| 1 |
dnl AMD64 mpn_addlshC_n, mpn_sublshC_n -- rp[] = up[] +- (vp[] << C), where |
| 2 |
dnl C is 1, 2, 3. Optimized for Pentium 4. |
| 3 |
|
| 4 |
dnl Contributed to the GNU project by Torbjorn Granlund. |
| 5 |
|
| 6 |
dnl Copyright 2008, 2010-2012 Free Software Foundation, Inc. |
| 7 |
|
| 8 |
dnl This file is part of the GNU MP Library. |
| 9 |
dnl |
| 10 |
dnl The GNU MP Library is free software; you can redistribute it and/or modify |
| 11 |
dnl it under the terms of either: |
| 12 |
dnl |
| 13 |
dnl * the GNU Lesser General Public License as published by the Free |
| 14 |
dnl Software Foundation; either version 3 of the License, or (at your |
| 15 |
dnl option) any later version. |
| 16 |
dnl |
| 17 |
dnl or |
| 18 |
dnl |
| 19 |
dnl * the GNU General Public License as published by the Free Software |
| 20 |
dnl Foundation; either version 2 of the License, or (at your option) any |
| 21 |
dnl later version. |
| 22 |
dnl |
| 23 |
dnl or both in parallel, as here. |
| 24 |
dnl |
| 25 |
dnl The GNU MP Library is distributed in the hope that it will be useful, but |
| 26 |
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
| 27 |
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 28 |
dnl for more details. |
| 29 |
dnl |
| 30 |
dnl You should have received copies of the GNU General Public License and the |
| 31 |
dnl GNU Lesser General Public License along with the GNU MP Library. If not, |
| 32 |
dnl see https://www.gnu.org/licenses/. |
| 33 |
|
| 34 |
C cycles/limb |
| 35 |
C AMD K8,K9 3.8 |
| 36 |
C AMD K10 3.8 |
| 37 |
C Intel P4 5.8 |
| 38 |
C Intel core2 4.75 |
| 39 |
C Intel corei 4.75 |
| 40 |
C Intel atom ? |
| 41 |
C VIA nano 4.75 |
| 42 |
|
| 43 |
|
| 44 |
C INPUT PARAMETERS |
| 45 |
define(`rp',`%rdi') |
| 46 |
define(`up',`%rsi') |
| 47 |
define(`vp',`%rdx') |
| 48 |
define(`n', `%rcx') |
| 49 |
|
| 50 |
define(M, eval(m4_lshift(1,LSH))) |
| 51 |
|
| 52 |
ABI_SUPPORT(DOS64) |
| 53 |
ABI_SUPPORT(STD64) |
| 54 |
|
| 55 |
ASM_START() |
| 56 |
TEXT |
| 57 |
ALIGN(16) |
| 58 |
PROLOGUE(func) |
| 59 |
FUNC_ENTRY(4) |
| 60 |
push %rbx |
| 61 |
push %r12 |
| 62 |
push %rbp |
| 63 |
|
| 64 |
mov (vp), %r9 |
| 65 |
shl $LSH, %r9 |
| 66 |
mov 4(vp), R32(%rbp) |
| 67 |
|
| 68 |
xor R32(%rbx), R32(%rbx) |
| 69 |
|
| 70 |
mov R32(n), R32(%rax) |
| 71 |
and $3, R32(%rax) |
| 72 |
jne L(n00) C n = 0, 4, 8, ... |
| 73 |
|
| 74 |
mov (up), %r8 |
| 75 |
mov 8(up), %r10 |
| 76 |
shr $RSH, R32(%rbp) |
| 77 |
ADDSUB %r9, %r8 |
| 78 |
mov 8(vp), %r9 |
| 79 |
lea (%rbp,%r9,M), %r9 |
| 80 |
setc R8(%rax) |
| 81 |
mov 12(vp), R32(%rbp) |
| 82 |
lea -16(rp), rp |
| 83 |
jmp L(L00) |
| 84 |
|
| 85 |
L(n00): cmp $2, R32(%rax) |
| 86 |
jnc L(n01) C n = 1, 5, 9, ... |
| 87 |
mov (up), %r11 |
| 88 |
lea -8(rp), rp |
| 89 |
shr $RSH, R32(%rbp) |
| 90 |
ADDSUB %r9, %r11 |
| 91 |
setc R8(%rbx) |
| 92 |
dec n |
| 93 |
jz L(1) C jump for n = 1 |
| 94 |
mov 8(up), %r8 |
| 95 |
mov 8(vp), %r9 |
| 96 |
lea (%rbp,%r9,M), %r9 |
| 97 |
mov 12(vp), R32(%rbp) |
| 98 |
lea 8(up), up |
| 99 |
lea 8(vp), vp |
| 100 |
jmp L(L01) |
| 101 |
|
| 102 |
L(n01): jne L(n10) C n = 2, 6, 10, ... |
| 103 |
mov (up), %r12 |
| 104 |
mov 8(up), %r11 |
| 105 |
shr $RSH, R32(%rbp) |
| 106 |
ADDSUB %r9, %r12 |
| 107 |
mov 8(vp), %r9 |
| 108 |
lea (%rbp,%r9,M), %r9 |
| 109 |
setc R8(%rax) |
| 110 |
mov 12(vp), R32(%rbp) |
| 111 |
lea 16(up), up |
| 112 |
lea 16(vp), vp |
| 113 |
jmp L(L10) |
| 114 |
|
| 115 |
L(n10): mov (up), %r10 |
| 116 |
mov 8(up), %r12 |
| 117 |
shr $RSH, R32(%rbp) |
| 118 |
ADDSUB %r9, %r10 |
| 119 |
mov 8(vp), %r9 |
| 120 |
lea (%rbp,%r9,M), %r9 |
| 121 |
setc R8(%rbx) |
| 122 |
mov 12(vp), R32(%rbp) |
| 123 |
lea -24(rp), rp |
| 124 |
lea -8(up), up |
| 125 |
lea -8(vp), vp |
| 126 |
jmp L(L11) |
| 127 |
|
| 128 |
L(c0): mov $1, R8(%rbx) |
| 129 |
jmp L(rc0) |
| 130 |
L(c1): mov $1, R8(%rax) |
| 131 |
jmp L(rc1) |
| 132 |
L(c2): mov $1, R8(%rbx) |
| 133 |
jmp L(rc2) |
| 134 |
|
| 135 |
ALIGN(16) |
| 136 |
L(top): mov (up), %r8 C not on critical path |
| 137 |
shr $RSH, R32(%rbp) |
| 138 |
ADDSUB %r9, %r11 C not on critical path |
| 139 |
mov (vp), %r9 |
| 140 |
lea (%rbp,%r9,M), %r9 |
| 141 |
setc R8(%rbx) C save carry out |
| 142 |
mov 4(vp), R32(%rbp) |
| 143 |
mov %r12, (rp) |
| 144 |
ADDSUB %rax, %r11 C apply previous carry out |
| 145 |
jc L(c0) C jump if ripple |
| 146 |
L(rc0): |
| 147 |
L(L01): mov 8(up), %r10 |
| 148 |
shr $RSH, R32(%rbp) |
| 149 |
ADDSUB %r9, %r8 |
| 150 |
mov 8(vp), %r9 |
| 151 |
lea (%rbp,%r9,M), %r9 |
| 152 |
setc R8(%rax) |
| 153 |
mov 12(vp), R32(%rbp) |
| 154 |
mov %r11, 8(rp) |
| 155 |
ADDSUB %rbx, %r8 |
| 156 |
jc L(c1) |
| 157 |
L(rc1): |
| 158 |
L(L00): mov 16(up), %r12 |
| 159 |
shr $RSH, R32(%rbp) |
| 160 |
ADDSUB %r9, %r10 |
| 161 |
mov 16(vp), %r9 |
| 162 |
lea (%rbp,%r9,M), %r9 |
| 163 |
setc R8(%rbx) |
| 164 |
mov 20(vp), R32(%rbp) |
| 165 |
mov %r8, 16(rp) |
| 166 |
ADDSUB %rax, %r10 |
| 167 |
jc L(c2) |
| 168 |
L(rc2): |
| 169 |
L(L11): mov 24(up), %r11 |
| 170 |
shr $RSH, R32(%rbp) |
| 171 |
ADDSUB %r9, %r12 |
| 172 |
mov 24(vp), %r9 |
| 173 |
lea (%rbp,%r9,M), %r9 |
| 174 |
lea 32(up), up |
| 175 |
lea 32(vp), vp |
| 176 |
setc R8(%rax) |
| 177 |
mov -4(vp), R32(%rbp) |
| 178 |
mov %r10, 24(rp) |
| 179 |
ADDSUB %rbx, %r12 |
| 180 |
jc L(c3) |
| 181 |
L(rc3): lea 32(rp), rp |
| 182 |
L(L10): sub $4, n |
| 183 |
ja L(top) |
| 184 |
|
| 185 |
L(end): |
| 186 |
shr $RSH, R32(%rbp) |
| 187 |
ADDSUB %r9, %r11 |
| 188 |
setc R8(%rbx) |
| 189 |
mov %r12, (rp) |
| 190 |
ADDSUB %rax, %r11 |
| 191 |
jnc L(1) |
| 192 |
mov $1, R8(%rbx) |
| 193 |
L(1): mov %r11, 8(rp) |
| 194 |
lea (%rbx,%rbp), R32(%rax) |
| 195 |
pop %rbp |
| 196 |
pop %r12 |
| 197 |
pop %rbx |
| 198 |
FUNC_EXIT() |
| 199 |
ret |
| 200 |
L(c3): mov $1, R8(%rax) |
| 201 |
jmp L(rc3) |
| 202 |
EPILOGUE() |
| 203 |
ASM_END() |