2004-06-30 23:43:49 +00:00
|
|
|
/* lround function. PowerPC32 version.
|
2016-01-04 16:05:18 +00:00
|
|
|
Copyright (C) 2004-2016 Free Software Foundation, Inc.
|
2004-06-30 23:43:49 +00:00
|
|
|
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
|
2012-02-09 23:18:22 +00:00
|
|
|
License along with the GNU C Library; if not, see
|
|
|
|
|
<http://www.gnu.org/licenses/>. */
|
2004-06-30 23:43:49 +00:00
|
|
|
|
|
|
|
|
#include <sysdep.h>
|
2006-01-28 00:15:15 +00:00
|
|
|
#include <math_ldbl_opt.h>
|
2004-12-22 20:10:10 +00:00
|
|
|
|
2008-04-11 19:32:37 +00:00
|
|
|
.section .rodata.cst4,"aM",@progbits,4
|
2005-05-21 19:29:13 +00:00
|
|
|
.align 2
|
2008-04-11 19:32:37 +00:00
|
|
|
.LC0: /* 0.5 */
|
2005-05-21 19:29:13 +00:00
|
|
|
.long 0x3f000000
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
.LC1: /* 2^52. */
|
|
|
|
|
.long 0x59800000
|
|
|
|
|
.section .rodata.cst8,"aM",@progbits,8
|
|
|
|
|
.align 3
|
|
|
|
|
.LC2: /* 0x7fffffff.8p0. */
|
|
|
|
|
.long 0x41dfffff
|
|
|
|
|
.long 0xffe00000
|
|
|
|
|
.LC3: /* -0x80000000.8p0. */
|
|
|
|
|
.long 0xc1e00000
|
|
|
|
|
.long 0x00100000
|
2004-06-30 23:43:49 +00:00
|
|
|
.section ".text"
|
2013-06-05 20:44:03 +00:00
|
|
|
|
2004-06-30 23:43:49 +00:00
|
|
|
/* long [r3] lround (float x [fp1])
|
2013-06-05 20:44:03 +00:00
|
|
|
IEEE 1003.1 lround function. IEEE specifies "round to the nearest
|
2004-06-30 23:43:49 +00:00
|
|
|
integer value, rounding halfway cases away from zero, regardless of
|
|
|
|
|
the current rounding mode." However PowerPC Architecture defines
|
2013-06-05 20:44:03 +00:00
|
|
|
"round to Nearest" as "Choose the best approximation. In case of a
|
|
|
|
|
tie, choose the one that is even (least significant bit o).".
|
2004-06-30 23:43:49 +00:00
|
|
|
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
|
2008-04-11 19:32:37 +00:00
|
|
|
to the integer value. It is necessary to detect when x is
|
|
|
|
|
(+-)0x1.fffffffffffffp-2 because adding +-0.5 in this case will
|
|
|
|
|
cause an erroneous shift, carry and round. We simply return 0 if
|
|
|
|
|
0.5 > x > -0.5. */
|
2004-06-30 23:43:49 +00:00
|
|
|
|
|
|
|
|
ENTRY (__lround)
|
2006-01-07 03:58:05 +00:00
|
|
|
stwu r1,-16(r1)
|
|
|
|
|
cfi_adjust_cfa_offset (16)
|
2004-06-30 23:43:49 +00:00
|
|
|
#ifdef SHARED
|
|
|
|
|
mflr r11
|
2006-01-07 03:58:05 +00:00
|
|
|
cfi_register(lr,r11)
|
2011-12-17 20:07:31 +00:00
|
|
|
SETUP_GOT_ACCESS(r9,got_label)
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
addis r10,r9,.LC0-got_label@ha
|
|
|
|
|
lfs fp10,.LC0-got_label@l(r10)
|
|
|
|
|
addis r10,r9,.LC1-got_label@ha
|
|
|
|
|
lfs fp11,.LC1-got_label@l(r10)
|
|
|
|
|
addis r10,r9,.LC2-got_label@ha
|
|
|
|
|
lfd fp9,.LC2-got_label@l(r10)
|
|
|
|
|
addis r10,r9,.LC3-got_label@ha
|
|
|
|
|
lfd fp8,.LC3-got_label@l(r10)
|
2004-06-30 23:43:49 +00:00
|
|
|
mtlr r11
|
2006-01-07 03:58:05 +00:00
|
|
|
cfi_same_value (lr)
|
2004-06-30 23:43:49 +00:00
|
|
|
#else
|
|
|
|
|
lis r9,.LC0@ha
|
2008-04-11 19:32:37 +00:00
|
|
|
lfs fp10,.LC0@l(r9)
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
lis r9,.LC1@ha
|
|
|
|
|
lfs fp11,.LC1@l(r9)
|
|
|
|
|
lis r9,.LC2@ha
|
|
|
|
|
lfd fp9,.LC2@l(r9)
|
|
|
|
|
lis r9,.LC3@ha
|
|
|
|
|
lfd fp8,.LC3@l(r9)
|
2004-06-30 23:43:49 +00:00
|
|
|
#endif
|
2008-04-11 19:32:37 +00:00
|
|
|
fabs fp2, fp1 /* Get the absolute value of x. */
|
|
|
|
|
fsub fp12,fp10,fp10 /* Compute 0.0. */
|
|
|
|
|
fcmpu cr6, fp2, fp10 /* if |x| < 0.5 */
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
fcmpu cr5, fp1, fp9 /* if x >= 0x7fffffff.8p0 */
|
|
|
|
|
fcmpu cr1, fp1, fp8 /* if x <= -0x80000000.8p0 */
|
2008-06-13 01:18:22 +00:00
|
|
|
fcmpu cr7, fp1, fp12 /* x is negative? x < 0.0 */
|
2008-04-11 19:32:37 +00:00
|
|
|
blt- cr6,.Lretzero
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
bge- cr5,.Loflow
|
|
|
|
|
ble- cr1,.Loflow
|
|
|
|
|
/* Test whether an integer to avoid spurious "inexact". */
|
|
|
|
|
fadd fp3,fp2,fp11
|
|
|
|
|
fsub fp3,fp3,fp11
|
|
|
|
|
fcmpu cr5, fp2, fp3
|
|
|
|
|
beq cr5,.Lnobias
|
2008-04-11 19:32:37 +00:00
|
|
|
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
|
2008-06-13 01:18:22 +00:00
|
|
|
bge cr7,.Lconvert /* x is positive so don't negate x. */
|
2013-06-05 20:44:03 +00:00
|
|
|
fnabs fp3,fp3 /* -(|x|+=0.5) */
|
2008-04-11 19:32:37 +00:00
|
|
|
.Lconvert:
|
|
|
|
|
fctiwz fp4,fp3 /* Convert to Integer word lround toward 0. */
|
|
|
|
|
stfd fp4,8(r1)
|
2005-05-21 19:01:22 +00:00
|
|
|
nop /* Ensure the following load is in a different dispatch */
|
|
|
|
|
nop /* group to avoid pipe stall on POWER4&5. */
|
2004-06-30 23:43:49 +00:00
|
|
|
nop
|
2013-08-17 09:01:45 +00:00
|
|
|
lwz r3,8+LOWORD(r1) /* Load return as integer. */
|
2008-04-11 19:32:37 +00:00
|
|
|
.Lout:
|
2006-01-07 03:58:05 +00:00
|
|
|
addi r1,r1,16
|
2004-06-30 23:43:49 +00:00
|
|
|
blr
|
2008-04-11 19:32:37 +00:00
|
|
|
.Lretzero: /* when 0.5 > x > -0.5 */
|
|
|
|
|
li r3,0 /* return 0. */
|
|
|
|
|
b .Lout
|
Fix powerpc32 lround, lroundf spurious exceptions (bug 19134).
The powerpc32 implementation of lround and lroundf can produce
spurious exceptions from adding 0.5 then converting to integer. This
includes "inexact" from the conversion to integer (not allowed for
integer arguments to these functions), and, for larger integer
arguments, "inexact", and "overflow" when rounding upward, from the
addition. In addition, "inexact" is not allowed together with
"invalid" and so inexact addition must be avoided when the integer
will be out of range of 32-bit long, whether or not the argument is an
integer.
This patch fixes these problems. As in the powerpc64 llround
implementation, a check is added for too-large arguments; in the
powerpc64 case that means arguments at least 2^52 in magnitude (so
that 0.5 cannot be added exactly), while in this case it means
arguments for which the result would overflow "long". In those cases
a suitable overflowing value is used for the integer conversion
without adding 0.5, while for smaller arguments it's tested whether
the argument is an integer (by adding and subtracting 2^52 to the
absolute value and comparing with the original absolute value) to
avoid adding 0.5 to integers and generating spurious "inexact".
This code is not used when the power5+ sysdeps directories are used,
as there's a separate power5+ version of these functions..
Tested for powerpc. This gets test-float (for a default powerpc32
hard-float build without any --with-cpu) back to the point where it
should pass once powerpc ulps are regenerated; test-double still needs
another problem with exceptions fixed to get back to that point (and I
haven't looked lately at what default powerpc64 results are like).
[BZ #19134]
* sysdeps/powerpc/powerpc32/fpu/s_lround.S (.LC1): New object.
(.LC2): Likewise.
(.LC3): Likewise.
(__lround): Do not add 0.5 to integer or out-of-range arguments.
2015-10-14 21:13:42 +00:00
|
|
|
.Lnobias:
|
|
|
|
|
fmr fp3,fp1
|
|
|
|
|
b .Lconvert
|
|
|
|
|
.Loflow:
|
|
|
|
|
fmr fp3,fp11
|
|
|
|
|
bge cr7,.Lconvert
|
|
|
|
|
fnabs fp3,fp3
|
|
|
|
|
b .Lconvert
|
2004-06-30 23:43:49 +00:00
|
|
|
END (__lround)
|
|
|
|
|
|
|
|
|
|
weak_alias (__lround, lround)
|
|
|
|
|
|
|
|
|
|
strong_alias (__lround, __lroundf)
|
|
|
|
|
weak_alias (__lround, lroundf)
|
|
|
|
|
|
|
|
|
|
#ifdef NO_LONG_DOUBLE
|
|
|
|
|
weak_alias (__lround, lroundl)
|
|
|
|
|
strong_alias (__lround, __lroundl)
|
|
|
|
|
#endif
|
2006-01-28 00:15:15 +00:00
|
|
|
#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
|
|
|
|
|
compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
|
|
|
|
|
#endif
|
