mul_1.asm revision 1.1.1.1.4.2
1 dnl AMD64 mpn_mul_1 optimised for AMD Bulldozer. 2 3 dnl Copyright 2003, 2004, 2005, 2007, 2008, 2011, 2012 Free Software 4 dnl Foundation, Inc. 5 6 dnl This file is part of the GNU MP Library. 7 8 dnl The GNU MP Library is free software; you can redistribute it and/or modify 9 dnl it under the terms of the GNU Lesser General Public License as published 10 dnl by the Free Software Foundation; either version 3 of the License, or (at 11 dnl your option) any later version. 12 13 dnl The GNU MP Library is distributed in the hope that it will be useful, but 14 dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 15 dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public 16 dnl License for more details. 17 18 dnl You should have received a copy of the GNU Lesser General Public License 19 dnl along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. 20 21 include(`../config.m4') 22 23 C cycles/limb 24 C AMD K8,K9 25 C AMD K10 26 C AMD bd1 4 27 C AMD bobcat 28 C Intel P4 29 C Intel core2 30 C Intel NHM 31 C Intel SBR 32 C Intel atom 33 C VIA nano 34 35 C The loop of this code is the result of running a code generation and 36 C optimisation tool suite written by David Harvey and Torbjorn Granlund. 37 38 C TODO 39 C * Move loop code into feed-in blocks, to save insn for zeroing regs. 40 41 define(`rp', `%rdi') C rcx 42 define(`up', `%rsi') C rdx 43 define(`n_param', `%rdx') C r8 44 define(`v0', `%rcx') C r9 45 46 define(`n', `%rbx') 47 48 ABI_SUPPORT(DOS64) 49 ABI_SUPPORT(STD64) 50 51 IFDOS(` define(`up', ``%rsi'') ') dnl 52 IFDOS(` define(`rp', ``%rcx'') ') dnl 53 IFDOS(` define(`v0', ``%r9'') ') dnl 54 IFDOS(` define(`r9', ``rdi'') ') dnl 55 IFDOS(` define(`n', ``%r8'') ') dnl 56 IFDOS(` define(`r8', ``rbx'') ') dnl 57 58 ASM_START() 59 TEXT 60 ALIGN(16) 61 PROLOGUE(mpn_mul_1c) 62 IFDOS(``push %rsi '') 63 IFDOS(``push %rdi '') 64 IFDOS(``mov %rdx, %rsi '') 65 66 mov (up), %rax C read first u limb early 67 push %rbx 68 IFSTD(` mov n_param, %r11 ') C move away n from rdx, mul uses it 69 IFDOS(` mov n, %r11 ') 70 mul v0 71 72 IFSTD(` add %r8, %rax ') 73 IFDOS(` add 64(%rsp), %rax ') C 40 + 3*8 (3 push insns) 74 adc $0, %rdx 75 jmp L(common) 76 77 EPILOGUE() 78 79 ALIGN(16) 80 PROLOGUE(mpn_mul_1) 81 IFDOS(``push %rsi '') 82 IFDOS(``push %rdi '') 83 IFDOS(``mov %rdx, %rsi '') 84 85 mov (up), %rax C read first u limb early 86 push %rbx 87 IFSTD(` mov n_param, %r11 ') C move away n from rdx, mul uses it 88 IFDOS(` mov n, %r11 ') 89 mul v0 90 91 L(common): 92 IFSTD(` mov %r11, n ') 93 94 and $3, R32(%r11) 95 lea -16(rp,n,8), rp 96 jz L(b0) 97 cmp $2, R32(%r11) 98 jb L(b1) 99 jz L(b2) 100 101 L(b3): mov %rax, %r10 102 mov %rdx, %r11 103 mov 8(up), %rax 104 mul v0 105 lea (up,n,8), up 106 not n 107 jmp L(L3) 108 109 L(b0): mov %rax, %r9 110 mov %rdx, %r10 111 mov 8(up), %rax 112 lea (up,n,8), up 113 neg n 114 jmp L(L0) 115 116 L(b1): mov %rax, %r8 117 cmp $1, n 118 jz L(n1) 119 mov %rdx, %r9 120 lea (up,n,8), up 121 neg n 122 mov %r8, 16(rp,n,8) 123 inc n 124 jmp L(L1) 125 126 L(b2): mov %rax, %r11 127 mov %rdx, %r8 128 mov 8(up), %rax 129 lea (up,n,8), up 130 neg n 131 add $2, n 132 jns L(end) 133 134 ALIGN(16) 135 L(top): mul v0 136 mov %rdx, %r9 137 add %rax, %r8 138 adc $0, %r9 139 mov %r8, 8(rp,n,8) 140 mov %r11, (rp,n,8) 141 L(L1): mov (up,n,8), %rax 142 mul v0 143 add %rax, %r9 144 mov %rdx, %r10 145 mov 8(up,n,8), %rax 146 adc $0, %r10 147 L(L0): mul v0 148 add %rax, %r10 149 mov %rdx, %r11 150 mov 16(up,n,8), %rax 151 adc $0, %r11 152 mul v0 153 mov %r9, 16(rp,n,8) 154 L(L3): add %rax, %r11 155 mov %r10, 24(rp,n,8) 156 mov %rdx, %r8 157 adc $0, %r8 158 add $4, n 159 mov -8(up,n,8), %rax 160 js L(top) 161 162 L(end): mul v0 163 add %rax, %r8 164 adc $0, %rdx 165 mov %r11, (rp) 166 L(n1): mov %r8, 8(rp) 167 mov %rdx, %rax 168 169 pop %rbx 170 IFDOS(``pop %rdi '') 171 IFDOS(``pop %rsi '') 172 ret 173 EPILOGUE() 174 ASM_END() 175
Indexes created Sat Feb 28 05:31:39 UTC 2026