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

   1 dnl  PowerPC-64 mpn_add_n, mpn_sub_n optimised for POWER7.
   2
   3 dnl  Copyright 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/limb
  34 C POWER3/PPC630          ?
  35 C POWER4/PPC970          ?
  36 C POWER5                 ?
  37 C POWER6                 ?
  38 C POWER7                 2.18
  39
  40 C This is a tad bit slower than the cnd_aors_n.asm code, which is of course an
  41 C anomaly.
  42
  43 ifdef(`OPERATION_add_n',`
  44   define(ADDSUBC,       adde)
  45   define(ADDSUB,        addc)
  46   define(func,          mpn_add_n)
  47   define(func_nc,       mpn_add_nc)
  48   define(GENRVAL,       `addi   r3, r3, 1')
  49   define(SETCBR,        `addic  r0, $1, -1')
  50   define(CLRCB,         `addic  r0, r0, 0')
  51 ')
  52 ifdef(`OPERATION_sub_n',`
  53   define(ADDSUBC,       subfe)
  54   define(ADDSUB,        subfc)
  55   define(func,          mpn_sub_n)
  56   define(func_nc,       mpn_sub_nc)
  57   define(GENRVAL,       `neg    r3, r3')
  58   define(SETCBR,        `subfic r0, $1, 0')
  59   define(CLRCB,         `addic  r0, r1, -1')
  60 ')
  61
  62 MULFUNC_PROLOGUE(mpn_add_n mpn_add_nc mpn_sub_n mpn_sub_nc)
  63
  64 C INPUT PARAMETERS
  65 define(`rp',    `r3')
  66 define(`up',    `r4')
  67 define(`vp',    `r5')
  68 define(`n',     `r6')
  69
  70 ASM_START()
  71 PROLOGUE(func_nc)
  72         SETCBR(r7)
  73         b       L(ent)
  74 EPILOGUE()
  75
  76 PROLOGUE(func)
  77         CLRCB
  78 L(ent):
  79         andi.   r7, n, 1
  80         beq     L(bx0)
  81
  82 L(bx1): ld      r7, 0(up)
  83         ld      r9, 0(vp)
  84         ADDSUBC r11, r9, r7
  85         std     r11, 0(rp)
  86         cmpldi  cr6, n, 1
  87         beq     cr6, L(end)
  88         addi    up, up, 8
  89         addi    vp, vp, 8
  90         addi    rp, rp, 8
  91
  92 L(bx0): addi    r0, n, 2        C compute branch...
  93         srdi    r0, r0, 2       C ...count
  94         mtctr   r0
  95
  96         andi.   r7, n, 2
  97         bne     L(mid)
  98
  99         addi    up, up, 16
 100         addi    vp, vp, 16
 101         addi    rp, rp, 16
 102
 103         ALIGN(32)
 104 L(top): ld      r6, -16(up)
 105         ld      r7, -8(up)
 106         ld      r8, -16(vp)
 107         ld      r9, -8(vp)
 108         ADDSUBC r10, r8, r6
 109         ADDSUBC r11, r9, r7
 110         std     r10, -16(rp)
 111         std     r11, -8(rp)
 112 L(mid): ld      r6, 0(up)
 113         ld      r7, 8(up)
 114         ld      r8, 0(vp)
 115         ld      r9, 8(vp)
 116         ADDSUBC r10, r8, r6
 117         ADDSUBC r11, r9, r7
 118         std     r10, 0(rp)
 119         std     r11, 8(rp)
 120         addi    up, up, 32
 121         addi    vp, vp, 32
 122         addi    rp, rp, 32
 123         bdnz    L(top)
 124
 125 L(end): subfe   r3, r0, r0      C -cy
 126         GENRVAL
 127         blr
 128 EPILOGUE()