Home | History | Annotate | Line # | Download | only in ultrasparct3
aorslsh_n.asm revision 1.1.1.1
      1 dnl  SPARC v9 mpn_addlsh_n and mpn_sublsh_n for T3/T4/T5.
      2 
      3 dnl  Contributed to the GNU project by Torbjrn Granlund.
      4 
      5 dnl  Copyright 2013 Free Software Foundation, Inc.
      6 
      7 dnl  This file is part of the GNU MP Library.
      8 dnl
      9 dnl  The GNU MP Library is free software; you can redistribute it and/or modify
     10 dnl  it under the terms of either:
     11 dnl
     12 dnl    * the GNU Lesser General Public License as published by the Free
     13 dnl      Software Foundation; either version 3 of the License, or (at your
     14 dnl      option) any later version.
     15 dnl
     16 dnl  or
     17 dnl
     18 dnl    * the GNU General Public License as published by the Free Software
     19 dnl      Foundation; either version 2 of the License, or (at your option) any
     20 dnl      later version.
     21 dnl
     22 dnl  or both in parallel, as here.
     23 dnl
     24 dnl  The GNU MP Library is distributed in the hope that it will be useful, but
     25 dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
     26 dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
     27 dnl  for more details.
     28 dnl
     29 dnl  You should have received copies of the GNU General Public License and the
     30 dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
     31 dnl  see https://www.gnu.org/licenses/.
     32 
     33 include(`../config.m4')
     34 
     35 C		   cycles/limb
     36 C UltraSPARC T3:	11
     37 C UltraSPARC T4:	 4
     38 
     39 C For sublsh_n we combine the two shifted limbs using xnor, using the identity
     40 C (a xor not b) = (not (a xor b)) which equals (not (a or b)) when (a and b) =
     41 C 0 as it is in our usage.  This gives us the ones complement for free.
     42 C Unfortunately, the same trick will not work for rsblsh_n, which will instead
     43 C require a separate negation.
     44 C
     45 C FIXME: Add rsblsh_n to this file.
     46 
     47 define(`rp', `%i0')
     48 define(`up', `%i1')
     49 define(`vp', `%i2')
     50 define(`n',  `%i3')
     51 define(`cnt',`%i4')
     52 
     53 define(`tnc',`%o5')
     54 
     55 ifdef(`OPERATION_addlsh_n',`
     56   define(`INITCY', `subcc	%g0, 0, %g0')
     57   define(`MERGE',  `or')
     58   define(`func',   `mpn_addlsh_n')
     59 ')
     60 ifdef(`OPERATION_sublsh_n',`
     61   define(`INITCY', `subcc	%g0, 1, %g0')
     62   define(`MERGE',  `xnor')
     63   define(`func',   `mpn_sublsh_n')
     64 ')
     65 
     66 define(`rp0',  `rp')
     67 define(`rp1',  `%o2')
     68 define(`up0',  `up')
     69 define(`up1',  `%o3')
     70 define(`vp0',  `vp')
     71 define(`vp1',  `%o4')
     72 
     73 MULFUNC_PROLOGUE(mpn_addlsh_n mpn_sublsh_n)
     74 ASM_START()
     75 	REGISTER(%g2,#scratch)
     76 	REGISTER(%g3,#scratch)
     77 PROLOGUE(func)
     78 	save	%sp, -176, %sp
     79 	mov	64, tnc
     80 	sub	tnc, cnt, tnc
     81 
     82 	andcc	n, 1, %g0
     83 	sllx	n, 3, n
     84 	add	n, -16, n
     85 	add	up, n, up0
     86 	add	vp, n, vp0
     87 	add	rp, n, rp0
     88 	add	up0, 8, up1
     89 	add	vp0, 8, vp1
     90 	add	rp0, -8, rp1
     91 	add	rp0, -16, rp0
     92 	neg	n, n
     93 	be	L(evn)
     94 	 INITCY
     95 
     96 L(odd):	ldx	[vp0 + n], %l1
     97 	mov	0, %l2
     98 	ldx	[up0 + n], %l5
     99 	sllx	%l1, cnt, %g3
    100 	brgez	n, L(wd1)
    101 	 add	n, 8, n
    102 	ldx	[vp0 + n], %l0
    103 	b	L(lo1)
    104 	 sllx	%l1, cnt, %g3
    105 
    106 L(evn):	ldx	[vp0 + n], %l0
    107 	mov	0, %l3
    108 	ldx	[up0 + n], %l4
    109 	ldx	[vp1 + n], %l1
    110 	b	L(lo0)
    111 	 sllx	%l0, cnt, %g1
    112 
    113 L(top):	addxccc(%l6, %l4, %o0)
    114 	ldx	[vp0 + n], %l0
    115 	sllx	%l1, cnt, %g3
    116 	stx	%o0, [rp0 + n]
    117 L(lo1):	srlx	%l1, tnc, %l3
    118 	MERGE	%l2, %g3, %l7
    119 	ldx	[up0 + n], %l4
    120 	addxccc(%l7, %l5, %o1)
    121 	ldx	[vp1 + n], %l1
    122 	sllx	%l0, cnt, %g1
    123 	stx	%o1, [rp1 + n]
    124 L(lo0):	srlx	%l0, tnc, %l2
    125 	MERGE	%l3, %g1, %l6
    126 	ldx	[up1 + n], %l5
    127 	brlz,pt	n, L(top)
    128 	 add	n, 16, n
    129 
    130 	addxccc(%l6, %l4, %o0)
    131 	sllx	%l1, cnt, %g3
    132 	stx	%o0, [rp0 + n]
    133 L(wd1):	srlx	%l1, tnc, %l3
    134 	MERGE	%l2, %g3, %l7
    135 	addxccc(%l7, %l5, %o1)
    136 	stx	%o1, [rp1 + n]
    137 
    138 ifdef(`OPERATION_addlsh_n',
    139 `	addxc(	%l3, %g0, %i0)')
    140 ifdef(`OPERATION_sublsh_n',
    141 `	addxc(	%g0, %g0, %g1)
    142 	add	%g1, -1, %g1
    143 	sub	%l3, %g1, %i0')
    144 
    145 	ret
    146 	 restore
    147 EPILOGUE()
    148