| Austin Schuh | dace2a6 | 2020-08-18 10:56:48 -0700 | [diff] [blame] | 1 | dnl Copyright 1999, 2000, 2002, 2003 Free Software Foundation, Inc. |
| 2 | |
| 3 | dnl This file is part of the GNU MP Library. |
| 4 | dnl |
| 5 | dnl The GNU MP Library is free software; you can redistribute it and/or modify |
| 6 | dnl it under the terms of either: |
| 7 | dnl |
| 8 | dnl * the GNU Lesser General Public License as published by the Free |
| 9 | dnl Software Foundation; either version 3 of the License, or (at your |
| 10 | dnl option) any later version. |
| 11 | dnl |
| 12 | dnl or |
| 13 | dnl |
| 14 | dnl * the GNU General Public License as published by the Free Software |
| 15 | dnl Foundation; either version 2 of the License, or (at your option) any |
| 16 | dnl later version. |
| 17 | dnl |
| 18 | dnl or both in parallel, as here. |
| 19 | dnl |
| 20 | dnl The GNU MP Library is distributed in the hope that it will be useful, but |
| 21 | dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
| 22 | dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 23 | dnl for more details. |
| 24 | dnl |
| 25 | dnl You should have received copies of the GNU General Public License and the |
| 26 | dnl GNU Lesser General Public License along with the GNU MP Library. If not, |
| 27 | dnl see https://www.gnu.org/licenses/. |
| 28 | |
| 29 | |
| 30 | dnl Optimizations: |
| 31 | dnl * Avoid skip instructions |
| 32 | dnl * Put carry-generating and carry-consuming insns consecutively |
| 33 | dnl * Don't allocate any stack, "home" positions for parameters could be used. |
| 34 | |
| 35 | include(`../config.m4') |
| 36 | |
| 37 | define(`p0',`%r28') |
| 38 | define(`p1',`%r29') |
| 39 | define(`t32',`%r19') |
| 40 | define(`t0',`%r20') |
| 41 | define(`t1',`%r21') |
| 42 | define(`x',`%r22') |
| 43 | define(`m0',`%r23') |
| 44 | define(`m1',`%r24') |
| 45 | |
| 46 | ifdef(`HAVE_ABI_2_0w', |
| 47 | ` .level 2.0w |
| 48 | ',` .level 2.0 |
| 49 | ') |
| 50 | PROLOGUE(mpn_umul_ppmm_r) |
| 51 | ldo 128(%r30),%r30 |
| 52 | ifdef(`HAVE_ABI_2_0w', |
| 53 | ` std %r26,-64(%r30) |
| 54 | std %r25,-56(%r30) |
| 55 | copy %r24,%r31 |
| 56 | ',` |
| 57 | depd %r25,31,32,%r26 |
| 58 | std %r26,-64(%r30) |
| 59 | depd %r23,31,32,%r24 |
| 60 | std %r24,-56(%r30) |
| 61 | ldw -180(%r30),%r31 |
| 62 | ') |
| 63 | |
| 64 | fldd -64(%r30),%fr4 |
| 65 | fldd -56(%r30),%fr5 |
| 66 | |
| 67 | xmpyu %fr5R,%fr4R,%fr6 |
| 68 | fstd %fr6,-128(%r30) |
| 69 | xmpyu %fr5R,%fr4L,%fr7 |
| 70 | fstd %fr7,-120(%r30) |
| 71 | xmpyu %fr5L,%fr4R,%fr8 |
| 72 | fstd %fr8,-112(%r30) |
| 73 | xmpyu %fr5L,%fr4L,%fr9 |
| 74 | fstd %fr9,-104(%r30) |
| 75 | |
| 76 | depdi,z 1,31,1,t32 C t32 = 2^32 |
| 77 | |
| 78 | ldd -128(%r30),p0 C lo = low 64 bit of product |
| 79 | ldd -120(%r30),m0 C m0 = mid0 64 bit of product |
| 80 | ldd -112(%r30),m1 C m1 = mid1 64 bit of product |
| 81 | ldd -104(%r30),p1 C hi = high 64 bit of product |
| 82 | |
| 83 | add,l,*nuv m0,m1,x C x = m1+m0 |
| 84 | add,l t32,p1,p1 C propagate carry to mid of p1 |
| 85 | depd,z x,31,32,t0 C lo32(m1+m0) |
| 86 | add t0,p0,p0 |
| 87 | extrd,u x,31,32,t1 C hi32(m1+m0) |
| 88 | add,dc t1,p1,p1 |
| 89 | |
| 90 | std p0,0(%r31) C store low half of product |
| 91 | ifdef(`HAVE_ABI_2_0w', |
| 92 | ` copy p1,%r28 C return val in %r28 |
| 93 | ',` extrd,u p1,31,32,%r28 C return val in %r28,%r29 |
| 94 | ') |
| 95 | bve (%r2) |
| 96 | ldo -128(%r30),%r30 |
| 97 | EPILOGUE(mpn_umul_ppmm_r) |