source/libs/gmp/gmp-src/mpn/arm/cnd_aors_n.asm

   1 dnl  ARM mpn_cnd_add_n, mpn_cnd_sub_n
   2
   3 dnl  Copyright 2012, 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 StrongARM      ?
  35 C XScale         ?
  36 C Cortex-A7      ?
  37 C Cortex-A8      ?
  38 C Cortex-A9      3
  39 C Cortex-A15     2.5
  40
  41 define(`cnd',   `r0')
  42 define(`rp',    `r1')
  43 define(`up',    `r2')
  44 define(`vp',    `r3')
  45
  46 define(`n',     `r12')
  47
  48
  49 ifdef(`OPERATION_cnd_add_n', `
  50         define(`ADDSUB',      adds)
  51         define(`ADDSUBC',      adcs)
  52         define(`INITCY',      `cmn      r0, #0')
  53         define(`RETVAL',      `adc      r0, n, #0')
  54         define(func,          mpn_cnd_add_n)')
  55 ifdef(`OPERATION_cnd_sub_n', `
  56         define(`ADDSUB',      subs)
  57         define(`ADDSUBC',      sbcs)
  58         define(`INITCY',      `cmp      r0, #0')
  59         define(`RETVAL',      `adc      r0, n, #0
  60                               rsb       r0, r0, #1')
  61         define(func,          mpn_cnd_sub_n)')
  62
  63 MULFUNC_PROLOGUE(mpn_cnd_add_n mpn_cnd_sub_n)
  64
  65 ASM_START()
  66 PROLOGUE(func)
  67         push    {r4-r11}
  68         ldr     n, [sp, #32]
  69
  70         cmp     cnd, #1
  71         sbc     cnd, cnd, cnd           C conditionally set to 0xffffffff
  72
  73         INITCY                          C really only needed for n = 0 (mod 4)
  74
  75         ands    r4, n, #3
  76         beq     L(top)
  77         cmp     r4, #2
  78         bcc     L(b1)
  79         beq     L(b2)
  80
  81 L(b3):  ldm     vp!, {r4,r5,r6}
  82         ldm     up!, {r8,r9,r10}
  83         bic     r4, r4, cnd
  84         bic     r5, r5, cnd
  85         bic     r6, r6, cnd
  86         ADDSUB  r8, r8, r4
  87         ADDSUBC r9, r9, r5
  88         ADDSUBC r10, r10, r6
  89         stm     rp!, {r8,r9,r10}
  90         sub     n, n, #3
  91         teq     n, #0
  92         bne     L(top)
  93         b       L(end)
  94
  95 L(b2):  ldm     vp!, {r4,r5}
  96         ldm     up!, {r8,r9}
  97         bic     r4, r4, cnd
  98         bic     r5, r5, cnd
  99         ADDSUB  r8, r8, r4
 100         ADDSUBC r9, r9, r5
 101         stm     rp!, {r8,r9}
 102         sub     n, n, #2
 103         teq     n, #0
 104         bne     L(top)
 105         b       L(end)
 106
 107 L(b1):  ldr     r4, [vp], #4
 108         ldr     r8, [up], #4
 109         bic     r4, r4, cnd
 110         ADDSUB  r8, r8, r4
 111         str     r8, [rp], #4
 112         sub     n, n, #1
 113         teq     n, #0
 114         beq     L(end)
 115
 116 L(top): ldm     vp!, {r4,r5,r6,r7}
 117         ldm     up!, {r8,r9,r10,r11}
 118         bic     r4, r4, cnd
 119         bic     r5, r5, cnd
 120         bic     r6, r6, cnd
 121         bic     r7, r7, cnd
 122         ADDSUBC r8, r8, r4
 123         ADDSUBC r9, r9, r5
 124         ADDSUBC r10, r10, r6
 125         ADDSUBC r11, r11, r7
 126         sub     n, n, #4
 127         stm     rp!, {r8,r9,r10,r11}
 128         teq     n, #0
 129         bne     L(top)
 130
 131 L(end): RETVAL
 132         pop     {r4-r11}
 133         bx      r14
 134 EPILOGUE()