glibc/sysdeps/m88k/m88110/sub_n.S
1996-05-07 21:04:52 +00:00

277 lines
6.4 KiB
ArmAsm

; mc88110 __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
; store difference in a third limb vector.
; Copyright (C) 1995, 1996 Free Software Foundation, Inc.
; This file is part of the GNU MP Library.
; The GNU MP 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 MP 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 MP 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.
; INPUT PARAMETERS
#define res_ptr r2
#define s1_ptr r3
#define s2_ptr r4
#define size r5
#include "sysdep.h"
text
align 16
global C_SYMBOL_NAME(__mpn_sub_n)
C_SYMBOL_NAME(__mpn_sub_n):
subu.co r0,r0,r0 ; set cy flag
xor r12,s2_ptr,res_ptr
bb1 2,r12,L1
; ** V1a **
L0: bb0 2,res_ptr,L_v1 ; branch if res_ptr is aligned
/* Add least significant limb separately to align res_ptr and s2_ptr */
ld r10,s1_ptr,0
addu s1_ptr,s1_ptr,4
ld r8,s2_ptr,0
addu s2_ptr,s2_ptr,4
subu size,size,1
subu.co r6,r10,r8
st r6,res_ptr,0
addu res_ptr,res_ptr,4
L_v1: cmp r12,size,2
bb1 lt,r12,Lend2
ld r10,s1_ptr,0
ld r12,s1_ptr,4
ld.d r8,s2_ptr,0
subu size,size,10
bcnd lt0,size,Lfin1
/* Add blocks of 8 limbs until less than 8 limbs remain */
align 8
Loop1: subu size,size,8
subu.cio r6,r10,r8
ld r10,s1_ptr,8
subu.cio r7,r12,r9
ld r12,s1_ptr,12
ld.d r8,s2_ptr,8
st.d r6,res_ptr,0
subu.cio r6,r10,r8
ld r10,s1_ptr,16
subu.cio r7,r12,r9
ld r12,s1_ptr,20
ld.d r8,s2_ptr,16
st.d r6,res_ptr,8
subu.cio r6,r10,r8
ld r10,s1_ptr,24
subu.cio r7,r12,r9
ld r12,s1_ptr,28
ld.d r8,s2_ptr,24
st.d r6,res_ptr,16
subu.cio r6,r10,r8
ld r10,s1_ptr,32
subu.cio r7,r12,r9
ld r12,s1_ptr,36
addu s1_ptr,s1_ptr,32
ld.d r8,s2_ptr,32
addu s2_ptr,s2_ptr,32
st.d r6,res_ptr,24
addu res_ptr,res_ptr,32
bcnd ge0,size,Loop1
Lfin1: addu size,size,8-2
bcnd lt0,size,Lend1
/* Add blocks of 2 limbs until less than 2 limbs remain */
Loope1: subu.cio r6,r10,r8
ld r10,s1_ptr,8
subu.cio r7,r12,r9
ld r12,s1_ptr,12
ld.d r8,s2_ptr,8
st.d r6,res_ptr,0
subu size,size,2
addu s1_ptr,s1_ptr,8
addu s2_ptr,s2_ptr,8
addu res_ptr,res_ptr,8
bcnd ge0,size,Loope1
Lend1: subu.cio r6,r10,r8
subu.cio r7,r12,r9
st.d r6,res_ptr,0
bb0 0,size,Lret1
/* Add last limb */
ld r10,s1_ptr,8
ld r8,s2_ptr,8
subu.cio r6,r10,r8
st r6,res_ptr,8
Lret1: addu.ci r2,r0,r0 ; return carry-out from most sign. limb
jmp.n r1
xor r2,r2,1
L1: xor r12,s1_ptr,res_ptr
bb1 2,r12,L2
; ** V1b **
bb0 2,res_ptr,L_v1b ; branch if res_ptr is aligned
/* Add least significant limb separately to align res_ptr and s1_ptr */
ld r10,s2_ptr,0
addu s2_ptr,s2_ptr,4
ld r8,s1_ptr,0
addu s1_ptr,s1_ptr,4
subu size,size,1
subu.co r6,r8,r10
st r6,res_ptr,0
addu res_ptr,res_ptr,4
L_v1b: cmp r12,size,2
bb1 lt,r12,Lend2
ld r10,s2_ptr,0
ld r12,s2_ptr,4
ld.d r8,s1_ptr,0
subu size,size,10
bcnd lt0,size,Lfin1b
/* Add blocks of 8 limbs until less than 8 limbs remain */
align 8
Loop1b: subu size,size,8
subu.cio r6,r8,r10
ld r10,s2_ptr,8
subu.cio r7,r9,r12
ld r12,s2_ptr,12
ld.d r8,s1_ptr,8
st.d r6,res_ptr,0
subu.cio r6,r8,r10
ld r10,s2_ptr,16
subu.cio r7,r9,r12
ld r12,s2_ptr,20
ld.d r8,s1_ptr,16
st.d r6,res_ptr,8
subu.cio r6,r8,r10
ld r10,s2_ptr,24
subu.cio r7,r9,r12
ld r12,s2_ptr,28
ld.d r8,s1_ptr,24
st.d r6,res_ptr,16
subu.cio r6,r8,r10
ld r10,s2_ptr,32
subu.cio r7,r9,r12
ld r12,s2_ptr,36
addu s2_ptr,s2_ptr,32
ld.d r8,s1_ptr,32
addu s1_ptr,s1_ptr,32
st.d r6,res_ptr,24
addu res_ptr,res_ptr,32
bcnd ge0,size,Loop1b
Lfin1b: addu size,size,8-2
bcnd lt0,size,Lend1b
/* Add blocks of 2 limbs until less than 2 limbs remain */
Loope1b:subu.cio r6,r8,r10
ld r10,s2_ptr,8
subu.cio r7,r9,r12
ld r12,s2_ptr,12
ld.d r8,s1_ptr,8
st.d r6,res_ptr,0
subu size,size,2
addu s1_ptr,s1_ptr,8
addu s2_ptr,s2_ptr,8
addu res_ptr,res_ptr,8
bcnd ge0,size,Loope1b
Lend1b: subu.cio r6,r8,r10
subu.cio r7,r9,r12
st.d r6,res_ptr,0
bb0 0,size,Lret1b
/* Add last limb */
ld r10,s2_ptr,8
ld r8,s1_ptr,8
subu.cio r6,r8,r10
st r6,res_ptr,8
Lret1b: addu.ci r2,r0,r0 ; return carry-out from most sign. limb
jmp.n r1
xor r2,r2,1
; ** V2 **
/* If we come here, the alignment of s1_ptr and res_ptr as well as the
alignment of s2_ptr and res_ptr differ. Since there are only two ways
things can be aligned (that we care about) we now know that the alignment
of s1_ptr and s2_ptr are the same. */
L2: cmp r12,size,1
bb1 eq,r12,Ljone
bb0 2,s1_ptr,L_v2 ; branch if s1_ptr is aligned
/* Add least significant limb separately to align res_ptr and s2_ptr */
ld r10,s1_ptr,0
addu s1_ptr,s1_ptr,4
ld r8,s2_ptr,0
addu s2_ptr,s2_ptr,4
subu size,size,1
subu.co r6,r10,r8
st r6,res_ptr,0
addu res_ptr,res_ptr,4
L_v2: subu size,size,8
bcnd lt0,size,Lfin2
/* Add blocks of 8 limbs until less than 8 limbs remain */
align 8
Loop2: subu size,size,8
ld.d r8,s1_ptr,0
ld.d r6,s2_ptr,0
subu.cio r8,r8,r6
st r8,res_ptr,0
subu.cio r9,r9,r7
st r9,res_ptr,4
ld.d r8,s1_ptr,8
ld.d r6,s2_ptr,8
subu.cio r8,r8,r6
st r8,res_ptr,8
subu.cio r9,r9,r7
st r9,res_ptr,12
ld.d r8,s1_ptr,16
ld.d r6,s2_ptr,16
subu.cio r8,r8,r6
st r8,res_ptr,16
subu.cio r9,r9,r7
st r9,res_ptr,20
ld.d r8,s1_ptr,24
ld.d r6,s2_ptr,24
subu.cio r8,r8,r6
st r8,res_ptr,24
subu.cio r9,r9,r7
st r9,res_ptr,28
addu s1_ptr,s1_ptr,32
addu s2_ptr,s2_ptr,32
addu res_ptr,res_ptr,32
bcnd ge0,size,Loop2
Lfin2: addu size,size,8-2
bcnd lt0,size,Lend2
Loope2: ld.d r8,s1_ptr,0
ld.d r6,s2_ptr,0
subu.cio r8,r8,r6
st r8,res_ptr,0
subu.cio r9,r9,r7
st r9,res_ptr,4
subu size,size,2
addu s1_ptr,s1_ptr,8
addu s2_ptr,s2_ptr,8
addu res_ptr,res_ptr,8
bcnd ge0,size,Loope2
Lend2: bb0 0,size,Lret2
/* Add last limb */
Ljone: ld r10,s1_ptr,0
ld r8,s2_ptr,0
subu.cio r6,r10,r8
st r6,res_ptr,0
Lret2: addu.ci r2,r0,r0 ; return carry-out from most sign. limb
jmp.n r1
xor r2,r2,1