Update.cvs/libc-ud-970924

1997-09-25 00:23  Ulrich Drepper  <drepper@cygnus.com>

	* elf/dl-profile.c: Correct implementation.

	* io/ftwtest-sh: Don't depend on TMPDIR == /tmp.

	* locale/setlocale.c: Rewrite a bit for more clarity.

	* math/Makefile (libm-calls): Add w_exp2.
	* math/math_private.h: Add prototypes for __ieee754_exp2{,f,l}.
	* sysdeps/libm-i387/s_exp2.S: Change name to __ieee754_exp2.
	* sysdeps/libm-i387/s_exp2f.S: Likewise.
	* sysdeps/libm-i387/s_exp2l.S: Likewise.
	* sysdeps/libm-ieee754/k_standard.c: Add error cases for exp2.

	* string/bits/string2.h (__strcpy_small): Optimize.
	(__stpcpy_small): Likewise.
	(strncpy): Use variable for dest argument since it's used more than
	once.
	(strncat): Likewise.
	(strcmp): Add optimization for this function.
	* sysdeps/i386/i486/string.h (strlen): Correctly use __builtin_strlen.
	(__strcpy_small): Optimize.
	(__stpcpy_small): Likewise.
	(__stpcpy_c): Correctly use __mempcpy_* macros.
	(__mempcpy_by2, __mempcpy_by4, __mempcpy_byn): Return pointer to
	byte following last copied.
	(strncat): Use variable for dest argument since it's used more than
	once.
	(strcmp): Add optimization for this function.

	* sysdeps/i386/dl-machine.h (ELF_MACHINE_RUNTIME_TRAMPOLINE): Use
	.text and .previous to select correct section.

1997-09-23 19:56  Philip Blundell  <Philip.Blundell@pobox.com>

	* configure.in: Add `--disable-versioning' option to suppress the
	use of symbol versions even if binutils claims to support it.

1997-09-24 20:10  Philip Blundell  <Philip.Blundell@pobox.com>

	* csu/Makefile (before-compile): Don't try to build abi-tags.h if
	not using ELF.

1997-09-21  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* elf/sofini.c [HAVE_DWARF2_UNWIND_INFO]: Supply termination for
	the frame unwind info section.

	* elf/soinit.c [HAVE_DWARF2_UNWIND_INFO]: Register and unregister
	the frame unwind info.

	* config.h.in (HAVE_DWARF2_UNWIND_INFO): Add #undef.

	* configure.in: Check whether gcc supports DWARF2 unwind info.

	* libc.map: Export frame handling tables.

1997-09-10 06:56  Geoff Keating  <geoffk@ozemail.com.au>

	* sysdeps/unix/sysv/linux/powerpc/bits/kernel_termios.h:
	Use the size of the kernel's termios structure for ioctls.

	* sysdeps/powerpc/dl-machine.h: Prepare for library profiling.

	* sysdeps/powerpc/bits/mathinline.h: Add slightly slower versions
	of the C9X FP comparison macros. Delete 'fabs' and 'sqrt' inline
	routines, because gcc has them as internals.

	* sysdeps/powerpc/Makefile (pic-ccflags): Define this instead of
	CFLAGS-.os.

	* sysdeps/powerpc/bzero.S: New file.
	* sysdeps/powerpc/strcat.c: New file.
	* sysdeps/powerpc/strcpy.S: New file.
	* sysdeps/powerpc/stpcpy.S: New file.

	* math/Makefile: Add atest_exp2, test-reduce.
	* math/atest_exp2.c: New file.
	* math/test-reduce.c: New file.
	* sysdeps/libm-ieee754/Dist: New file.
	* sysdeps/libm-ieee754/s_exp2.c: New file.
	* sysdeps/libm-ieee754/s_exp2f.c: New file.
	* sysdeps/libm-ieee754/t_exp2.h: New file.
	* sysdeps/libm-ieee754/t_exp2f.h: New file.
	* math/libm-test.c (exp2_test): Add some more tests.

	* Rules: Use empty.os instead of empty.o, since it gets linked into
	libc.so...
	* configure.in: Add --disable-static to disable building .a files.
	* config.make.in: Substitute the new variable.
	* Makeconfig: Don't build .o files if not building .a files.

	* elf/dl-runtime.c (fixup): Factor out call to elf_machine_relplt.
	(profile_fixup): Likewise.

1997-09-23  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* posix/globtest.c (main): Change to directory passed as first
	argument.

	* posix/globtest.sh: Don't cd before running the program, instead
	pass testdir as argument, so that $common_objpfx remains valid.

1997-09-23 18:01  Thorsten Kukuk  <kukuk@vt.uni-paderborn.de>

	* nis/nis_call.c (__do_niscall2): Fix multiple Server support.

	* nis/nis_findserver.c: Save latest working sockaddr_in.

1997-09-17 22:07  Zack Weinberg  <zack@rabi.phys.columbia.edu>

	* configure.in: Automatically determine whether as and ld are the
	GNU versions.
	(options):  --with-gnu-{as,ld,binutils} replaced by
	single option --with-binutils=PATH specifying a -B option to gcc.
	* aclocal.m4: Two new macros defined, LIBC_PROG_FOO_GNU and
	LIBC_PROG_BINUTILS.

	* configure.in: Allow the user to force configuration for
	unsupported platforms with an undocumented option.

1997-09-22 16:55  Thorsten Kukuk  <kukuk@vt.uni-paderborn.de>

	* sunrpc/get_myaddr.c (get_myaddress): Avoid loopback interfaces,
	return loopback address only if there is no other interface.

	* nis/Makefile: Add nis_callback libnsl-routines.

	* nis/nis_add.c: Use new __do_niscall* interface.
	* nis/nis_cache.c: Likewise.
	* nis/nis_checkpoint.c: Likewise.
	* nis/nis_lookup.c: Likewise.
	* nis/nis_mkdir.c: Likewise.
	* nis/nis_modify.c: Likewise.
	* nis/nis_ping.c: Likewise.
	* nis/nis_remove.c: Likewise.
	* nis/nis_rmdir.c: Likewise.
	* nis/nis_server.c: Likewise.
	* nis/nis_util.c: Likewise.

	* nis/rpcsvc/nis.h: Make C++ safe.

	* nis/nss_nisplus/nisplus-publickey.c (getsecretkey): Fix use
	of variables.

	* nis/nis_findserv.c: Make thread safe.

	* nis/nis_call.c: Add support for callback, Fix use of variables.
	* nis/nis_table.c: Add support for callback, FOLLOW_PATH and
	ALL_RESULTS.
	* nis/nis_callback.c: New, callback functions.
	* nis/nis_intern.h: Add callback declarations.

1997-09-20  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* string/bits/string2.h: Fix logic in preprocessor directive.
	(__strsep_1c, __strsep_g): Don't declare __retval as pointing to
	const, to save a cast and a possible warning.

1997-09-22 04:12  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/i386/i486/atomicity.h (exchange_and_add): Really address
	memory in xadd not %1.

1997-09-21 13:56  Ulrich Drepper  <drepper@cygnus.com>

	* manual/maint.texi (Supported Configurations): Add SPARC64 to
	list of supported platforms.

1 files changed, 126 insertions, 0 deletions

diff --git a/sysdeps/libm-ieee754/s_exp2f.c b/sysdeps/libm-ieee754/s_exp2f.c
new file mode 100644
index 0000000000..e50ac77792
--- /dev/null
+++ b/sysdeps/libm-ieee754/s_exp2f.c
@@ -0,0 +1,126 @@
+/* Single-precision floating point 2^x.
+   Copyright (C) 1997 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Geoffrey Keating <geoffk@ozemail.com.au>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Library General Public License as
+   published by the Free Software Foundation; either version 2 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
+   Library General Public License for more details.
+
+   You should have received a copy of the GNU Library General Public
+   License along with the GNU C Library; see the file COPYING.LIB.  If not,
+   write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+   Boston, MA 02111-1307, USA.  */
+
+/* The basic design here is from
+   Shmuel Gal and Boris Bachelis, "An Accurate Elementary Mathematical
+   Library for the IEEE Floating Point Standard", ACM Trans. Math. Soft.,
+   17 (1), March 1991, pp. 26-45.
+   It has been slightly modified to compute 2^x instead of e^x, and for
+   single-precision.
+   */
+#define _GNU_SOURCE
+#include <float.h>
+#include <ieee754.h>
+#include <math.h>
+#include <fenv.h>
+#include <inttypes.h>
+#include <math_private.h>
+
+#include "t_exp2f.h"
+
+static const volatile float TWOM100 = 7.88860905e-31;
+static const volatile float huge = 1e+30;
+
+float
+__ieee754_exp2f (float x)
+{
+  static const uint32_t a_inf = 0x7f800000;
+  /* Check for usual case.  */
+  if (isless (x, (float) FLT_MAX_EXP)
+      && isgreater (x, (float) (FLT_MIN_EXP - 1)))
+    {
+      static const float TWO16 = 65536.0;
+      int tval;
+      float rx, x22;
+      union ieee754_float ex2_u;
+      fenv_t oldenv;
+
+      feholdexcept (&oldenv);
+      fesetround (FE_TONEAREST);
+
+      /* 1. Argument reduction.
+	 Choose integers ex, -128 <= t < 128, and some real
+	 -1/512 <= x1 <= 1/512 so that
+	 x = ex + t/512 + x1.
+
+	 First, calculate rx = ex + t/256.  */
+      if (x >= 0)
+	{
+	  rx = x + TWO16;
+	  rx -= TWO16;
+	}
+      else
+	{
+	  rx = x - TWO16;
+	  rx += TWO16;
+	}
+      x -= rx;  /* Compute x=x1. */
+      /* Compute tval = (ex*256 + t)+128.
+	 Now, t = (tval mod 256)-128 and ex=tval/256  [that's mod, NOT %; and
+	 /-round-to-nearest not the usual c integer /].  */
+      tval = (int) (rx * 256.0f + 128.0f);
+
+      /* 2. Adjust for accurate table entry.
+	 Find e so that
+	 x = ex + t/256 + e + x2
+	 where -7e-4 < e < 7e-4, and
+	 (float)(2^(t/256+e))
+	 is accurate to one part in 2^-64.  */
+
+      /* 'tval & 255' is the same as 'tval%256' except that it's always
+	 positive.
+	 Compute x = x2.  */
+      x -= exp2_deltatable[tval & 255];
+
+      /* 3. Compute ex2 = 2^(t/255+e+ex).  */
+      ex2_u.f = exp2_accuratetable[tval & 255];
+      ex2_u.ieee.exponent += tval >> 8;
+
+      /* 4. Approximate 2^x2 - 1, using a second-degree polynomial,
+	 2^x2 ~= sum(k=0..2 | (x2 * ln(2))^k / k! ) +
+	 so
+	 2^x2 - 1 ~= sum(k=1..4 | (x2 * ln(2))^k / k! )
+	 with error less than 2^(1/512+7e-4) * (x2 * ln(2))^3 / 3! < 1.2e-18.  */
+
+      x22 = (.240226507f * x + .6931471806f) * ex2_u.f;
+
+      /* 5. Return (2^x2-1) * 2^(t/512+e+ex) + 2^(t/512+e+ex).  */
+      fesetenv (&oldenv);
+
+      /* Need to check: does this set FE_INEXACT correctly? */
+      return x22 * x + ex2_u.f;
+    }
+  /* 2^inf == inf, with no error.  */
+  else if (x == *(const float *)&a_inf)
+    {
+      return x;
+    }
+  /* Check for overflow.  */
+  else if (isgreaterequal (x, (float) FLT_MAX_EXP))
+    return huge * huge;
+  /* And underflow (including -inf).  */
+  else if (isless (x, (float) (FLT_MIN_EXP - FLT_MANT_DIF)))
+    return TWOM100 * TWOM100;
+  /* Maybe the result needs to be a denormalised number...  */
+  else if (!isnan (x))
+    return __ieee754_exp2f (x + 100.0) * TWOM100;
+  else /* isnan(x) */
+    return x + x;
+}