source/libs/gmp/gmp-src/mpn/powerpc64/mode64/gcd_1.asm

   1 dnl  PowerPC-64 mpn_gcd_1.
   2
   3 dnl  Copyright 2000-2002, 2005, 2009, 2011-2013 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 include(`../config.m4')
  32
  33 C                   cycles/bit (approx)
  34 C POWER3/PPC630          ?
  35 C POWER4/PPC970          8.5
  36 C POWER5                 ?
  37 C POWER6                10.1
  38 C POWER7                 9.4
  39 C Numbers measured with: speed -CD -s16-64 -t48 mpn_gcd_1
  40
  41 C INPUT PARAMETERS
  42 define(`up',    `r3')
  43 define(`n',     `r4')
  44 define(`v0',    `r5')
  45
  46 ifdef(`BMOD_1_TO_MOD_1_THRESHOLD',,
  47   `define(`BMOD_1_TO_MOD_1_THRESHOLD',30)')
  48
  49 EXTERN_FUNC(mpn_mod_1)
  50 EXTERN_FUNC(mpn_modexact_1c_odd)
  51
  52 ASM_START()
  53 PROLOGUE(mpn_gcd_1,toc)
  54         mflr    r0
  55         std     r30, -16(r1)
  56         std     r31, -8(r1)
  57         std     r0, 16(r1)
  58         stdu    r1, -128(r1)
  59
  60         ld      r7, 0(up)               C U low limb
  61         or      r0, r5, r7              C x | y
  62
  63         neg     r6, r0
  64         and     r6, r6, r0
  65         cntlzd  r31, r6                 C common twos
  66         subfic  r31, r31, 63
  67
  68         neg     r6, r5
  69         and     r6, r6, r5
  70         cntlzd  r8, r6
  71         subfic  r8, r8, 63
  72         srd     r5, r5, r8
  73         mr      r30, r5                 C v0 saved
  74
  75         cmpdi   r4, BMOD_1_TO_MOD_1_THRESHOLD
  76         blt     L(bmod)
  77         CALL(   mpn_mod_1)
  78         b       L(reduced)
  79 L(bmod):
  80         li      r6, 0
  81         CALL(   mpn_modexact_1c_odd)
  82 L(reduced):
  83
  84 define(`mask', `r0')dnl
  85 define(`a1',   `r4')dnl
  86 define(`a2',   `r5')dnl
  87 define(`d1',   `r6')dnl
  88 define(`d2',   `r7')dnl
  89 define(`cnt',  `r9')dnl
  90
  91         neg.    r6, r3
  92         and     r6, r6, r3
  93         cntlzd  cnt, r6
  94         subfic  cnt, cnt, 63
  95         li      r12, 63
  96         bne     L(mid)
  97         b       L(end)
  98
  99         ALIGN(16)
 100 L(top):
 101         and     a1, r10, mask           C d - a
 102         andc    a2, r11,  mask          C a - d
 103         and     d1, r3, mask            C a
 104         andc    d2, r30, mask           C d
 105         or      r3, a1, a2              C new a
 106         subf    cnt, cnt, r12
 107         or      r30, d1, d2             C new d
 108 L(mid): srd     r3, r3, cnt
 109         sub.    r10, r30, r3            C r10 = d - a
 110         subc    r11, r3, r30            C r11 = a - d
 111         neg     r8, r10
 112         and     r8, r8, r10
 113         subfe   mask, mask, mask
 114         cntlzd  cnt, r8
 115         bne     L(top)
 116
 117 L(end): sld     r3, r30, r31
 118
 119         addi    r1, r1, 128
 120         ld      r0, 16(r1)
 121         ld      r30, -16(r1)
 122         ld      r31, -8(r1)
 123         mtlr    r0
 124         blr
 125 EPILOGUE()