| Austin Schuh | dace2a6 | 2020-08-18 10:56:48 -0700 | [diff] [blame] | 1 | dnl PowerPC-32 mpn_divexact_by3 -- mpn by 3 exact division |
| 2 | |
| 3 | dnl Copyright 2002, 2003, 2005, 2006 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 603e: ? |
| 35 | C 604e: 5 |
| 36 | C 75x (G3): ? |
| 37 | C 7400,7410 (G4): 8 |
| 38 | C 744x,745x (G4+): 6 |
| 39 | C power4/ppc970: 12 |
| 40 | C power5: ? |
| 41 | |
| 42 | C void mpn_divexact_by3 (mp_ptr dst, mp_srcptr src, mp_size_t size); |
| 43 | C |
| 44 | C We avoid the slow subfe instruction and instead rely on an extremely unlikely |
| 45 | C branch. |
| 46 | C |
| 47 | C The mullw has the inverse in the first operand, since 0xAA..AB won't allow |
| 48 | C any early-out. The src[] data normally won't either, but there's at least |
| 49 | C a chance, whereas 0xAA..AB never will. If, for instance, src[] is all |
| 50 | C zeros (not a sensible input of course) we run at 7.0 c/l on ppc750. |
| 51 | C |
| 52 | C The mulhwu has the "3" multiplier in the second operand, which lets 750 and |
| 53 | C 7400 use an early-out. |
| 54 | |
| 55 | C INPUT PARAMETERS |
| 56 | define(`rp', `r3') |
| 57 | define(`up', `r4') |
| 58 | define(`n', `r5') |
| 59 | define(`cy', `r6') |
| 60 | |
| 61 | ASM_START() |
| 62 | PROLOGUE(mpn_divexact_by3c) |
| 63 | lwz r11, 0(up) |
| 64 | mtctr n |
| 65 | lis r12, 0xAAAA |
| 66 | ori r12, r12, 0xAAAB |
| 67 | li r10, 3 |
| 68 | |
| 69 | cmplw cr7, cy, r11 |
| 70 | subf r11, cy, r11 |
| 71 | |
| 72 | mullw r0, r11, r12 |
| 73 | stw r0, 0(rp) |
| 74 | bdz L(one) |
| 75 | |
| 76 | L(top): lwzu r9, 4(up) |
| 77 | mulhwu r7, r0, r10 |
| 78 | bgt- cr7, L(adj) C very unlikely branch |
| 79 | L(bko): cmplw cr7, r7, r9 |
| 80 | subf r0, r7, r9 |
| 81 | mullw r0, r12, r0 |
| 82 | stwu r0, 4(rp) |
| 83 | bdnz L(top) |
| 84 | |
| 85 | L(one): mulhwu r3, r0, r10 |
| 86 | blelr+ cr7 |
| 87 | addi r3, r3, 1 |
| 88 | blr |
| 89 | |
| 90 | L(adj): addi r7, r7, 1 |
| 91 | b L(bko) |
| 92 | EPILOGUE() |
| 93 | ASM_END() |