Home | History | Annotate | Line # | Download | only in v9
sub_n.asm revision 1.1
      1 dnl  SPARC mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
      2 dnl  store difference in a third limb vector.
      3 
      4 dnl  Copyright 2001 Free Software Foundation, Inc.
      5 
      6 dnl  This file is part of the GNU MP Library.
      7 
      8 dnl  The GNU MP Library is free software; you can redistribute it and/or modify
      9 dnl  it under the terms of the GNU Lesser General Public License as published
     10 dnl  by the Free Software Foundation; either version 3 of the License, or (at
     11 dnl  your option) any later version.
     12 
     13 dnl  The GNU MP Library is distributed in the hope that it will be useful, but
     14 dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
     15 dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
     16 dnl  License for more details.
     17 
     18 dnl  You should have received a copy of the GNU Lesser General Public License
     19 dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
     20 
     21 
     22 include(`../config.m4')
     23 
     24 C INPUT PARAMETERS
     25 define(rp,%o0)
     26 define(s1p,%o1)
     27 define(s2p,%o2)
     28 define(n,%o3)
     29 define(cy,%g1)
     30 
     31 C This code uses 64-bit operations on `o' and `g' registers.  It doesn't
     32 C require that `o' registers' upper 32 bits are preserved by the operating
     33 C system, but if they are not, they must be zeroed.  That is indeed what
     34 C happens at least on Slowaris 2.5 and 2.6.
     35 
     36 C On UltraSPARC 1 and 2, this code runs at 3 cycles/limb from the Dcache and at
     37 C about 10 cycles/limb from the Ecache.
     38 
     39 ASM_START()
     40 PROLOGUE(mpn_sub_n)
     41 	lduw	[s1p+0],%o4
     42 	lduw	[s2p+0],%o5
     43 	addcc	n,-2,n
     44 	bl,pn	%icc,L(end1)
     45 	lduw	[s1p+4],%g2
     46 	lduw	[s2p+4],%g3
     47 	be,pn	%icc,L(end2)
     48 	mov	0,cy
     49 
     50 	.align	16
     51 L(loop):
     52 	sub	%o4,%o5,%g4
     53 	add	rp,8,rp
     54 	lduw	[s1p+8],%o4
     55 	fitod	%f0,%f2
     56 C ---
     57 	sub	%g4,cy,%g4
     58 	addcc	n,-1,n
     59 	lduw	[s2p+8],%o5
     60 	fitod	%f0,%f2
     61 C ---
     62 	srlx	%g4,63,cy
     63 	add	s2p,8,s2p
     64 	stw	%g4,[rp-8]
     65 	be,pn	%icc,L(exito)+4
     66 C ---
     67 	sub	%g2,%g3,%g4
     68 	addcc	n,-1,n
     69 	lduw	[s1p+12],%g2
     70 	fitod	%f0,%f2
     71 C ---
     72 	sub	%g4,cy,%g4
     73 	add	s1p,8,s1p
     74 	lduw	[s2p+4],%g3
     75 	fitod	%f0,%f2
     76 C ---
     77 	srlx	%g4,63,cy
     78 	bne,pt	%icc,L(loop)
     79 	stw	%g4,[rp-4]
     80 C ---
     81 L(exite):
     82 	sub	%o4,%o5,%g4
     83 	sub	%g4,cy,%g4
     84 	srlx	%g4,63,cy
     85 	stw	%g4,[rp+0]
     86 	sub	%g2,%g3,%g4
     87 	sub	%g4,cy,%g4
     88 	stw	%g4,[rp+4]
     89 	retl
     90 	srlx	%g4,63,%o0
     91 
     92 L(exito):
     93 	sub	%g2,%g3,%g4
     94 	sub	%g4,cy,%g4
     95 	srlx	%g4,63,cy
     96 	stw	%g4,[rp-4]
     97 	sub	%o4,%o5,%g4
     98 	sub	%g4,cy,%g4
     99 	stw	%g4,[rp+0]
    100 	retl
    101 	srlx	%g4,63,%o0
    102 
    103 L(end1):
    104 	sub	%o4,%o5,%g4
    105 	stw	%g4,[rp+0]
    106 	retl
    107 	srlx	%g4,63,%o0
    108 
    109 L(end2):
    110 	sub	%o4,%o5,%g4
    111 	srlx	%g4,63,cy
    112 	stw	%g4,[rp+0]
    113 	sub	%g2,%g3,%g4
    114 	sub	%g4,cy,%g4
    115 	stw	%g4,[rp+4]
    116 	retl
    117 	srlx	%g4,63,%o0
    118 EPILOGUE(mpn_sub_n)
    119