mirror of
https://review.haiku-os.org/buildtools
synced 2025-02-12 08:47:41 +01:00
f133d007f9
Old version was from 2012-05-06, 6.1.2 is from 2016-12-16 A lot of support for newer processors and speedups since then See gmp/NEWS for details
86 lines
2.8 KiB
NASM
86 lines
2.8 KiB
NASM
dnl IBM POWER mpn_sub_n -- Subtract two limb vectors of equal, non-zero
|
|
dnl length.
|
|
|
|
dnl Copyright 1992, 1994-1996, 1999-2001, 2005 Free Software Foundation, Inc.
|
|
|
|
dnl This file is part of the GNU MP Library.
|
|
dnl
|
|
dnl The GNU MP Library is free software; you can redistribute it and/or modify
|
|
dnl it under the terms of either:
|
|
dnl
|
|
dnl * the GNU Lesser General Public License as published by the Free
|
|
dnl Software Foundation; either version 3 of the License, or (at your
|
|
dnl option) any later version.
|
|
dnl
|
|
dnl or
|
|
dnl
|
|
dnl * the GNU General Public License as published by the Free Software
|
|
dnl Foundation; either version 2 of the License, or (at your option) any
|
|
dnl later version.
|
|
dnl
|
|
dnl or both in parallel, as here.
|
|
dnl
|
|
dnl The GNU MP Library is distributed in the hope that it will be useful, but
|
|
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
|
|
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
dnl for more details.
|
|
dnl
|
|
dnl You should have received copies of the GNU General Public License and the
|
|
dnl GNU Lesser General Public License along with the GNU MP Library. If not,
|
|
dnl see https://www.gnu.org/licenses/.
|
|
|
|
|
|
dnl INPUT PARAMETERS
|
|
dnl res_ptr r3
|
|
dnl s1_ptr r4
|
|
dnl s2_ptr r5
|
|
dnl size r6
|
|
|
|
include(`../config.m4')
|
|
|
|
ASM_START()
|
|
PROLOGUE(mpn_sub_n)
|
|
andil. 10,6,1 C odd or even number of limbs?
|
|
l 8,0(4) C load least significant s1 limb
|
|
l 0,0(5) C load least significant s2 limb
|
|
cal 3,-4(3) C offset res_ptr, it's updated before it's used
|
|
sri 10,6,1 C count for unrolled loop
|
|
sf 7,0,8 C subtract least significant limbs, set cy
|
|
mtctr 10 C copy count into CTR
|
|
beq 0,Leven C branch if even # of limbs (# of limbs >= 2)
|
|
|
|
C We have an odd # of limbs. Add the first limbs separately.
|
|
cmpi 1,10,0 C is count for unrolled loop zero?
|
|
bc 4,6,L1 C bne cr1,L1 (misassembled by gas)
|
|
st 7,4(3)
|
|
sfe 3,0,0 C load !cy into ...
|
|
sfi 3,3,0 C ... return value register
|
|
br C return
|
|
|
|
C We added least significant limbs. Now reload the next limbs to enter loop.
|
|
L1: lu 8,4(4) C load s1 limb and update s1_ptr
|
|
lu 0,4(5) C load s2 limb and update s2_ptr
|
|
stu 7,4(3)
|
|
sfe 7,0,8 C subtract limbs, set cy
|
|
Leven: lu 9,4(4) C load s1 limb and update s1_ptr
|
|
lu 10,4(5) C load s2 limb and update s2_ptr
|
|
bdz Lend C If done, skip loop
|
|
|
|
Loop: lu 8,4(4) C load s1 limb and update s1_ptr
|
|
lu 0,4(5) C load s2 limb and update s2_ptr
|
|
sfe 11,10,9 C subtract previous limbs with cy, set cy
|
|
stu 7,4(3) C
|
|
lu 9,4(4) C load s1 limb and update s1_ptr
|
|
lu 10,4(5) C load s2 limb and update s2_ptr
|
|
sfe 7,0,8 C subtract previous limbs with cy, set cy
|
|
stu 11,4(3) C
|
|
bdn Loop C decrement CTR and loop back
|
|
|
|
Lend: sfe 11,10,9 C subtract limbs with cy, set cy
|
|
st 7,4(3) C
|
|
st 11,8(3) C
|
|
sfe 3,0,0 C load !cy into ...
|
|
sfi 3,3,0 C ... return value register
|
|
br
|
|
EPILOGUE(mpn_sub_n)
|