1 dnl Power9 mpn_addmul_1. 2 3 dnl Copyright 2017, 2018 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 - 39 C POWER8 - 40 C POWER9 2.5 41 42 C TODO 43 C * Schedule for Power9 pipeline. 44 C * Unroll 4x if that proves beneficial. 45 C * This is marginally faster (but much smaller) than ../aorsmul_1.asm. 46 47 C INPUT PARAMETERS 48 define(`rp', `r3') 49 define(`up', `r4') 50 define(`n', `r5') 51 define(`v0', `r6') 52 53 ASM_START() 54 PROLOGUE(mpn_addmul_1) 55 cmpdi cr6, n, 2 56 addi r0, n, -1 C FIXME: postpone 57 srdi r0, r0, 1 C FIXME: postpone 58 mtctr r0 C FIXME: postpone 59 rldicl. r0, n, 0,63 C r0 = n & 3, set cr0 60 bne cr0, L(b1) 61 62 L(b0): ld r10, 0(rp) 63 ld r12, 0(up) 64 ld r11, 8(rp) 65 ld r0, 8(up) 66 maddld( r9, r12, v0, r10) 67 maddhdu(r7, r12, v0, r10) 68 ble cr6, L(2) 69 ld r10, 16(rp) 70 ld r12, 16(up) 71 maddld( r8, r0, v0, r11) 72 maddhdu(r5, r0, v0, r11) 73 addic up, up, 16 74 addi rp, rp, -8 75 b L(mid) 76 77 L(b1): ld r11, 0(rp) 78 ld r0, 0(up) 79 ble cr6, L(1) 80 ld r10, 8(rp) 81 ld r12, 8(up) 82 maddld( r8, r0, v0, r11) 83 maddhdu(r5, r0, v0, r11) 84 ld r11, 16(rp) 85 ld r0, 16(up) 86 maddld( r9, r12, v0, r10) 87 maddhdu(r7, r12, v0, r10) 88 addic up, up, 24 89 bdz L(end) 90 91 ALIGN(16) 92 L(top): ld r10, 24(rp) 93 ld r12, 0(up) 94 std r8, 0(rp) 95 adde r9, r5, r9 96 maddld( r8, r0, v0, r11) C W:0,2,4 97 maddhdu(r5, r0, v0, r11) C W:1,3,5 98 L(mid): ld r11, 32(rp) 99 ld r0, 8(up) 100 std r9, 8(rp) 101 adde r8, r7, r8 102 maddld( r9, r12, v0, r10) C W:1,3,5 103 maddhdu(r7, r12, v0, r10) C W:2,4,6 104 addi rp, rp, 16 105 addi up, up, 16 106 bdnz L(top) 107 108 L(end): std r8, 0(rp) 109 maddld( r8, r0, v0, r11) 110 adde r9, r5, r9 111 maddhdu(r5, r0, v0, r11) 112 std r9, 8(rp) 113 adde r8, r7, r8 114 std r8, 16(rp) 115 addze r3, r5 116 blr 117 118 L(2): maddld( r8, r0, v0, r11) 119 maddhdu(r5, r0, v0, r11) 120 std r9, 0(rp) 121 addc r8, r7, r8 122 std r8, 8(rp) 123 addze r3, r5 124 blr 125 126 L(1): maddld( r8, r0, v0, r11) 127 std r8, 0(rp) 128 maddhdu(r3, r0, v0, r11) 129 blr 130 EPILOGUE() 131