Update. cvs/fedora-glibc-20050106T1443

2004-12-29  Jakub Jelinek  <jakub@redhat.com>

	* sysdeps/ia64/fpu/libm_support.h (__libm_error_support): Use
	libc_hidden_proto instead of HIDDEN_PROTO.
	* sysdeps/ia64/fpu/libm-symbols.h (HIDDEN_PROTO): Remove.
	(__libm_error_support): If ASSEMBLER and in libc, define to
	HIDDEN_JUMPTARGET(__libm_error_support).

2004-12-28  David Mosberger  <davidm@hpl.hp.com>

	* sysdeps/ia64/fpu/Makefile (duplicated-routines): New macro.
	(sysdep_routines): Replace libm_ldexp{,f,l} and libm_scalbn{,f,l}
	with $(duplicated-routines).
	(libm-sysdep_routines): Likewise, but substitute "s_" prefix for
	"m_" prefix.

2004-12-27  David Mosberger  <davidm@hpl.hp.com>

	* sysdeps/ia64/fpu/libm-symbols.h: Add include of <sysdep.h> and
	undefine "ret" macro.  Add __libm_error_support hidden definitions.

	* sysdeps/ia64/fpu/e_lgamma_r.c: Remove CVS-id comment.  Add
	missing portion of copyright statement.
	* sysdeps/ia64/fpu/e_lgammaf_r.c: Likewise.
	* sysdeps/ia64/fpu/e_lgammal_r.c: Likewise.

	* sysdeps/ia64/fpu/w_lgamma.c: Remove CVS-id comment.  Add
	missing portion of copyright statement.
	(__ieee754_lgamma): Rename from lgamma().  Make lgamma() a weak alias.
	(__ieee754_gamma): Likewise.
	* sysdeps/ia64/fpu/w_lgammaf.c: Likewise.
	* sysdeps/ia64/fpu/w_lgammal.c: Likewise.

2004-12-09  H. J. Lu <hjl@lucon.org>

	* sysdeps/ia64/fpu/s_nextafterl.c: Remove.
	* sysdeps/ia64/fpu/s_nexttoward.c: Likewise.
	* sysdeps/ia64/fpu/s_nexttowardf.c: Likewise.
	* sysdeps/ia64/fpu/e_atan2l.S: Remove (duplicate of e_atan2l.c).
	* sysdeps/ia64/fpu/e_expl.S: Likewise.
	* sysdeps/ia64/fpu/e_logl.c: Remove (conflicts with e_logl.S).

2004-11-18  David Mosberger  <davidm@hpl.hp.com>

	* sysdeps/ia64/fpu/README: New file.
	* sysdeps/ia64/fpu/gen_import_file_list: New file.
	* sysdeps/ia64/fpu/import_check: Likewise.
	* sysdeps/ia64/fpu/import_diffs: Likewise.
	* sysdeps/ia64/fpu/import_file.awk: Likewise.
	* sysdeps/ia64/fpu/import_intel_libm: Likewise.
	* sysdeps/ia64/fpu/libm-symbols.h: Likewise.

	* sysdeps/ia64/fpu/e_acos.S: Update from Intel libm v2.1+.
	* sysdeps/ia64/fpu/e_acosf.S: Likewise.
	* sysdeps/ia64/fpu/e_acosl.S: Likewise.
	* sysdeps/ia64/fpu/e_asin.S: Likewise.
	* sysdeps/ia64/fpu/e_asinf.S: Likewise.
	* sysdeps/ia64/fpu/e_asinl.S: Likewise.
	* sysdeps/ia64/fpu/e_atan2.S: Likewise.
	* sysdeps/ia64/fpu/e_atan2f.S: Likewise.
	* sysdeps/ia64/fpu/e_cosh.S: Likewise.
	* sysdeps/ia64/fpu/e_coshf.S: Likewise.
	* sysdeps/ia64/fpu/e_coshl.S: Likewise.
	* sysdeps/ia64/fpu/e_exp.S: Likewise.
	* sysdeps/ia64/fpu/e_expf.S: Likewise.
	* sysdeps/ia64/fpu/e_fmod.S: Likewise.
	* sysdeps/ia64/fpu/e_fmodf.S: Likewise.
	* sysdeps/ia64/fpu/e_fmodl.S: Likewise.
	* sysdeps/ia64/fpu/e_hypot.S: Likewise.
	* sysdeps/ia64/fpu/e_hypotf.S: Likewise.
	* sysdeps/ia64/fpu/e_hypotl.S: Likewise.
	* sysdeps/ia64/fpu/e_log.S: Likewise.
	* sysdeps/ia64/fpu/e_log2.S: Likewise.
	* sysdeps/ia64/fpu/e_log2f.S: Likewise.
	* sysdeps/ia64/fpu/e_log2l.S: Likewise.
	* sysdeps/ia64/fpu/e_logf.S: Likewise.
	* sysdeps/ia64/fpu/e_pow.S: Likewise.
	* sysdeps/ia64/fpu/e_powf.S: Likewise.
	* sysdeps/ia64/fpu/e_powl.S: Likewise.
	* sysdeps/ia64/fpu/e_remainder.S: Likewise.
	* sysdeps/ia64/fpu/e_remainderf.S: Likewise.
	* sysdeps/ia64/fpu/e_remainderl.S: Likewise.
	* sysdeps/ia64/fpu/e_scalb.S: Likewise.
	* sysdeps/ia64/fpu/e_scalbf.S: Likewise.
	* sysdeps/ia64/fpu/e_scalbl.S: Likewise.
	* sysdeps/ia64/fpu/e_sinh.S: Likewise.
	* sysdeps/ia64/fpu/e_sinhf.S: Likewise.
	* sysdeps/ia64/fpu/e_sinhl.S: Likewise.
	* sysdeps/ia64/fpu/e_sqrt.S: Likewise.
	* sysdeps/ia64/fpu/e_sqrtf.S: Likewise.
	* sysdeps/ia64/fpu/e_sqrtl.S: Likewise.
	* sysdeps/ia64/fpu/libm_error.c: Likewise.
	* sysdeps/ia64/fpu/libm_reduce.c: Likewise.
	* sysdeps/ia64/fpu/libm_support.h: Likewise.
	* sysdeps/ia64/fpu/s_atan.S: Likewise.
	* sysdeps/ia64/fpu/s_atanf.S: Likewise.
	* sysdeps/ia64/fpu/s_atanl.S: Likewise.
	* sysdeps/ia64/fpu/s_cbrt.S: Likewise.
	* sysdeps/ia64/fpu/s_cbrtf.S: Likewise.
	* sysdeps/ia64/fpu/s_cbrtl.S: Likewise.
	* sysdeps/ia64/fpu/s_ceil.S: Likewise.
	* sysdeps/ia64/fpu/s_ceilf.S: Likewise.
	* sysdeps/ia64/fpu/s_ceill.S: Likewise.
	* sysdeps/ia64/fpu/s_cos.S: Likewise.
	* sysdeps/ia64/fpu/s_cosf.S: Likewise.
	* sysdeps/ia64/fpu/s_cosl.S: Likewise.
	* sysdeps/ia64/fpu/s_expm1.S: Likewise.
	* sysdeps/ia64/fpu/s_expm1f.S: Likewise.
	* sysdeps/ia64/fpu/s_expm1l.S: Likewise.
	* sysdeps/ia64/fpu/s_fabs.S: Likewise.
	* sysdeps/ia64/fpu/s_fabsf.S: Likewise.
	* sysdeps/ia64/fpu/s_fabsl.S: Likewise.
	* sysdeps/ia64/fpu/s_floor.S: Likewise.
	* sysdeps/ia64/fpu/s_floorf.S: Likewise.
	* sysdeps/ia64/fpu/s_floorl.S: Likewise.
	* sysdeps/ia64/fpu/s_frexp.c: Likewise.
	* sysdeps/ia64/fpu/s_frexpf.c: Likewise.
	* sysdeps/ia64/fpu/s_frexpl.c: Likewise.
	* sysdeps/ia64/fpu/s_ilogb.S: Likewise.
	* sysdeps/ia64/fpu/s_ilogbf.S: Likewise.
	* sysdeps/ia64/fpu/s_ilogbl.S: Likewise.
	* sysdeps/ia64/fpu/s_log1p.S: Likewise.
	* sysdeps/ia64/fpu/s_log1pf.S: Likewise.
	* sysdeps/ia64/fpu/s_log1pl.S: Likewise.
	* sysdeps/ia64/fpu/s_logb.S: Likewise.
	* sysdeps/ia64/fpu/s_logbf.S: Likewise.
	* sysdeps/ia64/fpu/s_logbl.S: Likewise.
	* sysdeps/ia64/fpu/s_modf.S: Likewise.
	* sysdeps/ia64/fpu/s_modff.S: Likewise.
	* sysdeps/ia64/fpu/s_modfl.S: Likewise.
	* sysdeps/ia64/fpu/s_nearbyint.S: Likewise.
	* sysdeps/ia64/fpu/s_nearbyintf.S: Likewise.
	* sysdeps/ia64/fpu/s_nearbyintl.S: Likewise.
	* sysdeps/ia64/fpu/s_rint.S: Likewise.
	* sysdeps/ia64/fpu/s_rintf.S: Likewise.
	* sysdeps/ia64/fpu/s_rintl.S: Likewise.
	* sysdeps/ia64/fpu/s_round.S: Likewise.
	* sysdeps/ia64/fpu/s_roundf.S: Likewise.
	* sysdeps/ia64/fpu/s_roundl.S: Likewise.
	* sysdeps/ia64/fpu/s_significand.S: Likewise.
	* sysdeps/ia64/fpu/s_significandf.S: Likewise.
	* sysdeps/ia64/fpu/s_significandl.S: Likewise.
	* sysdeps/ia64/fpu/s_tan.S: Likewise.
	* sysdeps/ia64/fpu/s_tanf.S: Likewise.
	* sysdeps/ia64/fpu/s_tanl.S: Likewise.
	* sysdeps/ia64/fpu/s_trunc.S: Likewise.
	* sysdeps/ia64/fpu/s_truncf.S: Likewise.
	* sysdeps/ia64/fpu/s_truncl.S: Likewise.

	* sysdeps/ia64/fpu/e_acosh.S: New file from Intel libm v2.1+.
	* sysdeps/ia64/fpu/e_acoshf.S: Likewise.
	* sysdeps/ia64/fpu/e_acoshl.S: Likewise.
	* sysdeps/ia64/fpu/e_atanh.S: Likewise.
	* sysdeps/ia64/fpu/e_atanhf.S: Likewise.
	* sysdeps/ia64/fpu/e_atanhl.S: Likewise.
	* sysdeps/ia64/fpu/e_exp10.S: Likewise.
	* sysdeps/ia64/fpu/e_exp10f.S: Likewise.
	* sysdeps/ia64/fpu/e_exp10l.S: Likewise.
	* sysdeps/ia64/fpu/e_exp2.S: Likewise.
	* sysdeps/ia64/fpu/e_exp2f.S: Likewise.
	* sysdeps/ia64/fpu/e_exp2l.S: Likewise.
	* sysdeps/ia64/fpu/e_lgamma_r.S: Likewise.
	* sysdeps/ia64/fpu/e_lgammaf_r.S: Likewise.
	* sysdeps/ia64/fpu/e_lgammal_r.S: Likewise.
	* sysdeps/ia64/fpu/e_logl.S: Likewise.
	* sysdeps/ia64/fpu/libm_frexp.S: Likewise.
	* sysdeps/ia64/fpu/libm_frexpf.S: Likewise.
	* sysdeps/ia64/fpu/libm_frexpl.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_ldexp.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_ldexpf.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_ldexpl.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_scalbn.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_scalbnf.S: Likewise.
	* sysdeps/ia64/fpu/s_libm_scalbnl.S: Likewise.
	* sysdeps/ia64/fpu/libm_lgamma.S: Likewise.
	* sysdeps/ia64/fpu/libm_lgammaf.S: Likewise.
	* sysdeps/ia64/fpu/libm_lgammal.S: Likewise.
	* sysdeps/ia64/fpu/libm_sincos.S: Likewise.
	* sysdeps/ia64/fpu/libm_sincos_large.S: Likewise.
	* sysdeps/ia64/fpu/libm_sincosf.S: Likewise.
	* sysdeps/ia64/fpu/libm_sincosl.S: Likewise.
	* sysdeps/ia64/fpu/libm_scalblnf.S: Likewise.
	* sysdeps/ia64/fpu/s_asinh.S: Likewise.
	* sysdeps/ia64/fpu/s_asinhf.S: Likewise.
	* sysdeps/ia64/fpu/s_asinhl.S: Likewise.
	* sysdeps/ia64/fpu/s_erf.S: Likewise.
	* sysdeps/ia64/fpu/s_erfc.S: Likewise.
	* sysdeps/ia64/fpu/s_erfcf.S: Likewise.
	* sysdeps/ia64/fpu/s_erfcl.S: Likewise.
	* sysdeps/ia64/fpu/s_erff.S: Likewise.
	* sysdeps/ia64/fpu/s_erfl.S: Likewise.
	* sysdeps/ia64/fpu/s_fdim.S: Likewise.
	* sysdeps/ia64/fpu/s_fdimf.S: Likewise.
	* sysdeps/ia64/fpu/s_fdiml.S: Likewise.
	* sysdeps/ia64/fpu/s_fma.S: Likewise.
	* sysdeps/ia64/fpu/s_fmaf.S: Likewise.
	* sysdeps/ia64/fpu/s_fmal.S: Likewise.
	* sysdeps/ia64/fpu/s_fmax.S: Likewise.
	* sysdeps/ia64/fpu/s_fmaxf.S: Likewise.
	* sysdeps/ia64/fpu/s_fmaxl.S: Likewise.
	* sysdeps/ia64/fpu/s_ldexp.c: Likewise.
	* sysdeps/ia64/fpu/s_ldexpf.c: Likewise.
	* sysdeps/ia64/fpu/s_ldexpl.c: Likewise.
	* sysdeps/ia64/fpu/s_nextafter.S: Likewise.
	* sysdeps/ia64/fpu/s_nextafterf.S: Likewise.
	* sysdeps/ia64/fpu/s_nextafterl.S: Likewise.
	* sysdeps/ia64/fpu/s_nexttoward.S: Likewise.
	* sysdeps/ia64/fpu/s_nexttowardf.S: Likewise.
	* sysdeps/ia64/fpu/s_nexttowardl.S: Likewise.
	* sysdeps/ia64/fpu/s_tanh.S: Likewise.
	* sysdeps/ia64/fpu/s_tanhf.S: Likewise.
	* sysdeps/ia64/fpu/s_tanhl.S: Likewise.
	* sysdeps/ia64/fpu/s_scalblnf.c: Likewise.
	* sysdeps/ia64/fpu/w_lgamma.c: Likewise.
	* sysdeps/ia64/fpu/w_lgammaf.c: Likewise.
	* sysdeps/ia64/fpu/w_lgammal.c: Likewise.
	* sysdeps/ia64/fpu/w_tgamma.S: Likewise.
	* sysdeps/ia64/fpu/w_tgammaf.S: Likewise.
	* sysdeps/ia64/fpu/w_tgammal.S: Likewise.

	* sysdeps/ia64/fpu/e_gamma_r.c: New empty dummy-file.
	* sysdeps/ia64/fpu/e_gammaf_r.c: Likewise.
	* sysdeps/ia64/fpu/e_gammal_r.c: Likewise.
	* sysdeps/ia64/fpu/w_acosh.c: Likewise.
	* sysdeps/ia64/fpu/w_acoshf.c: Likewise.
	* sysdeps/ia64/fpu/w_acoshl.c: Likewise.
	* sysdeps/ia64/fpu/w_atanh.c: Likewise.
	* sysdeps/ia64/fpu/w_atanhf.c: Likewise.
	* sysdeps/ia64/fpu/w_atanhl.c: Likewise.
	* sysdeps/ia64/fpu/w_exp10.c: Likewise.
	* sysdeps/ia64/fpu/w_exp10f.c: Likewise.
	* sysdeps/ia64/fpu/w_exp10l.c: Likewise.
	* sysdeps/ia64/fpu/w_exp2.c: Likewise.
	* sysdeps/ia64/fpu/w_exp2f.c: Likewise.
	* sysdeps/ia64/fpu/w_exp2l.c: Likewise.
	* sysdeps/ia64/fpu/w_expl.c: Likewise.
	* sysdeps/ia64/fpu/e_expl.S: Likewise.
	* sysdeps/ia64/fpu/w_lgamma_r.c: Likewise.
	* sysdeps/ia64/fpu/w_lgammaf_r.c: Likewise.
	* sysdeps/ia64/fpu/w_lgammal_r.c: Likewise.
	* sysdeps/ia64/fpu/w_log2.c: Likewise.
	* sysdeps/ia64/fpu/w_log2f.c: Likewise.
	* sysdeps/ia64/fpu/w_log2l.c: Likewise.
	* sysdeps/ia64/fpu/w_sinh.c: Likewise.
	* sysdeps/ia64/fpu/w_sinhf.c: Likewise.
	* sysdeps/ia64/fpu/w_sinhl.c: Likewise.

	* sysdeps/ia64/fpu/libm_atan2_reg.S: Remove.
	* sysdeps/ia64/fpu/s_ldexp.S: Likewise.
	* sysdeps/ia64/fpu/s_ldexpf.S: Likewise.
	* sysdeps/ia64/fpu/s_ldexpl.S: Likewise.
	* sysdeps/ia64/fpu/s_scalbn.S: Likewise.
	* sysdeps/ia64/fpu/s_scalbnf.S: Likewise.
	* sysdeps/ia64/fpu/s_scalbnl.S: Likewise.

	* sysdeps/ia64/fpu/s_sincos.c: Make it an empty dummy-file.
	* sysdeps/ia64/fpu/s_sincosf.c: Likewise.
	* sysdeps/ia64/fpu/s_sincosl.c: Likewise.

	* sysdeps/ia64/fpu/e_atan2l.S: Add "Not needed" comment.

	* sysdeps/ia64/fpu/s_copysign.S: Add __libm_copysign{,f,l}
	alias for use by libm_error.c

	* sysdeps/ia64/fpu/Makefile (libm-sysdep_routines): Remove
	libm_atan2_reg, libm_tan, libm_frexp4{f,l}.
	Mention s_erfc{,f,l}, libm_frexp{,f,l}, libm_ldexp{,f,l},
	libm_sincos{,f,l}, libm_sincos_large, libm_lgamma{,f,l},
	libm_scalbn{,f,l}, libm_scalblnf.
	(sysdep_routines): Remove libm_frexp4{,f,l}.
	Mention libm_frexp{,f,l}, libm_ldexp{,f,l}, and libm_scalbn{,f,l}.
	(sysdep-CPPFLAGS): Add -include libm-symbols.h, -D__POSIX__,
	_D_LIB_VERSIONIMF=_LIB_VERSION, -DSIZE_LONG_INT_64, and
	-DSIZE_LONG_LONG_INT_64.

1 files changed, 144 insertions, 172 deletions

diff --git a/sysdeps/ia64/fpu/s_roundl.S b/sysdeps/ia64/fpu/s_roundl.S
index b30f590917..79dff00c06 100644
--- a/sysdeps/ia64/fpu/s_roundl.S
+++ b/sysdeps/ia64/fpu/s_roundl.S
@@ -1,11 +1,10 @@
 .file "roundl.s"
 
-// Copyright (C) 2000, 2001, Intel Corporation
+
+// Copyright (c) 2000 - 2003, Intel Corporation
 // All rights reserved.
-// 
-// Contributed 10/25/2000 by John Harrison, Cristina Iordache, Ted Kubaska,
-// Bob Norin, Tom Rowan, Shane Story, and Ping Tak Peter Tang of the
-// Computational Software Lab, Intel Corporation.
+//
+// Contributed 2000 by the Intel Numerics Group, Intel Corporation
 //
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
@@ -21,229 +20,202 @@
 // * The name of Intel Corporation may not be used to endorse or promote
 // products derived from this software without specific prior written
 // permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 
+
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS 
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
-// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 
-// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 
-// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY 
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
 // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
-// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 
-// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 
-// 
+// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+//
 // Intel Corporation is the author of this code, and requests that all
-// problem reports or change requests be submitted to it directly at 
-// http://developer.intel.com/opensource.
+// problem reports or change requests be submitted to it directly at
+// http://www.intel.com/software/products/opensource/libraries/num.htm.
 //
 // History
 //==============================================================
-// 10/25/2000: Created
+// 10/25/00 Initial version
+// 06/14/01 Changed cmp to an equivalent form
+// 05/20/02 Cleaned up namespace and sf0 syntax
+// 01/20/03 Improved performance and reduced code size
+// 04/18/03 Eliminate possible WAW dependency warning
 //==============================================================
-//
+
 // API
 //==============================================================
 // long double roundl(long double x)
-//
+//==============================================================
 
-#include "libm_support.h"
+// general input registers:
+// r14 - r19
 
-// general input registers:  
-//
-roundl_GR_half      = r14
-roundl_GR_big       = r15
-roundl_GR_expmask   = r16
-roundl_GR_signexp   = r17
-roundl_GR_exp       = r18
-roundl_GR_expdiff   = r19
-
-// predicate registers used: 
-// p6 - p10
+rSignexp   = r14
+rExp       = r15
+rExpMask   = r16
+rBigexp    = r17
+rExpHalf   = r18
+rExpMHalf  = r19
+
+// floating-point registers:
+// f8 - f13
 
-// floating-point registers used: 
+fXtruncInt = f9
+fNormX     = f10
+fHalf      = f11
+fMHalf     = f12
+fRem       = f13
 
-ROUNDL_NORM_f8        = f9                        
-ROUNDL_TRUNC_f8       = f10
-ROUNDL_RINT_f8        = f11
-ROUNDL_FLOAT_TRUNC_f8 = f12
-ROUNDL_FLOAT_RINT_f8  = f13
-ROUNDL_REMAINDER      = f14
-ROUNDL_HALF           = f15
+// predicate registers used:
+// p6 - p10
 
 // Overview of operation
 //==============================================================
-
 // long double roundl(long double x)
-// Return an integer value (represented as a long double) that is x 
-// rounded to nearest integer, halfway cases rounded away from 
-// zero. 
+// Return an integer value (represented as a long double) that is x
+// rounded to nearest integer, halfway cases rounded away from
+// zero.
 //  if x>0   result = trunc(x+0.5)
 //  if x<0   result = trunc(x-0.5)
-// *******************************************************************************
-
-// Set denormal flag for denormal input and
-// and take denormal fault if necessary.
+//
+//==============================================================
 
-// If x is NAN, ZERO, INFINITY, or >= 2^63 then return
+// double_extended
+// if the exponent is > 1003e => 3F(true) = 63(decimal)
+// we have a significand of 64 bits 1.63-bits.
+// If we multiply by 2^63, we no longer have a fractional part
+// So input is an integer value already.
 
-// qnan snan inf norm     unorm 0 -+
-// 1    1    1   0        0     1 11     0xe7
+// double
+// if the exponent is >= 10033 => 34(true) = 52(decimal)
+// 34 + 3ff = 433
+// we have a significand of 53 bits 1.52-bits. (implicit 1)
+// If we multiply by 2^52, we no longer have a fractional part
+// So input is an integer value already.
 
+// single
+// if the exponent is > 10016 => 17(true) = 23(decimal)
+// we have a significand of 24 bits 1.23-bits. (implicit 1)
+// If we multiply by 2^23, we no longer have a fractional part
+// So input is an integer value already.
 
-.align 32
-.global roundl#
 
 .section .text
-.proc  roundl#
-.align 32
-
+GLOBAL_LIBM_ENTRY(roundl)
 
-roundl: 
-	
-// Get exponent for +0.5
-// Truncate x to integer
 { .mfi
-      addl           roundl_GR_half  = 0x0fffe, r0
-      fcvt.fx.trunc.s1     ROUNDL_TRUNC_f8 = f8
-      nop.i 999
-}
-	
-// Get signexp of x
-// Normalize input
-// Form exponent mask
-{ .mfi
-      getf.exp  roundl_GR_signexp = f8
-      fnorm     ROUNDL_NORM_f8 = f8                        
-      addl      roundl_GR_expmask  = 0x1ffff, r0 ;;
+      getf.exp         rSignexp  = f8        // Get signexp, recompute if unorm
+      fcvt.fx.trunc.s1 fXtruncInt  = f8      // Convert to int in significand
+      addl             rBigexp = 0x1003e, r0 // Set exponent at which is integer
 }
-
-// Form +0.5
-// Round x to integer
 { .mfi
-      setf.exp    ROUNDL_HALF  = roundl_GR_half                      
-      fcvt.fx.s1  ROUNDL_RINT_f8 = f8
-      nop.i 999 ;;
+      mov              rExpHalf    = 0x0FFFE // Form sign and exponent of 0.5
+      fnorm.s1         fNormX  = f8          // Normalize input
+      mov              rExpMask    = 0x1FFFF // Form exponent mask
 }
-// Get exp of x
-// Test for NAN, INF, ZERO
-// Get exponent at which input has no fractional part
-{ .mfi
-      and         roundl_GR_exp = roundl_GR_expmask, roundl_GR_signexp
-      fclass.m    p8,p9 = f8,0xe7
-      addl        roundl_GR_big  = 0x1003e, r0 ;;
-}
-
-// Get exp-bigexp
-// If exp is so big there is no fractional part, then turn on p8, off p9
-{ .mmi
-      sub    roundl_GR_expdiff = roundl_GR_exp, roundl_GR_big ;;
-#ifdef _LIBC
-(p9)  cmp.lt.or.andcm  p8,p9 = r0, roundl_GR_expdiff
-#else
-(p9)  cmp.ge.or.andcm  p8,p9 = roundl_GR_expdiff, r0
-#endif
-      nop.i 999 ;;
-}
-     
-// Set p6 if x<0, else set p7
-{ .mfi
-      nop.m 999
-(p9)  fcmp.lt.unc  p6,p7 = f8,f0
-      nop.i 999
+;;
+
+{ .mmf
+      setf.exp         fHalf = rExpHalf      // Form 0.5
+      mov              rExpMHalf   = 0x2FFFE // Form sign and exponent of -0.5
+      fclass.m         p7,p0 = f8, 0x0b      // Test x unorm
 }
-	
-// If NAN, INF, ZERO, or no fractional part, result is just normalized input
-{ .mfi
-      nop.m 999
-(p8)  fnorm.s0  f8 = f8
-      nop.i 999 ;;
+;;
+
+{ .mfb
+      setf.exp         fMHalf = rExpMHalf    // Form -0.5
+      fclass.m         p6,p0 = f8, 0x1e3     // Test x natval, nan, inf
+(p7)  br.cond.spnt     ROUND_UNORM           // Branch if x unorm
 }
+;;
 
-// Float the truncated integer
+ROUND_COMMON:
+// Return here from ROUND_UNORM
 { .mfi
-      nop.m 999
-(p9)  fcvt.xf     ROUNDL_FLOAT_TRUNC_f8 = ROUNDL_TRUNC_f8
-      nop.i 999 ;;
+      nop.m            0
+      fcmp.lt.s1       p8,p9 = f8, f0        // Test if x < 0
+      nop.i            0
+}
+{ .mfb
+      and              rExp = rSignexp, rExpMask // Get biased exponent
+(p6)  fma.s0           f8 = f8, f1, f0       // Result if x natval, nan, inf
+(p6)  br.ret.spnt      b0                    // Exit if x natval, nan, inf
 }
+;;
 
-// Float the rounded integer to get preliminary result
 { .mfi
-      nop.m 999
-(p9)  fcvt.xf     ROUNDL_FLOAT_RINT_f8 = ROUNDL_RINT_f8
-      nop.i 999 ;;
-}
-
-// If x<0 and the difference of the truncated input minus the input is 0.5
-//    then result = truncated input - 1.0
-// Else if x>0 and the difference of the input minus truncated input is 0.5
-//    then result = truncated input + 1.0
-// Else 
-//    result = rounded input
-// Endif
-{ .mfi
-      nop.m 999
-(p6)  fsub.s1   ROUNDL_REMAINDER = ROUNDL_FLOAT_TRUNC_f8, ROUNDL_NORM_f8 
-      nop.i 999
+      cmp.lt           p6,p0 = rExp, rExpHalf // Is |x| < 0.5?
+      fcvt.xf          f8 = fXtruncInt        // Pre-Result if 0.5 <= |x| < 2^63
+      cmp.ge           p7,p0 = rExp, rBigexp  // Is |x| >= 2^63?
 }
-	
 { .mfi
-      nop.m 999
-(p7)  fsub.s1   ROUNDL_REMAINDER = ROUNDL_NORM_f8, ROUNDL_FLOAT_TRUNC_f8
-      nop.i 999 ;;
+      cmp.lt           p10,p0 = rExp, rExpHalf // Is |x| < 0.5? 
+      nop.f            0
+      nop.i            0
 }
+;;
 
-// Assume preliminary result is rounded integer
+// We must correct result if |x| < 0.5, or |x| >= 2^63
+.pred.rel "mutex",p6,p7
 { .mfi
-      nop.m 999
-(p9)  fnorm.s0  f8 = ROUNDL_FLOAT_RINT_f8
-      nop.i 999
+      nop.m            0
+(p6)  fmerge.s         f8 = fNormX, f0        // If |x| < 0.5, result sgn(x)*0
+      nop.i            0
 }
-
-// If x<0, test if result=0
-{ .mfi
-      nop.m 999
-(p6)  fcmp.eq.unc  p10,p0 = ROUNDL_FLOAT_RINT_f8,f0
-      nop.i 999 ;;
+{ .mfb
+(p7)  cmp.eq           p10,p0 = r0, r0        // Also turn on p10 if |x| >= 2^63
+(p7)  fma.s0           f8 = fNormX, f1, f0    // If |x| >= 2^63, result x
+(p10) br.ret.spnt      b0                     // Exit |x| < 0.5 or |x| >= 2^63
 }
+;;
 
-// If x<0 and result=0, set result=-0
+// Here if 0.5 <= |x| < 2^63
 { .mfi
-      nop.m 999
-(p10) fmerge.ns  f8 = f1,f8
-      nop.i 999
+      nop.m            0
+      fms.s1           fRem = fNormX, f1, f8  // Get remainder = x - trunc(x)
+      nop.i            0
 }
-	
-// If x<0, test if remainder=0.5
+;;
+
 { .mfi
-      nop.m 999
-(p6)  fcmp.eq.unc  p6,p0 = ROUNDL_REMAINDER, ROUNDL_HALF
-      nop.i 999 ;;
+      nop.m            0
+(p8)  fcmp.le.s1       p8,p0 = fRem, fMHalf
+      nop.i            0
 }
-	
-// If x>0, test if remainder=0.5
 { .mfi
-      nop.m 999
-(p7)  fcmp.eq.unc  p7,p0 = ROUNDL_REMAINDER, ROUNDL_HALF
-      nop.i 999 ;;
+      nop.m            0
+(p9)  fcmp.ge.s1       p9,p0 = fRem, fHalf
+      nop.i            0
 }
+;;
 
-// If x<0 and remainder=0.5, result=truncated-1.0
-// If x>0 and remainder=0.5, result=truncated+1.0
-// Exit
-.pred.rel "mutex",p6,p7
+// If x < 0 and remainder <= -0.5, then subtract 1 from result
+// If x > 0 and remainder >= +0.5, then add 1 to result
+.pred.rel "mutex",p8,p9
 { .mfi
-      nop.m 999
-(p6)  fsub.s0  f8 = ROUNDL_FLOAT_TRUNC_f8,f1
-      nop.i 999 
+      nop.m            0
+(p8)  fms.s0           f8 = f8, f1, f1
+      nop.i            0
 }
-	
 { .mfb
-      nop.m 999
-(p7)  fadd.s0  f8 = ROUNDL_FLOAT_TRUNC_f8,f1
-      br.ret.sptk  b0 ;;
+      nop.m            0
+(p9)  fma.s0           f8 = f8, f1, f1
+      br.ret.sptk      b0
+}
+;;
+
+
+ROUND_UNORM:
+// Here if x unorm
+{ .mfb
+      getf.exp         rSignexp  = fNormX     // Get signexp, recompute if unorm
+      fcmp.eq.s0       p7,p0 = f8, f0         // Dummy op to set denormal flag
+      br.cond.sptk     ROUND_COMMON           // Return to main path
 }
+;;
 
-.endp roundl
-ASM_SIZE_DIRECTIVE(roundl)
+GLOBAL_LIBM_END(roundl)