glibc/sysdeps/powerpc/powerpc32/fpu/s_lround.S

130 lines
3.9 KiB
ArmAsm

/* lround function. PowerPC32 version.
Copyright (C) 2004-2016 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 .rodata.cst4,"aM",@progbits,4
.align 2
.LC0: /* 0.5 */
.long 0x3f000000
.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
.section ".text"
/* long [r3] lround (float x [fp1])
IEEE 1003.1 lround function. IEEE specifies "round to the nearest
integer value, rounding halfway cases away from zero, regardless of
the current rounding mode." However PowerPC Architecture defines
"round to Nearest" as "Choose the best approximation. In case of a
tie, choose the one that is even (least significant bit o).".
So we can't use the PowerPC "round to Nearest" mode. Instead we set
"round toward Zero" mode and round by adding +-0.5 before rounding
to the integer value. 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. */
ENTRY (__lround)
stwu r1,-16(r1)
cfi_adjust_cfa_offset (16)
#ifdef SHARED
mflr r11
cfi_register(lr,r11)
SETUP_GOT_ACCESS(r9,got_label)
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)
mtlr r11
cfi_same_value (lr)
#else
lis r9,.LC0@ha
lfs fp10,.LC0@l(r9)
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)
#endif
fabs fp2, fp1 /* Get the absolute value of x. */
fsub fp12,fp10,fp10 /* Compute 0.0. */
fcmpu cr6, fp2, fp10 /* if |x| < 0.5 */
fcmpu cr5, fp1, fp9 /* if x >= 0x7fffffff.8p0 */
fcmpu cr1, fp1, fp8 /* if x <= -0x80000000.8p0 */
fcmpu cr7, fp1, fp12 /* x is negative? x < 0.0 */
blt- cr6,.Lretzero
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
fadd fp3,fp2,fp10 /* |x|+=0.5 bias to prepare to round. */
bge cr7,.Lconvert /* x is positive so don't negate x. */
fnabs fp3,fp3 /* -(|x|+=0.5) */
.Lconvert:
fctiwz fp4,fp3 /* Convert to Integer word lround toward 0. */
stfd fp4,8(r1)
nop /* Ensure the following load is in a different dispatch */
nop /* group to avoid pipe stall on POWER4&5. */
nop
lwz r3,8+LOWORD(r1) /* Load return as integer. */
.Lout:
addi r1,r1,16
blr
.Lretzero: /* when 0.5 > x > -0.5 */
li r3,0 /* return 0. */
b .Lout
.Lnobias:
fmr fp3,fp1
b .Lconvert
.Loflow:
fmr fp3,fp11
bge cr7,.Lconvert
fnabs fp3,fp3
b .Lconvert
END (__lround)
weak_alias (__lround, lround)
strong_alias (__lround, __lroundf)
weak_alias (__lround, lroundf)
#ifdef NO_LONG_DOUBLE
weak_alias (__lround, lroundl)
strong_alias (__lround, __lroundl)
#endif
#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
compat_symbol (libm, __lround, lroundl, GLIBC_2_1)
#endif