source/libs/gmp/gmp-src/mpn/powerpc64/mode64/bdiv_dbm1c.asm

   1 dnl  PPC64 mpn_bdiv_dbm1c.
   2
   3 dnl  Copyright 2008, 2010 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/limb
  34 C POWER3/PPC630       6-18
  35 C POWER4/PPC970       8.25
  36 C POWER5              8.5  fluctuating as function of n % 3
  37 C POWER6             15
  38 C POWER7              4.75
  39
  40 C TODO
  41 C  * Nothing to do...
  42
  43 C INPUT PARAMETERS
  44 define(`rp', `r3')
  45 define(`up', `r4')
  46 define(`n',  `r5')
  47 define(`bd', `r6')
  48 define(`cy', `r7')
  49
  50 ASM_START()
  51 PROLOGUE(mpn_bdiv_dbm1c)
  52         ld      r0, 0(r4)
  53
  54         rldicl. r12, r5, 0,62
  55         cmpldi  cr6, r12, 2
  56         cmpldi  cr7, r5, 4
  57         addi    r5, r5, 1
  58         srwi    r5, r5, 2
  59         mtctr   r5
  60         beq     cr0, L(b00)
  61         blt     cr6, L(b01)
  62         beq     cr6, L(b10)
  63
  64         ALIGN(16)
  65 L(b11): mulld   r5, r0, r6
  66         mulhdu  r12, r0, r6
  67         ld      r0, 8(r4)
  68         addi    r4, r4, -24
  69         addi    r3, r3, -24
  70         b       L(3)
  71
  72         ALIGN(16)
  73 L(b00): mulld   r9, r0, r6
  74         mulhdu  r8, r0, r6
  75         addi    r4, r4, -16
  76         addi    r3, r3, -16
  77         b       L(0)
  78
  79         ALIGN(16)
  80 L(b01): mulld   r5, r0, r6
  81         mulhdu  r12, r0, r6
  82         addi    r3, r3, -8
  83         ble     cr7, L(e1)
  84         ld      r0, 8(r4)
  85         addi    r4, r4, -8
  86         b       L(1)
  87
  88         ALIGN(16)
  89 L(b10): mulld   r9, r0, r6
  90         mulhdu  r8, r0, r6
  91         ble     cr7, L(e2)
  92
  93         ALIGN(16)
  94 L(top): subfc   r11, r9, r7
  95         ld      r10, 8(r4)
  96         ld      r0, 16(r4)
  97         subfe   r7, r8, r11
  98         std     r11, 0(r3)
  99         mulld   r5, r10, r6
 100         mulhdu  r12, r10, r6
 101 L(1):   mulld   r9, r0, r6
 102         mulhdu  r8, r0, r6
 103         subfc   r11, r5, r7
 104         subfe   r7, r12, r11
 105         std     r11, 8(r3)
 106 L(0):   subfc   r11, r9, r7
 107         ld      r10, 24(r4)
 108         ld      r0, 32(r4)
 109         subfe   r7, r8, r11
 110         std     r11, 16(r3)
 111         mulld   r5, r10, r6
 112         mulhdu  r12, r10, r6
 113 L(3):   mulld   r9, r0, r6
 114         mulhdu  r8, r0, r6
 115         subfc   r11, r5, r7
 116         subfe   r7, r12, r11
 117         std     r11, 24(r3)
 118         addi    r4, r4, 32
 119         addi    r3, r3, 32
 120         bdnz    L(top)
 121
 122 L(e2):  ld      r10, 8(r4)
 123         mulld   r5, r10, r6
 124         mulhdu  r12, r10, r6
 125         subfc   r11, r9, r7
 126         subfe   r7, r8, r11
 127         std     r11, 0(r3)
 128 L(e1):  subfc   r11, r5, r7
 129         std     r11, 8(r3)
 130         subfe   r3, r12, r11
 131         blr
 132 EPILOGUE()