu/multiarch
+powerpc/powerpc64/power5/fpu
-powerpc/powerpc64/power4/fpu/multiarch
+powerpc/powerpc64/power4/fpu/
--- /dev/null
+powerpc/powerpc64/power6/fpu
-sysdeps/powerpc/powerpc64/power6/fpu/multiarch
+powerpc/powerpc64/power6/fpu/multiarch
--- /dev/null
+powerpc/powerpc64/power6/multiarch
--- /dev/null
+/* Implemented at memmove.S */
-powerpc/powerpc64/power6/fpu/multiarch
+powerpc/powerpc64/power6/fpu
ble cr1, L(copy_LT_32) /* If move < 32 bytes use short move
code. */
-#ifdef __LITTLE_ENDIAN__
-/* In little-endian mode, power7 takes an alignment trap on any lxvd2x
- or stxvd2x crossing a 32-byte boundary, so ensure the aligned_copy
- loop is only used for quadword aligned copies. */
+/* Align copies using VSX instructions to quadword. It is to avoid alignment
+ traps when memcpy is used on non-cacheable memory (for instance, memory
+ mapped I/O). */
andi. 10,3,15
clrldi 11,4,60
-#else
- andi. 10,3,7 /* Check alignment of DST. */
- clrldi 11,4,61 /* Check alignment of SRC. */
-#endif
cmpld cr6,10,11 /* SRC and DST alignments match? */
mr dst,3
beq L(aligned_copy)
mtocrf 0x01,0
-#ifdef __LITTLE_ENDIAN__
clrldi 0,0,60
-#else
- clrldi 0,0,61
-#endif
-/* Get the DST and SRC aligned to 8 bytes (16 for little-endian). */
+/* Get the DST and SRC aligned to 16 bytes. */
1:
bf 31,2f
lbz 6,0(src)
stw 6,0(dst)
addi dst,dst,4
8:
-#ifdef __LITTLE_ENDIAN__
bf 28,16f
ld 6,0(src)
addi src,src,8
std 6,0(dst)
addi dst,dst,8
16:
-#endif
subf cnt,0,cnt
/* Main aligned copy loop. Copies 128 bytes at a time. */
.align 4
L(copy_GE_32_unaligned):
clrldi 0,0,60 /* Number of bytes until the 1st dst quadword. */
-#ifndef __LITTLE_ENDIAN__
- andi. 10,3,15 /* Check alignment of DST (against quadwords). */
-#endif
srdi 9,cnt,4 /* Number of full quadwords remaining. */
beq L(copy_GE_32_unaligned_cont)
--- /dev/null
+/* Optimized memmove implementation for PowerPC64/POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+
+/* void* [r3] memmove (void *dest [r3], const void *src [r4], size_t len [r5])
+
+ This optimization check if memory 'dest' overlaps with 'src'. If it does
+ not then it calls an optimized memcpy call (similar to memcpy for POWER7,
+ embedded here to gain some cycles).
+ If source and destiny overlaps, a optimized backwards memcpy is used
+ instead. */
+
+ .machine power7
+EALIGN (memmove, 5, 0)
+ CALL_MCOUNT 3
+
+L(_memmove):
+ subf r9,r4,r3
+ cmpld cr7,r9,r5
+ blt cr7,L(memmove_bwd)
+
+ cmpldi cr1,r5,31
+ neg 0,3
+ ble cr1, L(copy_LT_32) /* If move < 32 bytes use short move
+ code. */
+
+ andi. 10,3,15
+ clrldi 11,4,60
+ cmpld cr6,10,11 /* SRC and DST alignments match? */
+
+ mr r11,3
+ bne cr6,L(copy_GE_32_unaligned)
+ beq L(aligned_copy)
+
+ mtocrf 0x01,0
+ clrldi 0,0,60
+
+/* Get the DST and SRC aligned to 8 bytes (16 for little-endian). */
+1:
+ bf 31,2f
+ lbz 6,0(r4)
+ addi r4,r4,1
+ stb 6,0(r11)
+ addi r11,r11,1
+2:
+ bf 30,4f
+ lhz 6,0(r4)
+ addi r4,r4,2
+ sth 6,0(r11)
+ addi r11,r11,2
+4:
+ bf 29,8f
+ lwz 6,0(r4)
+ addi r4,r4,4
+ stw 6,0(r11)
+ addi r11,r11,4
+8:
+ bf 28,16f
+ ld 6,0(r4)
+ addi r4,r4,8
+ std 6,0(r11)
+ addi r11,r11,8
+16:
+ subf r5,0,r5
+
+/* Main aligned copy loop. Copies 128 bytes at a time. */
+L(aligned_copy):
+ li 6,16
+ li 7,32
+ li 8,48
+ mtocrf 0x02,r5
+ srdi 12,r5,7
+ cmpdi 12,0
+ beq L(aligned_tail)
+ lxvd2x 6,0,r4
+ lxvd2x 7,r4,6
+ mtctr 12
+ b L(aligned_128loop)
+
+ .align 4
+L(aligned_128head):
+ /* for the 2nd + iteration of this loop. */
+ lxvd2x 6,0,r4
+ lxvd2x 7,r4,6
+L(aligned_128loop):
+ lxvd2x 8,r4,7
+ lxvd2x 9,r4,8
+ stxvd2x 6,0,r11
+ addi r4,r4,64
+ stxvd2x 7,r11,6
+ stxvd2x 8,r11,7
+ stxvd2x 9,r11,8
+ lxvd2x 6,0,r4
+ lxvd2x 7,r4,6
+ addi r11,r11,64
+ lxvd2x 8,r4,7
+ lxvd2x 9,r4,8
+ addi r4,r4,64
+ stxvd2x 6,0,r11
+ stxvd2x 7,r11,6
+ stxvd2x 8,r11,7
+ stxvd2x 9,r11,8
+ addi r11,r11,64
+ bdnz L(aligned_128head)
+
+L(aligned_tail):
+ mtocrf 0x01,r5
+ bf 25,32f
+ lxvd2x 6,0,r4
+ lxvd2x 7,r4,6
+ lxvd2x 8,r4,7
+ lxvd2x 9,r4,8
+ addi r4,r4,64
+ stxvd2x 6,0,r11
+ stxvd2x 7,r11,6
+ stxvd2x 8,r11,7
+ stxvd2x 9,r11,8
+ addi r11,r11,64
+32:
+ bf 26,16f
+ lxvd2x 6,0,r4
+ lxvd2x 7,r4,6
+ addi r4,r4,32
+ stxvd2x 6,0,r11
+ stxvd2x 7,r11,6
+ addi r11,r11,32
+16:
+ bf 27,8f
+ lxvd2x 6,0,r4
+ addi r4,r4,16
+ stxvd2x 6,0,r11
+ addi r11,r11,16
+8:
+ bf 28,4f
+ ld 6,0(r4)
+ addi r4,r4,8
+ std 6,0(r11)
+ addi r11,r11,8
+4: /* Copies 4~7 bytes. */
+ bf 29,L(tail2)
+ lwz 6,0(r4)
+ stw 6,0(r11)
+ bf 30,L(tail5)
+ lhz 7,4(r4)
+ sth 7,4(r11)
+ bflr 31
+ lbz 8,6(r4)
+ stb 8,6(r11)
+ /* Return original DST pointer. */
+ blr
+
+/* Handle copies of 0~31 bytes. */
+ .align 4
+L(copy_LT_32):
+ mr r11,3
+ cmpldi cr6,r5,8
+ mtocrf 0x01,r5
+ ble cr6,L(copy_LE_8)
+
+ /* At least 9 bytes to go. */
+ neg 8,4
+ andi. 0,8,3
+ cmpldi cr1,r5,16
+ beq L(copy_LT_32_aligned)
+
+ /* Force 4-byte alignment for SRC. */
+ mtocrf 0x01,0
+ subf r5,0,r5
+2:
+ bf 30,1f
+ lhz 6,0(r4)
+ addi r4,r4,2
+ sth 6,0(r11)
+ addi r11,r11,2
+1:
+ bf 31,L(end_4bytes_alignment)
+ lbz 6,0(r4)
+ addi r4,r4,1
+ stb 6,0(r11)
+ addi r11,r11,1
+
+ .align 4
+L(end_4bytes_alignment):
+ cmpldi cr1,r5,16
+ mtocrf 0x01,r5
+
+L(copy_LT_32_aligned):
+ /* At least 6 bytes to go, and SRC is word-aligned. */
+ blt cr1,8f
+
+ /* Copy 16 bytes. */
+ lwz 6,0(r4)
+ lwz 7,4(r4)
+ stw 6,0(r11)
+ lwz 8,8(r4)
+ stw 7,4(r11)
+ lwz 6,12(r4)
+ addi r4,r4,16
+ stw 8,8(r11)
+ stw 6,12(r11)
+ addi r11,r11,16
+8: /* Copy 8 bytes. */
+ bf 28,L(tail4)
+ lwz 6,0(r4)
+ lwz 7,4(r4)
+ addi r4,r4,8
+ stw 6,0(r11)
+ stw 7,4(r11)
+ addi r11,r11,8
+
+ .align 4
+/* Copies 4~7 bytes. */
+L(tail4):
+ bf 29,L(tail2)
+ lwz 6,0(r4)
+ stw 6,0(r11)
+ bf 30,L(tail5)
+ lhz 7,4(r4)
+ sth 7,4(r11)
+ bflr 31
+ lbz 8,6(r4)
+ stb 8,6(r11)
+ /* Return original DST pointer. */
+ blr
+
+ .align 4
+/* Copies 2~3 bytes. */
+L(tail2):
+ bf 30,1f
+ lhz 6,0(r4)
+ sth 6,0(r11)
+ bflr 31
+ lbz 7,2(r4)
+ stb 7,2(r11)
+ blr
+
+ .align 4
+L(tail5):
+ bflr 31
+ lbz 6,4(r4)
+ stb 6,4(r11)
+ blr
+
+ .align 4
+1:
+ bflr 31
+ lbz 6,0(r4)
+ stb 6,0(r11)
+ /* Return original DST pointer. */
+ blr
+
+/* Handles copies of 0~8 bytes. */
+ .align 4
+L(copy_LE_8):
+ bne cr6,L(tail4)
+
+ /* Though we could've used ld/std here, they are still
+ slow for unaligned cases. */
+
+ lwz 6,0(r4)
+ lwz 7,4(r4)
+ stw 6,0(r11)
+ stw 7,4(r11)
+ blr
+
+
+/* Handle copies of 32+ bytes where DST is aligned (to quadword) but
+ SRC is not. Use aligned quadword loads from SRC, shifted to realign
+ the data, allowing for aligned DST stores. */
+ .align 4
+L(copy_GE_32_unaligned):
+ clrldi 0,0,60 /* Number of bytes until the 1st r11 quadword. */
+ srdi 9,r5,4 /* Number of full quadwords remaining. */
+
+ beq L(copy_GE_32_unaligned_cont)
+
+ /* DST is not quadword aligned, get it aligned. */
+
+ mtocrf 0x01,0
+ subf r5,0,r5
+
+ /* Vector instructions work best when proper alignment (16-bytes)
+ is present. Move 0~15 bytes as needed to get DST quadword-aligned. */
+1:
+ bf 31,2f
+ lbz 6,0(r4)
+ addi r4,r4,1
+ stb 6,0(r11)
+ addi r11,r11,1
+2:
+ bf 30,4f
+ lhz 6,0(r4)
+ addi r4,r4,2
+ sth 6,0(r11)
+ addi r11,r11,2
+4:
+ bf 29,8f
+ lwz 6,0(r4)
+ addi r4,r4,4
+ stw 6,0(r11)
+ addi r11,r11,4
+8:
+ bf 28,0f
+ ld 6,0(r4)
+ addi r4,r4,8
+ std 6,0(r11)
+ addi r11,r11,8
+0:
+ srdi 9,r5,4 /* Number of full quadwords remaining. */
+
+ /* The proper alignment is present, it is OK to copy the bytes now. */
+L(copy_GE_32_unaligned_cont):
+
+ /* Setup two indexes to speed up the indexed vector operations. */
+ clrldi 10,r5,60
+ li 6,16 /* Index for 16-bytes offsets. */
+ li 7,32 /* Index for 32-bytes offsets. */
+ cmpldi cr1,10,0
+ srdi 8,r5,5 /* Setup the loop counter. */
+ mtocrf 0x01,9
+ cmpldi cr6,9,1
+#ifdef __LITTLE_ENDIAN__
+ lvsr 5,0,r4
+#else
+ lvsl 5,0,r4
+#endif
+ lvx 3,0,r4
+ li 0,0
+ bf 31,L(setup_unaligned_loop)
+
+ /* Copy another 16 bytes to align to 32-bytes due to the loop. */
+ lvx 4,r4,6
+#ifdef __LITTLE_ENDIAN__
+ vperm 6,4,3,5
+#else
+ vperm 6,3,4,5
+#endif
+ addi r4,r4,16
+ stvx 6,0,r11
+ addi r11,r11,16
+ vor 3,4,4
+ clrrdi 0,r4,60
+
+L(setup_unaligned_loop):
+ mtctr 8
+ ble cr6,L(end_unaligned_loop)
+
+ /* Copy 32 bytes at a time using vector instructions. */
+ .align 4
+L(unaligned_loop):
+
+ /* Note: vr6/vr10 may contain data that was already copied,
+ but in order to get proper alignment, we may have to copy
+ some portions again. This is faster than having unaligned
+ vector instructions though. */
+
+ lvx 4,r4,6
+#ifdef __LITTLE_ENDIAN__
+ vperm 6,4,3,5
+#else
+ vperm 6,3,4,5
+#endif
+ lvx 3,r4,7
+#ifdef __LITTLE_ENDIAN__
+ vperm 10,3,4,5
+#else
+ vperm 10,4,3,5
+#endif
+ addi r4,r4,32
+ stvx 6,0,r11
+ stvx 10,r11,6
+ addi r11,r11,32
+ bdnz L(unaligned_loop)
+
+ clrrdi 0,r4,60
+
+ .align 4
+L(end_unaligned_loop):
+
+ /* Check for tail bytes. */
+ mtocrf 0x01,r5
+ beqlr cr1
+
+ add r4,r4,0
+
+ /* We have 1~15 tail bytes to copy, and DST is quadword aligned. */
+ /* Copy 8 bytes. */
+ bf 28,4f
+ lwz 6,0(r4)
+ lwz 7,4(r4)
+ addi r4,r4,8
+ stw 6,0(r11)
+ stw 7,4(r11)
+ addi r11,r11,8
+4: /* Copy 4~7 bytes. */
+ bf 29,L(tail2)
+ lwz 6,0(r4)
+ stw 6,0(r11)
+ bf 30,L(tail5)
+ lhz 7,4(r4)
+ sth 7,4(r11)
+ bflr 31
+ lbz 8,6(r4)
+ stb 8,6(r11)
+ /* Return original DST pointer. */
+ blr
+
+ /* Start to memcpy backward implementation: the algorith first check if
+ src and dest have the same alignment and if it does align both to 16
+ bytes and copy using VSX instructions.
+ If does not, align dest to 16 bytes and use VMX (altivec) instruction
+ to read two 16 bytes at time, shift/permute the bytes read and write
+ aligned to dest. */
+L(memmove_bwd):
+ cmpldi cr1,r5,31
+ /* Copy is done backwards: update the pointers and check alignment. */
+ add r11,r3,r5
+ add r4,r4,r5
+ mr r0,r11
+ ble cr1, L(copy_LT_32_bwd) /* If move < 32 bytes use short move
+ code. */
+
+ andi. r10,r11,15 /* Check if r11 is aligned to 16 bytes */
+ clrldi r9,r4,60 /* Check if r4 is aligned to 16 bytes */
+ cmpld cr6,r10,r9 /* SRC and DST alignments match? */
+
+ bne cr6,L(copy_GE_32_unaligned_bwd)
+ beq L(aligned_copy_bwd)
+
+ mtocrf 0x01,r0
+ clrldi r0,r0,60
+
+/* Get the DST and SRC aligned to 16 bytes. */
+1:
+ bf 31,2f
+ lbz r6,-1(r4)
+ subi r4,r4,1
+ stb r6,-1(r11)
+ subi r11,r11,1
+2:
+ bf 30,4f
+ lhz r6,-2(r4)
+ subi r4,r4,2
+ sth r6,-2(r11)
+ subi r11,r11,2
+4:
+ bf 29,8f
+ lwz r6,-4(r4)
+ subi r4,r4,4
+ stw r6,-4(r11)
+ subi r11,r11,4
+8:
+ bf 28,16f
+ ld r6,-8(r4)
+ subi r4,r4,8
+ std r6,-8(r11)
+ subi r11,r11,8
+16:
+ subf r5,0,r5
+
+/* Main aligned copy loop. Copies 128 bytes at a time. */
+L(aligned_copy_bwd):
+ li r6,-16
+ li r7,-32
+ li r8,-48
+ li r9,-64
+ mtocrf 0x02,r5
+ srdi r12,r5,7
+ cmpdi r12,0
+ beq L(aligned_tail_bwd)
+ lxvd2x v6,r4,r6
+ lxvd2x v7,r4,r7
+ mtctr 12
+ b L(aligned_128loop_bwd)
+
+ .align 4
+L(aligned_128head_bwd):
+ /* for the 2nd + iteration of this loop. */
+ lxvd2x v6,r4,r6
+ lxvd2x v7,r4,r7
+L(aligned_128loop_bwd):
+ lxvd2x v8,r4,r8
+ lxvd2x v9,r4,r9
+ stxvd2x v6,r11,r6
+ subi r4,r4,64
+ stxvd2x v7,r11,r7
+ stxvd2x v8,r11,r8
+ stxvd2x v9,r11,r9
+ lxvd2x v6,r4,r6
+ lxvd2x v7,r4,7
+ subi r11,r11,64
+ lxvd2x v8,r4,r8
+ lxvd2x v9,r4,r9
+ subi r4,r4,64
+ stxvd2x v6,r11,r6
+ stxvd2x v7,r11,r7
+ stxvd2x v8,r11,r8
+ stxvd2x v9,r11,r9
+ subi r11,r11,64
+ bdnz L(aligned_128head_bwd)
+
+L(aligned_tail_bwd):
+ mtocrf 0x01,r5
+ bf 25,32f
+ lxvd2x v6,r4,r6
+ lxvd2x v7,r4,r7
+ lxvd2x v8,r4,r8
+ lxvd2x v9,r4,r9
+ subi r4,r4,64
+ stxvd2x v6,r11,r6
+ stxvd2x v7,r11,r7
+ stxvd2x v8,r11,r8
+ stxvd2x v9,r11,r9
+ subi r11,r11,64
+32:
+ bf 26,16f
+ lxvd2x v6,r4,r6
+ lxvd2x v7,r4,r7
+ subi r4,r4,32
+ stxvd2x v6,r11,r6
+ stxvd2x v7,r11,r7
+ subi r11,r11,32
+16:
+ bf 27,8f
+ lxvd2x v6,r4,r6
+ subi r4,r4,16
+ stxvd2x v6,r11,r6
+ subi r11,r11,16
+8:
+ bf 28,4f
+ ld r6,-8(r4)
+ subi r4,r4,8
+ std r6,-8(r11)
+ subi r11,r11,8
+4: /* Copies 4~7 bytes. */
+ bf 29,L(tail2_bwd)
+ lwz r6,-4(r4)
+ stw r6,-4(r11)
+ bf 30,L(tail5_bwd)
+ lhz r7,-6(r4)
+ sth r7,-6(r11)
+ bflr 31
+ lbz r8,-7(r4)
+ stb r8,-7(r11)
+ /* Return original DST pointer. */
+ blr
+
+/* Handle copies of 0~31 bytes. */
+ .align 4
+L(copy_LT_32_bwd):
+ cmpldi cr6,r5,8
+ mtocrf 0x01,r5
+ ble cr6,L(copy_LE_8_bwd)
+
+ /* At least 9 bytes to go. */
+ neg r8,r4
+ andi. r0,r8,3
+ cmpldi cr1,r5,16
+ beq L(copy_LT_32_aligned_bwd)
+
+ /* Force 4-byte alignment for SRC. */
+ mtocrf 0x01,0
+ subf r5,0,r5
+2:
+ bf 30,1f
+ lhz r6,-2(r4)
+ subi r4,r4,2
+ sth r6,-2(r11)
+ subi r11,r11,2
+1:
+ bf 31,L(end_4bytes_alignment_bwd)
+ lbz 6,-1(r4)
+ subi r4,r4,1
+ stb 6,-1(r11)
+ subi r11,r11,1
+
+ .align 4
+L(end_4bytes_alignment_bwd):
+ cmpldi cr1,r5,16
+ mtocrf 0x01,r5
+
+L(copy_LT_32_aligned_bwd):
+ /* At least 6 bytes to go, and SRC is word-aligned. */
+ blt cr1,8f
+
+ /* Copy 16 bytes. */
+ lwz r6,-4(r4)
+ lwz r7,-8(r4)
+ stw r6,-4(r11)
+ lwz r8,-12(r4)
+ stw r7,-8(r11)
+ lwz r6,-16(r4)
+ subi r4,r4,16
+ stw r8,-12(r11)
+ stw r6,-16(r11)
+ subi r11,r11,16
+8: /* Copy 8 bytes. */
+ bf 28,L(tail4_bwd)
+ lwz r6,-4(r4)
+ lwz r7,-8(r4)
+ subi r4,r4,8
+ stw r6,-4(r11)
+ stw r7,-8(r11)
+ subi r11,r11,8
+
+ .align 4
+/* Copies 4~7 bytes. */
+L(tail4_bwd):
+ bf 29,L(tail2_bwd)
+ lwz 6,-4(r4)
+ stw 6,-4(r11)
+ bf 30,L(tail5_bwd)
+ lhz 7,-6(r4)
+ sth 7,-6(r11)
+ bflr 31
+ lbz 8,-7(r4)
+ stb 8,-7(r11)
+ /* Return original DST pointer. */
+ blr
+
+ .align 4
+/* Copies 2~3 bytes. */
+L(tail2_bwd):
+ bf 30,1f
+ lhz 6,-2(r4)
+ sth 6,-2(r11)
+ bflr 31
+ lbz 7,-3(r4)
+ stb 7,-3(r11)
+ blr
+
+ .align 4
+L(tail5_bwd):
+ bflr 31
+ lbz 6,-5(r4)
+ stb 6,-5(r11)
+ blr
+
+ .align 4
+1:
+ bflr 31
+ lbz 6,-1(r4)
+ stb 6,-1(r11)
+ /* Return original DST pointer. */
+ blr
+
+
+/* Handles copies of 0~8 bytes. */
+ .align 4
+L(copy_LE_8_bwd):
+ bne cr6,L(tail4_bwd)
+
+ /* Though we could've used ld/std here, they are still
+ slow for unaligned cases. */
+ lwz 6,-8(r4)
+ lwz 7,-4(r4)
+ stw 6,-8(r11)
+ stw 7,-4(r11)
+ blr
+
+
+/* Handle copies of 32+ bytes where DST is aligned (to quadword) but
+ SRC is not. Use aligned quadword loads from SRC, shifted to realign
+ the data, allowing for aligned DST stores. */
+ .align 4
+L(copy_GE_32_unaligned_bwd):
+ andi. r10,r11,15 /* Check alignment of DST against 16 bytes.. */
+ srdi r9,r5,4 /* Number of full quadwords remaining. */
+
+ beq L(copy_GE_32_unaligned_cont_bwd)
+
+ /* DST is not quadword aligned and r10 holds the address masked to
+ compare alignments. */
+ mtocrf 0x01,r10
+ subf r5,r10,r5
+
+ /* Vector instructions work best when proper alignment (16-bytes)
+ is present. Move 0~15 bytes as needed to get DST quadword-aligned. */
+1:
+ bf 31,2f
+ lbz r6,-1(r4)
+ subi r4,r4,1
+ stb r6,-1(r11)
+ subi r11,r11,1
+2:
+ bf 30,4f
+ lhz r6,-2(r4)
+ subi r4,r4,2
+ sth r6,-2(r11)
+ subi r11,r11,2
+4:
+ bf 29,8f
+ lwz r6,-4(r4)
+ subi r4,r4,4
+ stw r6,-4(r11)
+ subi r11,r11,4
+8:
+ bf 28,0f
+ ld r6,-8(r4)
+ subi r4,r4,8
+ std r6,-8(r11)
+ subi r11,r11,8
+0:
+ srdi r9,r5,4 /* Number of full quadwords remaining. */
+
+ /* The proper alignment is present, it is OK to copy the bytes now. */
+L(copy_GE_32_unaligned_cont_bwd):
+
+ /* Setup two indexes to speed up the indexed vector operations. */
+ clrldi r10,r5,60
+ li r6,-16 /* Index for 16-bytes offsets. */
+ li r7,-32 /* Index for 32-bytes offsets. */
+ cmpldi cr1,10,0
+ srdi r8,r5,5 /* Setup the loop counter. */
+ mtocrf 0x01,9
+ cmpldi cr6,r9,1
+#ifdef __LITTLE_ENDIAN__
+ lvsr v5,r0,r4
+#else
+ lvsl v5,r0,r4
+#endif
+ lvx v3,0,r4
+ li r0,0
+ bf 31,L(setup_unaligned_loop_bwd)
+
+ /* Copy another 16 bytes to align to 32-bytes due to the loop. */
+ lvx v4,r4,r6
+#ifdef __LITTLE_ENDIAN__
+ vperm v6,v3,v4,v5
+#else
+ vperm v6,v4,v3,v5
+#endif
+ subi r4,r4,16
+ stvx v6,r11,r6
+ subi r11,r11,16
+ vor v3,v4,v4
+ clrrdi r0,r4,60
+
+L(setup_unaligned_loop_bwd):
+ mtctr r8
+ ble cr6,L(end_unaligned_loop_bwd)
+
+ /* Copy 32 bytes at a time using vector instructions. */
+ .align 4
+L(unaligned_loop_bwd):
+
+ /* Note: vr6/vr10 may contain data that was already copied,
+ but in order to get proper alignment, we may have to copy
+ some portions again. This is faster than having unaligned
+ vector instructions though. */
+
+ lvx v4,r4,r6
+#ifdef __LITTLE_ENDIAN__
+ vperm v6,v3,v4,v5
+#else
+ vperm v6,v4,v3,v5
+#endif
+ lvx v3,r4,r7
+#ifdef __LITTLE_ENDIAN__
+ vperm v10,v4,v3,v5
+#else
+ vperm v10,v3,v4,v5
+#endif
+ subi r4,r4,32
+ stvx v6,r11,r6
+ stvx v10,r11,r7
+ subi r11,r11,32
+ bdnz L(unaligned_loop_bwd)
+
+ clrrdi r0,r4,60
+
+ .align 4
+L(end_unaligned_loop_bwd):
+
+ /* Check for tail bytes. */
+ mtocrf 0x01,r5
+ beqlr cr1
+
+ add r4,r4,0
+
+ /* We have 1~15 tail bytes to copy, and DST is quadword aligned. */
+ /* Copy 8 bytes. */
+ bf 28,4f
+ lwz r6,-4(r4)
+ lwz r7,-8(r4)
+ subi r4,r4,8
+ stw r6,-4(r11)
+ stw r7,-8(r11)
+ subi r11,r11,8
+4: /* Copy 4~7 bytes. */
+ bf 29,L(tail2_bwd)
+ lwz r6,-4(r4)
+ stw r6,-4(r11)
+ bf 30,L(tail5_bwd)
+ lhz r7,-6(r4)
+ sth r7,-6(r11)
+ bflr 31
+ lbz r8,-7(r4)
+ stb r8,-7(r11)
+ /* Return original DST pointer. */
+ blr
+END_GEN_TB (memmove, TB_TOCLESS)
+libc_hidden_builtin_def (memmove)
+
+
+/* void bcopy(const void *src [r3], void *dest [r4], size_t n [r5])
+ Implemented in this file to avoid linker create a stub function call
+ in the branch to '_memmove'. */
+ENTRY (bcopy)
+ mr r6,r3
+ mr r3,r4
+ mr r4,r6
+ b L(_memmove)
+END (bcopy)
mr r10,r3
clrrdi r6,r7,7
li r9,3<<5
- dcbt r9,r6,16 /* Stream hint, decreasing addresses. */
+ dcbt r9,r6,8 /* Stream hint, decreasing addresses. */
/* Replicate BYTE to doubleword. */
insrdi r4,r4,8,48
--- /dev/null
+/* Optimized stpncpy implementation for PowerPC64/POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#define USE_AS_STPNCPY
+#include <sysdeps/powerpc/powerpc64/power7/strncpy.S>
+
+weak_alias (__stpncpy, stpncpy)
+libc_hidden_def (__stpncpy)
+libc_hidden_builtin_def (stpncpy)
--- /dev/null
+/* Optimized strcmp implementation for Power7 using 'cmpb' instruction
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* The optimization is achieved here through cmpb instruction.
+ 8byte aligned strings are processed with double word comparision
+ and unaligned strings are handled effectively with loop unrolling
+ technique */
+
+#include <sysdep.h>
+
+/* int [r3] strcmp (const char *s1 [r3], const char *s2 [r4]) */
+
+EALIGN (strcmp, 4, 0)
+ CALL_MCOUNT 2
+
+ or r9, r3, r4
+ rldicl. r10, r9, 0, 61 /* are s1 and s2 8 byte aligned..? */
+ bne cr0, L(process_unaligned_bytes)
+
+/* process input parameters on double word aligned boundary */
+ ld r9, 0(r4) /* load s2 at offset=0 */
+ li r10, 0 /* load mask=0 */
+ cmpb r10, r9, r10 /* compare bytes at s2 with mask */
+ cmpdi cr7, r10, 0 /* is NULL found ..? is end of string HIT */
+ bne cr7, L(process_unaligned_bytes) /* process byte by byte */
+
+ ld r10, 0(r3) /* load s1 at offset=0 */
+ li r8, 0 /* load mask=0 */
+ cmpb r8, r10, r8 /* compare bytes at s1 with mask */
+ cmpdi cr7, r8, 0 /* is NULL found ..? is end of string HIT */
+ bne cr7, L(process_unaligned_bytes) /* process byte by byte */
+
+/*s1 and s2 does not contain NULL now , so compare all 8 bytes in a GO */
+ cmpb r9, r10, r9 /* compare s1 and s2 */
+ cmpdi cr7, r9, -1 /* compare result with 0xFFFFFFFFFFFFFFFF */
+ bne cr7, L(process_unaligned_bytes) /* s1,s2 mismatch found */
+
+ addi r5, r3, 8 /* save next offset of s2 */
+ addi r11, r4, 8 /* save next offset of s1 */
+ ld r8, 8(r4) /* load s2 at offset=8 */
+ li r9, 0 /* load mask=0 */
+ cmpb r9, r8, r9 /* compare bytes at s2 with mask */
+ cmpdi cr7, r9, 0 /* NULL found ..? */
+ bne cr7, L(processBytes)/* update input and process bytes one by one */
+
+ mr r9, r4 /* save s2 */
+ li r10, 0 /* load mask=0 */
+
+ ld r7, 8(r3) /* load s1 at offset=8 */
+ cmpb r6, r7, r10 /* compare bytes at s1 with mask */
+ cmpdi cr7, r6, 0 /* is NULL found */
+ bne cr7, L(processBytes)/* mismatch, so process one by one */
+
+L(unrollDword):
+ cmpb r8, r7, r8 /* compare s1 and s2 */
+ cmpdi cr7, r8, -1 /* compare result with 0xFFFFFFFFFFFFFFFF */
+ bne cr7, L(processBytes)/* mismatch with s1 and s2 */
+
+ addi r5, r3, 16 /* save offset=16 of s1 */
+ addi r4, r9, 16 /* save offset=16 of s2 */
+ ld r8, 16(r9) /* load s2 at offset=16 */
+ cmpb r7, r8, r10 /* compare bytes at s2 with mask */
+ cmpdi cr7, r7, 0 /* NULL found ..? */
+ bne cr7, L(update2processBytes)
+
+ ld r7, 16(r3) /* load s1 at offset=16 */
+ cmpb r6, r7, r10 /* check s1 for end of string */
+ cmpdi cr7, r6, 0 /* end of s1 ?,then handle byte by byte */
+ bne 7,L(update2processBytes)
+
+ cmpb r8, r7, r8 /* compare s1 and s2 double words */
+ cmpdi cr7, r8, -1 /* compare results with 0xFFFFFFFFFFFFFFFF */
+ bne cr7,L(update2processBytes)
+
+ addi r5, r3, 24 /* update s1 to offset=24 */
+ addi r4, r9, 24 /* update s2 to offset=24 */
+
+ ld r8, 24(r9) /* load s2 */
+ cmpb r7, r8, r10 /* compare s2 for NULL */
+ cmpdi cr7, r7, 0 /* verify if s2 is ending now */
+ bne cr7,L(update2processBytes)
+
+ ld r7, 24(r3) /* load s1 at offset=24 */
+ cmpb r6, r7, r10 /* verify for NULL */
+ cmpdi cr7, r6, 0 /* is NULL found */
+ bne cr7, L(update2processBytes)
+
+ cmpb r8, r7, r8 /* compare s1 and s2 */
+ cmpdi cr7, r8, -1 /* are s1 and s2 same ..? */
+ bne cr7, L(update2processBytes)
+
+ addi r7, r9, 32 /* update s2 to next double word */
+ addi r3, r3, 32 /* update s1 to next double word */
+
+ ld r8, 32(r9) /* load s2 */
+ mr r4, r7 /* save s2 */
+ cmpb r6, r8, r10 /* compare s2 with NULL */
+ cmpdi cr7, r6, 0 /* end of s2 ..? */
+ bne cr7, L(process_unaligned_bytes)
+
+ ld r6, 0(r3) /* load and compare s1 for NULL */
+ cmpb r5, r6, r10
+ cmpdi cr7, r5, 0
+ bne cr7, L(process_unaligned_bytes)
+
+ cmpb r8, r6, r8 /* compare s1 and s2 */
+ cmpdi cr7, r8, -1
+ bne cr7, L(process_unaligned_bytes)
+
+ addi r5, r3, 8 /* increment s1 and d2 here */
+ addi r11, r9, 40
+
+ ld r8, 40(r9) /* process s2 now */
+ cmpb r9, r8, r10
+ cmpdi cr7, r9, 0
+ bne cr7, L(processBytes)
+
+ mr r9, r7
+ ld r7, 8(r3) /* process s1 now */
+ cmpb r6, r7, r10
+ cmpdi cr7, r6, 0
+ beq cr7, L(unrollDword) /* unroll to compare s1 and s2 */
+
+L(processBytes):
+ mr r4, r11 /* update input params */
+ mr r3, r5
+
+ .p2align 4
+L(process_unaligned_bytes):
+ lbz r9, 0(r3) /* load byte from s1 */
+ lbz r10, 0(r4) /* load byte from s2 */
+ cmpdi cr7, r9, 0 /* compare *s1 with NULL */
+ beq cr7, L(diffOfNULL) /* if *s1 is NULL , return *s1 - *s2 */
+ cmplw cr7, r9, r10 /* compare *s1 and *s2 */
+ bne cr7, L(ComputeDiff) /* branch to compute difference and return */
+
+ lbz r9, 1(r3) /* load next byte from s1 */
+ lbz r10, 1(r4) /* load next byte from s2 */
+ cmpdi cr7, r9, 0 /* compare *s1 with NULL */
+ beq cr7, L(diffOfNULL) /* if *s1 is NULL , return *s1 - *s2 */
+ cmplw cr7, r9, r10 /* compare *s1 and *s2 */
+ bne cr7, L(ComputeDiff) /* branch to compute difference and return */
+
+ lbz r9, 2(r3) /* unroll 3rd byte here */
+ lbz r10, 2(r4)
+ cmpdi cr7, r9, 0
+ beq cr7, L(diffOfNULL)
+ cmplw cr7, r9, r10
+ bne 7, L(ComputeDiff)
+
+ lbz r9, 3(r3) /* unroll 4th byte now */
+ lbz r10, 3(r4)
+ addi r3, r3, 4 /* increment s1 by unroll factor */
+ cmpdi cr7, r9, 0
+ cmplw cr6, 9, r10
+ beq cr7, L(diffOfNULL)
+ addi r4, r4, 4 /* increment s2 by unroll factor */
+ beq cr6, L(process_unaligned_bytes) /* unroll byte processing */
+
+ .p2align 4
+L(ComputeDiff):
+ extsw r9, r9
+ subf r10, r10, r9 /* compute s1 - s2 */
+ extsw r3, r10
+ blr /* return */
+
+ .p2align 4
+L(diffOfNULL):
+ li r9, 0
+ subf r10, r10, r9 /* compute s1 - s2 */
+ extsw r3, r10 /* sign extend result */
+ blr /* return */
+
+ .p2align 4
+L(update2processBytes):
+ mr r3, r5 /* update and proceed */
+ b L(process_unaligned_bytes)
+
+END (strcmp)
+libc_hidden_builtin_def (strcmp)
--- /dev/null
+/* Optimized strcspn implementation for PowerPC64.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+/* size_t [r3] strcspn (const char [r4] *s, const char [r5] *reject) */
+
+ .machine power7
+EALIGN (strcspn, 4, 0)
+ CALL_MCOUNT 3
+
+ /* The idea to speed up the algorithm is to create a lookup table
+ for fast check if input character should be considered. For ASCII
+ or ISO-8859-X character sets it has 256 positions. */
+ lbz r10,0(r4)
+
+ /* First the table should be cleared and to avoid unaligned accesses
+ when using the VSX stores the table address is aligned to 16
+ bytes. */
+ xxlxor v0,v0,v0
+
+ /* PPC64 ELF ABI stack is aligned to 16 bytes. */
+ addi r9,r1,-256
+
+ li r8,48
+ li r5,16
+ li r6,32
+ cmpdi cr7,r10,0 /* reject[0] == '\0' ? */
+ addi r12,r9,64
+ /* Clear the table with 0 values */
+ stxvw4x v0,r0,r9
+ addi r11,r9,128
+ addi r7,r9,192
+ stxvw4x v0,r9,r5
+ stxvw4x v0,r9,r6
+ stxvw4x v0,r9,r8
+ stxvw4x v0,r0,r12
+ stxvw4x v0,r12,r5
+ stxvw4x v0,r12,r6
+ stxvw4x v0,r12,r8
+ stxvw4x v0,r0,r11
+ stxvw4x v0,r11,r5
+ stxvw4x v0,r11,r6
+ stxvw4x v0,r11,r8
+ stxvw4x v0,r0,r7
+ stxvw4x v0,r7,r5
+ stxvw4x v0,r7,r6
+ stxvw4x v0,r7,r8
+ li r8,1
+ beq cr7,L(finish_table) /* If reject[0] == '\0' skip */
+
+ /* Initialize the table as:
+ for (i=0; reject[i]; i++
+ table[reject[i]]] = 1 */
+ .p2align 4,,15
+L(init_table):
+ stbx r8,r9,r10
+ lbzu r10,1(r4)
+ cmpdi cr7,r10,0 /* If reject[0] == '\0' finish */
+ bne cr7,L(init_table)
+L(finish_table):
+ /* set table[0] = 1 */
+ li r10,1
+ stb r10,0(r9)
+ li r10,0
+ b L(mainloop)
+
+ /* Unrool the loop 4 times and check using the table as:
+ i = 0;
+ while (1)
+ {
+ if (table[input[i++]] == 1)
+ return i - 1;
+ if (table[input[i++]] == 1)
+ return i - 1;
+ if (table[input[i++]] == 1)
+ return i - 1;
+ if (table[input[i++]] == 1)
+ return i - 1;
+ } */
+ .p2align 4,,15
+L(unroll):
+ lbz r8,1(r3)
+ addi r10,r10,4
+ lbzx r8,r9,r8
+ cmpwi r7,r8,1
+ beq cr7,L(end)
+ lbz r8,2(r3)
+ addi r3,r3,4
+ lbzx r8,r9,r8
+ cmpwi cr7,r8,1
+ beq cr7,L(end2)
+ lbz r8,3(r7)
+ lbzx r8,r9,r8
+ cmpwi cr7,r8,1
+ beq cr7,L(end3)
+L(mainloop):
+ lbz r8,0(r3)
+ mr r7,r3
+ addi r6,r10,1
+ addi r4,r10,2
+ addi r5,r10,3
+ lbzx r8,r9,8
+ cmpwi cr7,r8,1
+ bne cr7,L(unroll)
+ mr r3,r10
+ blr
+
+ .p2align 4,,15
+L(end):
+ mr r3,r6
+ blr
+
+ .p2align 4,,15
+L(end2):
+ mr r3,r4
+ blr
+
+ .p2align 4,,15
+L(end3):
+ mr r3,r5
+ blr
+END (strcspn)
+libc_hidden_builtin_def (strcspn)
--- /dev/null
+/* Optimized strncat implementation for PowerPC64/POWER7.
+
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* The algorithm is as follows for aligned memory access :
+
+ if address of s2 is divisible by 0x7UL,
+ perform aligned doubleword catenation
+ else
+ perform unaligned catenation
+
+ The aligned comparison are made using cmpb instructions. */
+
+/* char* [r3] strncat (const char *s1 [r3],
+ const char *s2 [r4],
+ size_t size [r5]) */
+
+#include <sysdep.h>
+
+#ifndef STRNCAT
+# undef strncat
+# define STRNCAT strncat
+#endif
+
+#ifndef STRLEN
+/* For builds with no IFUNC support, local calls should be made to internal
+ GLIBC symbol (created by libc_hidden_builtin_def). */
+# ifdef SHARED
+# define STRLEN __GI_strlen
+# else
+# define STRLEN strlen
+# endif
+#endif
+
+#define FRAMESIZE (FRAME_MIN_SIZE+32)
+
+ .machine power7
+EALIGN(STRNCAT, 4, 0)
+ CALL_MCOUNT 3
+
+ mflr r0 /* Load link register LR to r0. */
+
+/* We shall use r29, r30 and r31 non volatile register for retention.
+ Save all the callee registers in the GPR save area. */
+ std r29, -24(r1) /* Save callers register r29. */
+ std r30, -16(r1) /* Save callers register r30. */
+ std r31, -8(r1) /* Save callers register r31. */
+
+ std r0, 16(r1) /* Store the link register. */
+ stdu r1, -FRAMESIZE(r1) /* Create the stack frame. */
+
+/* Improve performance with CPU pre-fetch. */
+ dcbt 0, r3 /* Pre-fetch str to avoid cache
+ miss. */
+ dcbt 0, r4 /* Pre-fetch accept to avoid cache
+ miss. */
+
+ mr. r29, r5 /* Save "n" in r29. */
+ mr r30, r3 /* Save "s1" in r30 from r3. */
+ beq cr0,L(done)
+
+ mr r31, r4 /* Save "s2" in r31 from r4. */
+ bl STRLEN /* Call optimized strlen on s1; goto
+ end of s1. */
+ nop
+ cmpldi cr7, r29, 7 /* If s2 is <=7 process
+ byte-by-byte. */
+ add r3, r30, r3 /* Grab the last character of s1. */
+ bgt cr7,L(alignment) /* Process by aligned strings. */
+
+ cmpldi cr7, r29, 3 /* If n is >= 4, we can
+ byte-unroll. */
+ addi r9, r3, -1 /* Make "s1" point before next
+ character, increment when read. */
+ bgt cr7, L(bytes_unroll) /* Process each byte. */
+
+L(byte_by_byte):
+ lbz r10, 0(r31)
+ addi r8, r9, 1
+ cmpdi cr7, r10, 0 /* Check for NULL in "s2". */
+ stb r10, 1(r9)
+ beq cr7, L(done)
+ add r9, r9, r29
+ subf r9, r8, r9
+ addi r9, r9, 1
+ mtctr r9
+ b L(branch2)
+ .p2align 4
+L(branch1):
+ lbzu r10, 1(r31)
+ cmpdi cr7, r10, 0
+ stbu r10, 1(r8)
+ beq cr7,L(done)
+L(branch2):
+ mr r9, r8
+ bdnz L(branch1)
+ beq cr7,L(done)
+L(nullTerminate):
+ li r10, 0 /* Load NULL for termination. */
+ stb r10, 1(r9) /* Append or terminate s1 with
+ NULL. */
+ .p2align 4 /* A small section here. */
+L(done): /* We return now. */
+ addi r1, r1, FRAMESIZE /* Restore stack pointer. */
+ mr r3, r30 /* Set the return value length of
+ string. */
+ ld r0, 16(r1) /* Read the saved link register. */
+ ld r29, -24(r1) /* Restore save register r29. */
+ ld r30, -16(r1) /* Restore save register r30. */
+ ld r31, -8(r1) /* Restore save register r31. */
+ mtlr r0 /* Restore link register. */
+ blr /* Branch to link register. */
+
+ .p2align 4
+L(alignment):
+ rldicl. r9, r31, 0, 61 /* Check if s2 is 8byte aligned */
+ beq cr0,L(dwordAligned)
+
+ .p2align 4
+/* Unaligned bytes in string, so process byte by byte.
+ POWER7 has performance gains over loop unroll. */
+L(bytes_unroll):
+ addi r9, r3, -1
+ srdi r10, r29, 2
+ mtctr r10
+ b L(L10)
+ .p2align 4
+L(L44):
+ lbz r10, 1(r31) /* Load byte. */
+ cmpdi cr7, r10, 0 /* Compare ; if byte not zero,
+ continue. */
+ stb r10, 2(r9) /* Store byte */
+ beq cr7, L(done)
+ addi r31, r31, 4
+
+ lbz r10, -2(r31) /* Perform loop unroll here on byte
+ load and store. */
+ cmpdi cr7, r10, 0
+ stb r10, 3(r9)
+ beq cr7, L(done)
+
+ lbz r10, -1(r31) /* Loop unroll here. */
+ cmpdi cr7, r10, 0
+ stbu r10, 4(r9)
+ beq cr7, L(done)
+
+ bdz L(leftNbytes)
+
+L(L10):
+ lbz r10, 0(r31) /* Loop unroll here. */
+ cmpdi cr7, r10, 0
+ stb r10, 1(r9)
+ bne cr7,L(L44)
+ b L(done)
+ .p2align 4
+/* If s2 is double word aligned, we load and store double word. */
+L(dwordAligned):
+/* read, write 8 bytes at a time */
+ srdi r8, r29, 3 /* Compute count for CTR to loop;
+ count = n/8. */
+ li r7, 0 /* Load r7 with NULL. */
+ li r10, 0 /* Load r10 with MASK '0'. */
+
+ mtctr r8 /* Move count to CTR. */
+L(loop8):
+ ld r9, 0(r31) /* Read double word from s2. */
+ cmpb r6, r9, r10 /* Compare bytes in s2 we read
+ just now. */
+ cmpdi r6, 0 /* If cmpb returned NULL,
+ we continue. */
+ bne+ L(a8)
+ std r9, 0(r3) /* Append double word from s2
+ with s1. */
+ addi r3, r3, 8 /* Increment s1. */
+ addi r31, r31, 8 /* Increment s2. */
+ subi r29, r29, 8 /* Decrement count by 8. */
+ bdnz L(loop8) /* Continue until "count" is
+ non zero. */
+
+L(a8):
+ cmpdi r29, 0 /* If "n" is already zero, we skip. */
+ beq+ L(align8align)
+
+ mtctr r29 /* Process left over bytes in "n". */
+L(unaligned0):
+ lbz r9, 0(r31) /* Read a byte from s2. */
+ cmpw r9, r7 /* If byte is NULL, we stop here . */
+ beq+ L(align8align) /* Skip processing further if NULL. */
+ stb r9, 0(r3) /* If not NULL, store byte into s1. */
+ addi r3, r3, 1 /* Increment s1 by 1. */
+ addi r31, r31, 1 /* Increment s2 by 1. */
+ bdnz L(unaligned0) /* Decrement counter "n" and loop
+ until non zero. */
+L(align8align):
+ stb r7, 0(r3) /* Terminate s1 with NULL. */
+
+ addi r1, r1, FRAMESIZE /* Restore stack pointer. */
+ mr r3, r30 /* Set the return value, length of
+ string. */
+ ld r0, 16(r1) /* Read the saved link register. */
+ ld r29, -24(r1) /* Restore save register r29. */
+ ld r30, -16(r1) /* Restore save register r30. */
+ ld r31, -8(r1) /* Restore save register r31. */
+ mtlr r0 /* Restore link register. */
+ blr /* Branch to link register */
+
+ .p2align 4
+L(leftNbytes):
+ rldicl. r29, r29, 0, 62 /* Check if n>0 and n < 4 bytes. */
+ bne cr0,L(byte_by_byte) /* Process bytes one by one. */
+ b L(nullTerminate) /* Now, finish catenation with
+ NULL termination. */
+END(STRNCAT)
--- /dev/null
+/* Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+/* Implements the functions
+
+ char * [r3] strncpy (char *dst [r3], const char *src [r4], size_t n [r5])
+
+ AND
+
+ char * [r3] stpncpy (char *dst [r3], const char *src [r4], size_t n [r5])
+
+ The algorithm is as follows:
+ > if src and dest are 8 byte aligned, perform double word copy
+ else
+ > copy byte by byte on unaligned addresses.
+
+ The aligned comparison are made using cmpb instructions. */
+
+/* The focus on optimization for performance improvements are as follows:
+ 1. data alignment [gain from aligned memory access on read/write]
+ 2. POWER7 gains performance with loop unrolling/unwinding
+ [gain by reduction of branch penalty].
+ 3. The final pad with null bytes is done by calling an optimized
+ memset. */
+
+#ifdef USE_AS_STPNCPY
+# define FUNC_NAME __stpncpy
+#else
+# define FUNC_NAME strncpy
+#endif
+
+#define FRAMESIZE (FRAME_MIN_SIZE+32)
+
+#ifndef MEMSET
+/* For builds with no IFUNC support, local calls should be made to internal
+ GLIBC symbol (created by libc_hidden_builtin_def). */
+# ifdef SHARED
+# define MEMSET __GI_memset
+# else
+# define MEMSET memset
+# endif
+#endif
+
+ .machine power7
+EALIGN(FUNC_NAME, 4, 0)
+ CALL_MCOUNT 3
+
+ mflr r0 /* load link register LR to r0 */
+ or r10, r3, r4 /* to verify source and destination */
+ rldicl. r8, r10, 0, 61 /* is double word aligned .. ? */
+
+ std r19, -8(r1) /* save callers register , r19 */
+ std r18, -16(r1) /* save callers register , r18 */
+ std r0, 16(r1) /* store the link register */
+ stdu r1, -FRAMESIZE(r1) /* create the stack frame */
+
+ mr r9, r3 /* save r3 into r9 for use */
+ mr r18, r3 /* save r3 for retCode of strncpy */
+ bne 0, L(byte_by_byte)
+
+
+ srdi r11, r5, 3 /* compute count for CTR ; count = n/8 */
+ cmpldi cr7, r11, 3 /* if count > 4 ; perform unrolling 4 times */
+ ble 7, L(update1)
+
+ ld r10, 0(r4) /* load doubleWord from src */
+ cmpb r8, r10, r8 /* compare src with NULL ,we read just now */
+ cmpdi cr7, r8, 0 /* if cmpb returned NULL ; we continue */
+ bne cr7, L(update3)
+
+ std r10, 0(r3) /* copy doubleword at offset=0 */
+ ld r10, 8(r4) /* load next doubleword from offset=8 */
+ cmpb r8, r10, r8 /* compare src with NULL , we read just now */
+ cmpdi cr7, r8, 0 /* if cmpb returned NULL ; we continue */
+ bne 7,L(HopBy8)
+
+ addi r8, r11, -4
+ mr r7, r3
+ srdi r8, r8, 2
+ mr r6, r4
+ addi r8, r8, 1
+ li r12, 0
+ mtctr r8
+ b L(dwordCopy)
+
+ .p2align 4
+L(dWordUnroll):
+ std r8, 16(r9)
+ ld r8, 24(r4) /* load dword,perform loop unrolling again */
+ cmpb r10, r8, r10
+ cmpdi cr7, r10, 0
+ bne cr7, L(HopBy24)
+
+ std r8, 24(r7) /* copy dword at offset=24 */
+ addi r9, r9, 32
+ addi r4, r4, 32
+ bdz L(leftDwords) /* continue with loop on counter */
+
+ ld r3, 32(r6)
+ cmpb r8, r3, r10
+ cmpdi cr7, r8, 0
+ bne cr7, L(update2)
+
+ std r3, 32(r7)
+ ld r10, 40(r6)
+ cmpb r8, r10, r8
+ cmpdi cr7, r8, 0
+ bne cr7, L(HopBy40)
+
+ mr r6, r4 /* update values */
+ mr r7, r9
+ mr r11, r0
+ mr r5, r19
+
+L(dwordCopy):
+ std r10, 8(r9) /* copy dword at offset=8 */
+ addi r19, r5, -32
+ addi r0, r11, -4
+ ld r8, 16(r4)
+ cmpb r10, r8, r12
+ cmpdi cr7, r10, 0
+ beq cr7, L(dWordUnroll)
+
+ addi r9, r9, 16 /* increment dst by 16 */
+ addi r4, r4, 16 /* increment src by 16 */
+ addi r5, r5, -16 /* decrement length 'n' by 16 */
+ addi r0, r11, -2 /* decrement loop counter */
+
+L(dWordUnrollOFF):
+ ld r10, 0(r4) /* load first dword */
+ li r8, 0 /* load mask */
+ cmpb r8, r10, r8
+ cmpdi cr7, r8, 0
+ bne cr7, L(byte_by_byte)
+ mtctr r0
+ li r7, 0
+ b L(CopyDword)
+
+ .p2align 4
+L(loadDWordandCompare):
+ ld r10, 0(r4)
+ cmpb r8, r10, r7
+ cmpdi cr7, r8, 0
+ bne cr7, L(byte_by_byte)
+
+L(CopyDword):
+ addi r9, r9, 8
+ std r10, -8(r9)
+ addi r4, r4, 8
+ addi r5, r5, -8
+ bdnz L(loadDWordandCompare)
+
+L(byte_by_byte):
+ cmpldi cr7, r5, 3
+ ble cr7, L(verifyByte)
+ srdi r10, r5, 2
+ mr r19, r9
+ mtctr r10
+ b L(firstByteUnroll)
+
+ .p2align 4
+L(bytes_unroll):
+ lbz r10, 1(r4) /* load byte from src */
+ cmpdi cr7, r10, 0 /* compare for NULL */
+ stb r10, 1(r19) /* store byte to dst */
+ beq cr7, L(updtDestComputeN2ndByte)
+
+ addi r4, r4, 4 /* advance src */
+
+ lbz r10, -2(r4) /* perform loop unrolling for byte r/w */
+ cmpdi cr7, r10, 0
+ stb r10, 2(r19)
+ beq cr7, L(updtDestComputeN3rdByte)
+
+ lbz r10, -1(r4) /* perform loop unrolling for byte r/w */
+ addi r19, r19, 4
+ cmpdi cr7, r10, 0
+ stb r10, -1(r19)
+ beq cr7, L(ComputeNByte)
+
+ bdz L(update0)
+
+L(firstByteUnroll):
+ lbz r10, 0(r4) /* perform loop unrolling for byte r/w */
+ cmpdi cr7, 10, 0
+ stb r10, 0(r19)
+ bne cr7, L(bytes_unroll)
+ addi r19, r19, 1
+
+L(ComputeNByte):
+ subf r9, r19, r9 /* compute 'n'n bytes to fill */
+ add r8, r9, r5
+
+L(zeroFill):
+ cmpdi cr7, r8, 0 /* compare if length is zero */
+ beq cr7, L(update3return)
+
+ mr r3, r19 /* fill buffer with */
+ li r4, 0 /* zero fill buffer */
+ mr r5, r8 /* how many bytes to fill buffer with */
+ bl MEMSET /* call optimized memset */
+ nop
+
+L(update3return):
+#ifdef USE_AS_STPNCPY
+ addi r3, r19, -1 /* update return value */
+#endif
+
+L(hop2return):
+#ifndef USE_AS_STPNCPY
+ mr r3, r18 /* set return value */
+#endif
+ addi r1, r1, FRAMESIZE /* restore stack pointer */
+ ld r0, 16(r1) /* read the saved link register */
+ ld r18, -16(r1) /* restore callers save register, r18 */
+ ld r19, -8(r1) /* restore callers save register, r19 */
+ mtlr r0 /* branch to link register */
+ blr /* return */
+
+ .p2align 4
+L(update0):
+ mr r9, r19
+
+ .p2align 4
+L(verifyByte):
+ rldicl. r8, r5, 0, 62
+#ifdef USE_AS_STPNCPY
+ mr r3, r9
+#endif
+ beq cr0, L(hop2return)
+ mtctr r8
+ addi r4, r4, -1
+ mr r19, r9
+ b L(oneBYone)
+
+ .p2align 4
+L(proceed):
+ bdz L(done)
+
+L(oneBYone):
+ lbzu r10, 1(r4) /* copy byte */
+ addi r19, r19, 1
+ addi r8, r8, -1
+ cmpdi cr7, r10, 0
+ stb r10, -1(r19)
+ bne cr7, L(proceed)
+ b L(zeroFill)
+
+ .p2align 4
+L(done):
+ addi r1, r1, FRAMESIZE /* restore stack pointer */
+#ifdef USE_AS_STPNCPY
+ mr r3, r19 /* set the return value */
+#else
+ mr r3, r18 /* set the return value */
+#endif
+ ld r0, 16(r1) /* read the saved link register */
+ ld r18, -16(r1) /* restore callers save register, r18 */
+ ld r19, -8(r1) /* restore callers save register, r19 */
+ mtlr r0 /* branch to link register */
+ blr /* return */
+
+L(update1):
+ mr r0, r11
+ mr r19, r5
+
+ .p2align 4
+L(leftDwords):
+ cmpdi cr7, r0, 0
+ mr r5, r19
+ bne cr7, L(dWordUnrollOFF)
+ b L(byte_by_byte)
+
+ .p2align 4
+L(updtDestComputeN2ndByte):
+ addi r19, r19, 2 /* update dst by 2 */
+ subf r9, r19, r9 /* compute distance covered */
+ add r8, r9, r5
+ b L(zeroFill)
+
+ .p2align 4
+L(updtDestComputeN3rdByte):
+ addi r19, r19, 3 /* update dst by 3 */
+ subf r9, r19, r9 /* compute distance covered */
+ add r8, r9, r5
+ b L(zeroFill)
+
+ .p2align 4
+L(HopBy24):
+ addi r9, r9, 24 /* increment dst by 24 */
+ addi r4, r4, 24 /* increment src by 24 */
+ addi r5, r5, -24 /* decrement length 'n' by 24 */
+ addi r0, r11, -3 /* decrement loop counter */
+ b L(dWordUnrollOFF)
+
+ .p2align 4
+L(update2):
+ mr r5, r19
+ b L(dWordUnrollOFF)
+
+ .p2align 4
+L(HopBy40):
+ addi r9, r7, 40 /* increment dst by 40 */
+ addi r4, r6, 40 /* increment src by 40 */
+ addi r5, r5, -40 /* decrement length 'n' by 40 */
+ addi r0, r11, -5 /* decrement loop counter */
+ b L(dWordUnrollOFF)
+
+L(update3):
+ mr r0, r11
+ b L(dWordUnrollOFF)
+
+L(HopBy8):
+ addi r9, r3, 8 /* increment dst by 8 */
+ addi r4, r4, 8 /* increment src by 8 */
+ addi r5, r5, -8 /* decrement length 'n' by 8 */
+ addi r0, r11, -1 /* decrement loop counter */
+ b L(dWordUnrollOFF)
+END(FUNC_NAME)
+#ifndef USE_AS_STPNCPY
+libc_hidden_builtin_def (strncpy)
+#endif
--- /dev/null
+/* Optimized strpbrk implementation for PowerPC64/POWER7.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+/* char [r3] *strpbrk(const char [r4] *s, const char [r5] *accept) */
+
+ .machine power7
+EALIGN (strpbrk, 4, 0)
+ CALL_MCOUNT 3
+
+ lbz r10,0(r4)
+ cmpdi cr7,r10,0 /* accept[0] == '\0' ? */
+ beq cr7,L(nullfound)
+
+ /* The idea to speed up the algorithm is to create a lookup table
+ for fast check if input character should be considered. For ASCII
+ or ISO-8859-X character sets it has 256 positions. */
+
+ /* First the table should be cleared and to avoid unaligned accesses
+ when using the VSX stores the table address is aligned to 16
+ bytes. */
+ xxlxor v0,v0,v0
+
+ /* PPC64 ELF ABI stack is aligned to 16 bytes */
+ addi r9,r1,-256
+
+ li r5,16
+ li r6,32
+ li r8,48
+ addi r12,r9,64
+ /* Clear the table with 0 values */
+ stxvw4x v0,r0,r9
+ addi r11,r9,128
+ addi r7,r9,192
+ stxvw4x v0,r9,r5
+ li r0,1
+ stxvw4x v0,r9,r6
+ stxvw4x v0,r9,r8
+ stxvw4x v0,r0,r12
+ stxvw4x v0,r12,r5
+ stxvw4x v0,r12,r6
+ stxvw4x v0,r12,r8
+ stxvw4x v0,r0,r11
+ stxvw4x v0,r11,r5
+ stxvw4x v0,r11,r6
+ stxvw4x v0,r11,r8
+ stxvw4x v0,r0,r7
+ stxvw4x v0,r7,r5
+ stxvw4x v0,r7,r6
+ stxvw4x v0,r7,r8
+
+ /* Initialize the table as:
+ for (i=0; accept[i]; i++
+ table[accept[i]]] = 1 */
+ .p2align 4,,15
+L(init_table):
+ stbx r0,r9,r10
+ lbzu r10,1(r4)
+ cmpdi r0,r10,0
+ bne cr0,L(init_table)
+L(finish_table):
+ /* set table[0] = 1 */
+ li r4,1
+ stb r4,0(r9)
+ b L(mainloop)
+
+ /* Unrool the loop 4 times and check using the table as:
+ i = 0;
+ while (1)
+ {
+ if (table[input[i++]] == 1)
+ return (s[i -1] ? s + i - 1: NULL);
+ if (table[input[i++]] == 1)
+ return (s[i -1] ? s + i - 1: NULL);
+ if (table[input[i++]] == 1)
+ return (s[i -1] ? s + i - 1: NULL);
+ if (table[input[i++]] == 1)
+ return (s[i -1] ? s + i - 1: NULL);
+ } */
+ .p2align 4
+L(unroll):
+ lbz r0,1(r3)
+ lbzx r8,r9,r0
+ cmpwi cr6,r8,1
+ beq cr6,L(checkend2)
+ lbz r10,2(r3)
+ lbzx r4,r9,r10
+ cmpwi cr7,r4,1
+ beq cr7,L(checkend3)
+ lbz r12,3(r3)
+ addi r3,r3,4
+ lbzx r11,r9,r12
+ cmpwi cr0,r11,1
+ beq cr0,L(checkend)
+L(mainloop):
+ lbz r12,0(r3)
+ addi r11,r3,1
+ addi r5,r3,2
+ addi r7,r3,3
+ lbzx r6,r9,r12
+ cmpwi cr1,r6,1
+ bne cr1,L(unroll)
+ cmpdi cr0,r12,0
+ beq cr0,L(nullfound)
+L(end):
+ blr
+
+ .p2align 4
+L(checkend):
+ cmpdi cr1,r12,0
+ mr r3,r7
+ bne cr1,L(end)
+L(nullfound):
+ /* return NULL */
+ li 3,0
+ blr
+
+ .p2align 4
+L(checkend2):
+ cmpdi cr7,r0,0
+ mr r3,r11
+ beq cr7,L(nullfound)
+ blr
+
+ .p2align 4
+L(checkend3):
+ cmpdi cr6,r10,0
+ mr r3,r5
+ beq cr6,L(nullfound)
+ blr
+END (strpbrk)
+libc_hidden_builtin_def (strpbrk)
--- /dev/null
+/* Optimized strrchr implementation for PowerPC64/POWER7 using cmpb insn.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+
+/* int [r3] strrchr (char *s [r3], int c [r4]) */
+ .machine power7
+ENTRY (strrchr)
+ CALL_MCOUNT 2
+ dcbt 0,r3
+ clrrdi r8,r3,3 /* Align the address to doubleword boundary. */
+ cmpdi cr7,r4,0
+ ld r12,0(r8) /* Load doubleword from memory. */
+ li r9,0 /* used to store last occurence */
+ li r0,0 /* Doubleword with null chars to use
+ with cmpb. */
+
+ rlwinm r6,r3,3,26,28 /* Calculate padding. */
+
+ beq cr7,L(null_match)
+
+ /* Replicate byte to doubleword. */
+ insrdi r4,r4,8,48
+ insrdi r4,r4,16,32
+ insrdi r4,r4,32,0
+
+ /* r4 is changed now ,if its passed as more chars
+ check for null again */
+ cmpdi cr7,r4,0
+ beq cr7,L(null_match)
+ /* Now r4 has a doubleword of c bytes and r0 has
+ a doubleword of null bytes. */
+
+ cmpb r10,r12,r4 /* Compare each byte against c byte. */
+ cmpb r11,r12,r0 /* Compare each byte against null byte. */
+
+ /* Move the doublewords left and right to discard the bits that are
+ not part of the string and bring them back as zeros. */
+#ifdef __LITTLE_ENDIAN__
+ srd r10,r10,r6
+ srd r11,r11,r6
+ sld r10,r10,r6
+ sld r11,r11,r6
+#else
+ sld r10,r10,r6
+ sld r11,r11,r6
+ srd r10,r10,r6
+ srd r11,r11,r6
+#endif
+ or r5,r10,r11 /* OR the results to speed things up. */
+ cmpdi cr7,r5,0 /* If r5 == 0, no c or null bytes
+ have been found. */
+ bne cr7,L(done)
+
+L(align):
+ mtcrf 0x01,r8
+
+ /* Are we now aligned to a doubleword boundary? If so, skip to
+ the main loop. Otherwise, go through the alignment code. */
+
+ bt 28,L(loop)
+
+ /* Handle WORD2 of pair. */
+ ldu r12,8(r8)
+ cmpb r10,r12,r4
+ cmpb r11,r12,r0
+ or r5,r10,r11
+ cmpdi cr7,r5,0
+ bne cr7,L(done)
+ b L(loop) /* We branch here (rather than falling through)
+ to skip the nops due to heavy alignment
+ of the loop below. */
+ .p2align 5
+L(loop):
+ /* Load two doublewords, compare and merge in a
+ single register for speed. This is an attempt
+ to speed up the null-checking process for bigger strings. */
+ ld r12,8(r8)
+ ldu r7,16(r8)
+ cmpb r10,r12,r4
+ cmpb r11,r12,r0
+ cmpb r6,r7,r4
+ cmpb r7,r7,r0
+ or r12,r10,r11
+ or r5,r6,r7
+ or r5,r12,r5
+ cmpdi cr7,r5,0
+ beq cr7,L(loop)
+
+ /* OK, one (or both) of the doublewords contains a c/null byte. Check
+ the first doubleword and decrement the address in case the first
+ doubleword really contains a c/null byte. */
+ cmpdi cr6,r12,0
+ addi r8,r8,-8
+ bne cr6,L(done)
+
+ /* The c/null byte must be in the second doubleword. Adjust the
+ address again and move the result of cmpb to r10 so we can calculate
+ the pointer. */
+
+ mr r10,r6
+ mr r11,r7
+ addi r8,r8,8
+
+ /* r10/r11 have the output of the cmpb instructions, that is,
+ 0xff in the same position as the c/null byte in the original
+ doubleword from the string. Use that to calculate the pointer. */
+
+L(done):
+ /* if there are more than one 0xff in r11, find the first pos of ff
+ in r11 and fill r10 with 0 from that position */
+ cmpdi cr7,r11,0
+ beq cr7,L(no_null)
+#ifdef __LITTLE_ENDIAN__
+ addi r3,r11,-1
+ andc r3,r3,r11
+ popcntd r0,r3
+#else
+ cntlzd r0,r11
+#endif
+ subfic r0,r0,63
+ li r6,-1
+#ifdef __LITTLE_ENDIAN__
+ srd r0,r6,r0
+#else
+ sld r0,r6,r0
+#endif
+ and r10,r0,r10
+L(no_null):
+#ifdef __LITTLE_ENDIAN__
+ cntlzd r0,r10 /* Count leading zeros before c matches. */
+ addi r3,r10,-1
+ andc r3,r3,r10
+ addi r10,r11,-1
+ andc r10,r10,r11
+ cmpld cr7,r3,r10
+ bgt cr7,L(no_match)
+#else
+ addi r3,r10,-1 /* Count trailing zeros before c matches. */
+ andc r3,r3,r10
+ popcntd r0,r3
+ cmpld cr7,r11,r10
+ bgt cr7,L(no_match)
+#endif
+ srdi r0,r0,3 /* Convert trailing zeros to bytes. */
+ subfic r0,r0,7
+ add r9,r8,r0 /* Return address of the matching c byte
+ or null in case c was not found. */
+ li r0,0
+ cmpdi cr7,r11,0 /* If r11 == 0, no null's have been found. */
+ beq cr7,L(align)
+
+ .align 4
+L(no_match):
+ mr r3,r9
+ blr
+
+/* We are here because strrchr was called with a null byte. */
+ .align 4
+L(null_match):
+ /* r0 has a doubleword of null bytes. */
+
+ cmpb r5,r12,r0 /* Compare each byte against null bytes. */
+
+ /* Move the doublewords left and right to discard the bits that are
+ not part of the string and bring them back as zeros. */
+#ifdef __LITTLE_ENDIAN__
+ srd r5,r5,r6
+ sld r5,r5,r6
+#else
+ sld r5,r5,r6
+ srd r5,r5,r6
+#endif
+ cmpdi cr7,r5,0 /* If r10 == 0, no c or null bytes
+ have been found. */
+ bne cr7,L(done_null)
+
+ mtcrf 0x01,r8
+
+ /* Are we now aligned to a quadword boundary? If so, skip to
+ the main loop. Otherwise, go through the alignment code. */
+
+ bt 28,L(loop_null)
+
+ /* Handle WORD2 of pair. */
+ ldu r12,8(r8)
+ cmpb r5,r12,r0
+ cmpdi cr7,r5,0
+ bne cr7,L(done_null)
+ b L(loop_null) /* We branch here (rather than falling through)
+ to skip the nops due to heavy alignment
+ of the loop below. */
+
+ /* Main loop to look for the end of the string. Since it's a
+ small loop (< 8 instructions), align it to 32-bytes. */
+ .p2align 5
+L(loop_null):
+ /* Load two doublewords, compare and merge in a
+ single register for speed. This is an attempt
+ to speed up the null-checking process for bigger strings. */
+ ld r12,8(r8)
+ ldu r11,16(r8)
+ cmpb r5,r12,r0
+ cmpb r10,r11,r0
+ or r6,r5,r10
+ cmpdi cr7,r6,0
+ beq cr7,L(loop_null)
+
+ /* OK, one (or both) of the doublewords contains a null byte. Check
+ the first doubleword and decrement the address in case the first
+ doubleword really contains a null byte. */
+
+ cmpdi cr6,r5,0
+ addi r8,r8,-8
+ bne cr6,L(done_null)
+
+ /* The null byte must be in the second doubleword. Adjust the address
+ again and move the result of cmpb to r10 so we can calculate the
+ pointer. */
+
+ mr r5,r10
+ addi r8,r8,8
+
+ /* r5 has the output of the cmpb instruction, that is, it contains
+ 0xff in the same position as the null byte in the original
+ doubleword from the string. Use that to calculate the pointer. */
+L(done_null):
+#ifdef __LITTLE_ENDIAN__
+ addi r0,r5,-1
+ andc r0,r0,r5
+ popcntd r0,r0
+#else
+ cntlzd r0,r5 /* Count leading zeros before the match. */
+#endif
+ srdi r0,r0,3 /* Convert trailing zeros to bytes. */
+ add r3,r8,r0 /* Return address of the matching null byte. */
+ blr
+END (strrchr)
+weak_alias (strrchr, rindex)
+libc_hidden_builtin_def (strrchr)
--- /dev/null
+/* Optimized strspn implementation for PowerPC64/POWER7.
+
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* size_t [r3] strspn (const char *string [r3],
+ const char *needleAccept [r4] */
+
+/* Performance gains are grabbed through following techniques:
+
+ > hashing of needle.
+ > hashing avoids scanning of duplicate entries in needle
+ across the string.
+ > initializing the hash table with Vector instructions
+ by quadword access.
+ > unrolling when scanning for character in string
+ across hash table. */
+
+/* Algorithm is as below:
+ 1. A empty hash table/dictionary is created comprising of
+ 256 ascii character set
+ 2. When hash entry is found in needle , the hash index
+ is initialized to 1
+ 3. The string is scanned until end and for every character,
+ its corresponding hash index is compared.
+ 4. initial length of string (count) until first hit of
+ accept needle to be found is set to 0
+ 4. If hash index is set to 1 for the index of string,
+ count is returned.
+ 5. Otherwise count is incremented and scanning continues
+ until end of string. */
+
+#include <sysdep.h>
+
+#undef strspn
+
+ .machine power7
+EALIGN(strspn, 4, 0)
+ CALL_MCOUNT 2
+
+ lbz r10, 0(r4) /* load r10 with needle (r4) */
+ addi r9, r1, -256 /* r9 is a hash of 256 bytes */
+
+ li r5, 16 /* set r5 = 16 as offset */
+ li r6, 32 /* set r6 = 32 as offset */
+ li r8, 48 /* set r8 = 48 as offset */
+
+/*Iniatliaze hash table with Zeroes in double indexed quadword accesses */
+ xxlxor v0, v0, v0 /* prepare for initializing hash */
+
+ stxvd2x v0, r0, r9 /* initialize 1st quadword */
+ stxvd2x v0, r9, r5
+ stxvd2x v0, r9, r6
+ stxvd2x v0, r9, r8 /* initialize 4th quadword */
+
+ addi r11, r9, 64 /* r11 is index to hash */
+
+ stxvd2x v0, r0, r11 /* initialize 5th quadword */
+ stxvd2x v0, r11, r5
+ stxvd2x v0, r11, r6
+ stxvd2x v0, r11, r8 /* initialize 8th quadword */
+
+ addi r11, r9, 128 /* r11 is index to hash */
+
+ stxvd2x v0, r0, r11 /* initialize 9th quadword */
+ stxvd2x v0, r11, r5
+ stxvd2x v0, r11, r6
+ stxvd2x v0, r11, r8 /* initialize 12th quadword */
+
+ addi r11, r9, 192 /* r11 is index to hash */
+
+ stxvd2x v0, r0, r11 /* initialize 13th quadword */
+ stxvd2x v0, r11, r5
+ stxvd2x v0, r11, r6
+ stxvd2x v0, r11, r8 /* initialize 16th quadword */
+
+ li r8, 1 /* r8=1, marker into hash if found in
+ needle */
+
+ cmpdi cr7, r10, 0 /* accept needle is NULL */
+ beq cr7, L(skipHashing) /* if needle is NULL, skip hashing */
+
+ .p2align 4 /* align section to 16 byte boundary */
+L(hashing):
+ stbx r8, r9, r10 /* update hash with marker for the pivot of
+ the needle */
+ lbzu r10, 1(r4) /* load needle into r10 and update to next */
+ cmpdi cr7, r10, 0 /* if needle is has reached NULL, continue */
+ bne cr7, L(hashing) /* loop to hash the needle */
+
+L(skipHashing):
+ li r10, 0 /* load counter = 0 */
+ b L(beginScan)
+
+ .p2align 4 /* align section to 16 byte boundary */
+L(scanUnroll):
+ lbzx r8, r9, r8 /* load r8 with hash value at index */
+ cmpwi cr7, r8, 0 /* if we hit marker in hash, we have found
+ accept needle */
+ beq cr7, L(ret1stIndex) /* we have hit accept needle, return the
+ count */
+
+ lbz r8, 1(r3) /* load string[1] into r8 */
+ addi r10, r10, 4 /* increment counter */
+ lbzx r8, r9, r8 /* load r8 with hash value at index */
+ cmpwi cr7, r8, 0 /* if we hit marker in hash, we have found
+ accept needle */
+ beq cr7, L(ret2ndIndex) /* we have hit accept needle, return the
+ count */
+
+ lbz r8, 2(r3) /* load string[2] into r8 */
+ lbzx r8, r9, r8 /* load r8 with hash value at index */
+ cmpwi cr7, r8, 0 /* if we hit marker in hash, we have found
+ accept needle */
+ beq cr7, L(ret3rdIndex) /* we have hit accept needle, return the
+ count */
+
+ lbz r8, 3(r3) /* load string[3] into r8 */
+ lbzx r8, r9, r8 /* load r8 with hash value at index */
+ addi r3, r3, 4 /* unroll factor , increment string by 4 */
+ cmpwi cr7, r8, 0 /* if we hit marker in hash, we have found
+ accept needle */
+ beq cr7,L(ret4thIndex) /* we have hit accept needle, return the
+ count */
+
+L(beginScan):
+ lbz r8, 0(r3) /* load string[0] into r8 */
+ addi r6, r10, 1 /* place holder for counter + 1 */
+ addi r5, r10, 2 /* place holder for counter + 2 */
+ addi r4, r10, 3 /* place holder for counter + 3 */
+ cmpdi cr7, r8, 0 /* if we hit marker in hash, we have found
+ accept needle */
+ bne cr7, L(scanUnroll) /* continue scanning */
+
+L(ret1stIndex):
+ mr r3, r10 /* update r3 for return */
+ blr /* return */
+
+L(ret2ndIndex):
+ mr r3, r6 /* update r3 for return */
+ blr /* return */
+
+L(ret3rdIndex):
+ mr r3, r5 /* update r3 for return */
+ blr /* return */
+
+L(ret4thIndex):
+ mr r3, r4 /* update r3 for return */
+ blr /* done */
+END(strspn)
+libc_hidden_builtin_def (strspn)
-powerpc/powerpc64/power7/fpu/multiarch
+powerpc/powerpc64/power7/fpu/
--- /dev/null
+/* isfinite(). PowerPC64/POWER8 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <endian.h>
+#include <math_ldbl_opt.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define MFVSRD_R3_V1 .byte 0x66,0x00,0x23,0x7c /* mfvsrd r3,vs1 */
+#else
+#define MFVSRD_R3_V1 .byte 0x7c,0x23,0x00,0x66 /* mfvsrd r3,vs1 */
+#endif
+
+/* int [r3] __finite ([fp1] x) */
+
+EALIGN (__finite, 4, 0)
+ CALL_MCOUNT 0
+ MFVSRD_R3_V1
+ lis r9,0x8010
+ clrldi r3,r3,1 /* r3 = r3 & 0x8000000000000000 */
+ rldicr r9,r9,32,31 /* r9 = (r9 << 32) & 0xffffffff */
+ add r3,r3,r9
+ rldicl r3,r3,1,63
+ blr
+END (__finite)
+
+hidden_def (__finite)
+weak_alias (__finite, finite)
+
+/* It turns out that the 'double' version will also always work for
+ single-precision. */
+strong_alias (__finite, __finitef)
+hidden_def (__finitef)
+weak_alias (__finitef, finitef)
+
+#ifdef IS_IN_libm
+# if LONG_DOUBLE_COMPAT (libm, GLIBC_2_0)
+compat_symbol (libm, __finite, __finitel, GLIBC_2_0)
+compat_symbol (libm, finite, finitel, GLIBC_2_0)
+# endif
+#else
+# if LONG_DOUBLE_COMPAT (libc, GLIBC_2_0)
+compat_symbol (libc, __finite, __finitel, GLIBC_2_0);
+compat_symbol (libc, finite, finitel, GLIBC_2_0);
+# endif
+#endif
--- /dev/null
+/* This function uses the same code as s_finite.S. */
--- /dev/null
+/* isinf(). PowerPC64/POWER8 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <endian.h>
+#include <math_ldbl_opt.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define MFVSRD_R3_V1 .byte 0x66,0x00,0x23,0x7c /* mfvsrd r3,vs1 */
+#else
+#define MFVSRD_R3_V1 .byte 0x7c,0x23,0x00,0x66 /* mfvsrd r3,vs1 */
+#endif
+
+/* int [r3] __isinf([fp1] x) */
+
+EALIGN (__isinf, 4, 0)
+ CALL_MCOUNT 0
+ MFVSRD_R3_V1
+ lis r9,0x7ff0 /* r9 = 0x7ff0 */
+ rldicl r10,r3,0,1 /* r10 = r3 & (0x8000000000000000) */
+ sldi r9,r9,32 /* r9 = r9 << 52 */
+ cmpd cr7,r10,r9 /* fp1 & 0x7ff0000000000000 ? */
+ beq cr7,L(inf)
+ li r3,0 /* Not inf */
+ blr
+L(inf):
+ sradi r3,r3,63 /* r3 = r3 >> 63 */
+ ori r3,r3,1 /* r3 = r3 | 0x1 */
+ blr
+END (__isinf)
+
+hidden_def (__isinf)
+weak_alias (__isinf, isinf)
+
+/* It turns out that the 'double' version will also always work for
+ single-precision. */
+strong_alias (__isinf, __isinff)
+hidden_def (__isinff)
+weak_alias (__isinff, isinff)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__isinf, __isinfl)
+weak_alias (__isinf, isinfl)
+#endif
+
+#ifndef IS_IN_libm
+# if LONG_DOUBLE_COMPAT(libc, GLIBC_2_0)
+compat_symbol (libc, __isinf, __isinfl, GLIBC_2_0);
+compat_symbol (libc, isinf, isinfl, GLIBC_2_0);
+# endif
+#endif
--- /dev/null
+/* This function uses the same code as s_isinf.S. */
--- /dev/null
+/* isnan(). PowerPC64/POWER8 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <endian.h>
+#include <math_ldbl_opt.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define MFVSRD_R3_V1 .byte 0x66,0x00,0x23,0x7c /* mfvsrd r3,vs1 */
+#else
+#define MFVSRD_R3_V1 .byte 0x7c,0x23,0x00,0x66 /* mfvsrd r3,vs1 */
+#endif
+
+/* int [r3] __isnan([f1] x) */
+
+EALIGN (__isnan, 4, 0)
+ CALL_MCOUNT 0
+ MFVSRD_R3_V1
+ lis r9,0x7ff0
+ clrldi r3,r3,1 /* r3 = r3 & 0x8000000000000000 */
+ rldicr r9,r9,32,31 /* r9 = (r9 << 32) & 0xffffffff */
+ subf r3,r3,r9
+ rldicl r3,r3,1,63
+ blr
+END (__isnan)
+
+hidden_def (__isnan)
+weak_alias (__isnan, isnan)
+
+/* It turns out that the 'double' version will also always work for
+ single-precision. */
+strong_alias (__isnan, __isnanf)
+hidden_def (__isnanf)
+weak_alias (__isnanf, isnanf)
+
+#ifdef NO_LONG_DOUBLE
+strong_alias (__isnan, __isnanl)
+weak_alias (__isnan, isnanl)
+#endif
+
+#ifndef IS_IN_libm
+# if LONG_DOUBLE_COMPAT(libc, GLIBC_2_0)
+compat_symbol (libc, __isnan, __isnanl, GLIBC_2_0);
+compat_symbol (libc, isnan, isnanl, GLIBC_2_0);
+# endif
+#endif
--- /dev/null
+/* This function uses the same code as s_isnan.S. */
--- /dev/null
+/* Round double to long int. POWER8 PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <endian.h>
+#include <math_ldbl_opt.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define MFVSRD_R3_V1 .byte 0x66,0x00,0x23,0x7c /* mfvsrd r3,vs1 */
+#else
+#define MFVSRD_R3_V1 .byte 0x7c,0x23,0x00,0x66 /* mfvsrd r3,vs1 */
+#endif
+
+/* long long int[r3] __llrint (double x[fp1]) */
+ENTRY (__llrint)
+ CALL_MCOUNT 0
+ fctid fp1,fp1
+ MFVSRD_R3_V1
+ 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
--- /dev/null
+/* llround function. POWER8 PowerPC64 version.
+ Copyright (C) 2014 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <sysdep.h>
+#include <endian.h>
+#include <math_ldbl_opt.h>
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define MFVSRD_R3_V1 .byte 0x66,0x00,0x23,0x7c /* mfvsrd r3,vs1 */
+#else
+#define MFVSRD_R3_V1 .byte 0x7c,0x23,0x00,0x66 /* mfvsrd r3,vs1 */
+#endif
+
+/* long long [r3] llround (float x [fp1]) */
+
+ENTRY (__llround)
+ CALL_MCOUNT 0
+ frin fp1,fp1 /* Round to nearest +-0.5. */
+ fctidz fp1,fp1 /* Convert To Integer DW round toward 0. */
+ MFVSRD_R3_V1
+ 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
/* put the address of start_addresses in r8... **
** PPC64 ABI uses R13 for thread local, so we leave it alone */
- ld r8,.L01(r2)
+ ld r8,.L01@toc(r2)
/* and continue in libc-start, in glibc. */
b JUMPTARGET(__libc_start_main)
#include <string.h>
-#undef strcat
+#ifndef STRCAT
+# undef strcat
+# define STRCAT strcat
+#endif
/* Append SRC on the end of DEST. */
char *
-strcat (char *dest, const char *src)
+STRCAT(char *dest, const char *src)
{
strcpy (dest + strlen (dest), src);
return dest;
}
-libc_hidden_builtin_def (strcat)
+libc_hidden_builtin_def (STRCAT)
projects / thirdparty / glibc.git / commitdiff
