source/libs/gmp/gmp-src/mpn/powerpc64/mode64/p7/aorsorrlshC_n.asm

   1 dnl  PowerPC-64 mpn_addlshC_n, mpn_sublshC_n, mpn_rsblshC_n.
   2
   3 dnl  Copyright 2003, 2005, 2009, 2010, 2013 Free Software Foundation, Inc.
   4
   5 dnl  This file is part of the GNU MP Library.
   6 dnl
   7 dnl  The GNU MP Library is free software; you can redistribute it and/or modify
   8 dnl  it under the terms of either:
   9 dnl
  10 dnl    * the GNU Lesser General Public License as published by the Free
  11 dnl      Software Foundation; either version 3 of the License, or (at your
  12 dnl      option) any later version.
  13 dnl
  14 dnl  or
  15 dnl
  16 dnl    * the GNU General Public License as published by the Free Software
  17 dnl      Foundation; either version 2 of the License, or (at your option) any
  18 dnl      later version.
  19 dnl
  20 dnl  or both in parallel, as here.
  21 dnl
  22 dnl  The GNU MP Library is distributed in the hope that it will be useful, but
  23 dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  24 dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  25 dnl  for more details.
  26 dnl
  27 dnl  You should have received copies of the GNU General Public License and the
  28 dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
  29 dnl  see https://www.gnu.org/licenses/.
  30
  31 C                  cycles/limb
  32 C POWER3/PPC630          ?
  33 C POWER4/PPC970          ?
  34 C POWER5                 ?
  35 C POWER6                 ?
  36 C POWER7                 2.5
  37
  38 C INPUT PARAMETERS
  39 define(`rp', `r3')
  40 define(`up', `r4')
  41 define(`vp', `r5')
  42 define(`n',  `r6')
  43
  44 ifdef(`DO_add', `
  45   define(`ADDSUBC',     `addc   $1, $2, $3')
  46   define(`ADDSUBE',     `adde   $1, $2, $3')
  47   define(INITCY,        `addic  $1, r1, 0')
  48   define(RETVAL,        `addze  r3, $1')
  49   define(`func',        mpn_addlsh`'LSH`'_n)')
  50 ifdef(`DO_sub', `
  51   define(`ADDSUBC',     `subfc  $1, $2, $3')
  52   define(`ADDSUBE',     `subfe  $1, $2, $3')
  53   define(INITCY,        `addic  $1, r1, -1')
  54   define(RETVAL,        `subfze r3, $1
  55                         neg     r3, r3')
  56   define(`func',        mpn_sublsh`'LSH`'_n)')
  57 ifdef(`DO_rsb', `
  58   define(`ADDSUBC',     `subfc  $1, $3, $2')
  59   define(`ADDSUBE',     `subfe  $1, $3, $2')
  60   define(INITCY,        `addic  $1, r1, -1')
  61   define(RETVAL,        `addme  r3, $1')
  62   define(`func',        mpn_rsblsh`'LSH`'_n)')
  63
  64 define(`s0', `r0')  define(`s1', `r9')
  65 define(`u0', `r6')  define(`u1', `r7')
  66 define(`v0', `r10') define(`v1', `r11')
  67
  68
  69 ASM_START()
  70 PROLOGUE(func)
  71         rldic   r7, n, 3, 59
  72         add     up, up, r7
  73         add     vp, vp, r7
  74         add     rp, rp, r7
  75
  76 ifdef(`DO_add', `
  77         addic   r0, n, 3        C set cy flag as side effect
  78 ',`
  79         subfc   r0, r0, r0      C set cy flag
  80         addi    r0, n, 3
  81 ')
  82         srdi    r0, r0, 2
  83         mtctr   r0
  84
  85         andi.   r0, n, 1
  86         beq     L(bx0)
  87
  88 L(bx1): andi.   r0, n, 2
  89         li      s0, 0
  90         bne     L(lo3)
  91         b       L(lo1)
  92
  93 L(bx0): andi.   r0, n, 2
  94         li      s1, 0
  95         bne     L(lo2)
  96
  97         ALIGN(32)
  98 L(top): addi    rp, rp, 32
  99         ld      v0, 0(vp)
 100         addi    vp, vp, 32
 101         rldimi  s1, v0, LSH, 0
 102         ld      u0, 0(up)
 103         addi    up, up, 32
 104         srdi    s0, v0, RSH
 105         ADDSUBE(s1, s1, u0)
 106         std     s1, -32(rp)
 107 L(lo3): ld      v1, -24(vp)
 108         rldimi  s0, v1, LSH, 0
 109         ld      u1, -24(up)
 110         srdi    s1, v1, RSH
 111         ADDSUBE(s0, s0, u1)
 112         std     s0, -24(rp)
 113 L(lo2): ld      v0, -16(vp)
 114         rldimi  s1, v0, LSH, 0
 115         ld      u0, -16(up)
 116         srdi    s0, v0, RSH
 117         ADDSUBE(s1, s1, u0)
 118         std     s1, -16(rp)
 119 L(lo1): ld      v1, -8(vp)
 120         rldimi  s0, v1, LSH, 0
 121         ld      u1, -8(up)
 122         srdi    s1, v1, RSH
 123         ADDSUBE(s0, s0, u1)
 124         std     s0, -8(rp)
 125         bdnz    L(top)          C decrement CTR and loop back
 126
 127         RETVAL( s1)
 128         blr
 129 EPILOGUE()