1dnl Power9 mpn_mul_1. 2 3dnl Copyright 2017, 2018 Free Software Foundation, Inc. 4 5dnl This file is part of the GNU MP Library. 6dnl 7dnl The GNU MP Library is free software; you can redistribute it and/or modify 8dnl it under the terms of either: 9dnl 10dnl * the GNU Lesser General Public License as published by the Free 11dnl Software Foundation; either version 3 of the License, or (at your 12dnl option) any later version. 13dnl 14dnl or 15dnl 16dnl * the GNU General Public License as published by the Free Software 17dnl Foundation; either version 2 of the License, or (at your option) any 18dnl later version. 19dnl 20dnl or both in parallel, as here. 21dnl 22dnl The GNU MP Library is distributed in the hope that it will be useful, but 23dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 24dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 25dnl for more details. 26dnl 27dnl You should have received copies of the GNU General Public License and the 28dnl GNU Lesser General Public License along with the GNU MP Library. If not, 29dnl see https://www.gnu.org/licenses/. 30 31include(`../config.m4') 32 33C cycles/limb 34C POWER3/PPC630 ? 35C POWER4/PPC970 ? 36C POWER5 ? 37C POWER6 ? 38C POWER7 ? 39C POWER8 ? 40C POWER9 2.47 41 42C TODO 43C * Schedule for Power9 pipeline. 44C * Unroll 4x if that proves beneficial. 45C * This is marginally faster (but much smaller) than ../mul_1.asm. 46 47C INPUT PARAMETERS 48define(`rp', `r3') 49define(`up', `r4') 50define(`n', `r5') 51define(`v0', `r6') 52 53ASM_START() 54PROLOGUE(mpn_mul_1c) 55 b L(ent) 56EPILOGUE() 57PROLOGUE(mpn_mul_1) 58 li r7, 0 59L(ent): ld r11, 0(up) 60 cmpdi cr6, n, 2 61 addi r0, n, -1 C FIXME: postpone 62 srdi r0, r0, 1 C FIXME: postpone 63 mtctr r0 C FIXME: postpone 64 rldicl. r12, n, 0,63 C r0 = n & 3, set cr0 65 bne cr0, L(b1) 66 67L(b0): ld r0, 8(up) 68 maddld( r9, r11, v0, r7) 69 maddhdu(r7, r11, v0, r7) 70 ble cr6, L(2) 71 ld r12, 16(up) 72 mulld r8, r0, v0 73 mulhdu r5, r0, v0 74 addic up, up, 16 75 addi rp, rp, -8 76 b L(mid) 77 78L(b1): ld r0, 0(up) 79 ble cr6, L(1) 80 ld r12, 8(up) 81 maddld( r8, r11, v0, r7) 82 maddhdu(r5, r11, v0, r7) 83 ld r0, 16(up) 84 mulld r9, r12, v0 85 mulhdu r7, r12, v0 86 addic up, up, 24 87 bdz L(end) 88 89 ALIGN(16) 90L(top): ld r12, 0(up) 91 std r8, 0(rp) 92 adde r9, r5, r9 93 mulld r8, r0, v0 94 mulhdu r5, r0, v0 95L(mid): ld r0, 8(up) 96 std r9, 8(rp) 97 adde r8, r7, r8 98 mulld r9, r12, v0 99 mulhdu r7, r12, v0 100 addi rp, rp, 16 101 addi up, up, 16 102 bdnz L(top) 103 104L(end): std r8, 0(rp) 105 mulld r8, r0, v0 106 adde r9, r5, r9 107 mulhdu r5, r0, v0 108 std r9, 8(rp) 109 adde r8, r7, r8 110 std r8, 16(rp) 111 addze r3, r5 112 blr 113 114L(2): mulld r8, r0, v0 115 mulhdu r5, r0, v0 116 std r9, 0(rp) 117 addc r8, r7, r8 118 std r8, 8(rp) 119 addze r3, r5 120 blr 121 122L(1): maddld( r8, r0, v0, r7) 123 std r8, 0(rp) 124 maddhdu(r3, r0, v0, r7) 125 blr 126EPILOGUE() 127