1 dnl HP-PA 2.0 64-bit mpn_sqr_diagonal.
3 dnl Copyright 2001, 2002, 2003 Free Software Foundation, Inc.
5 dnl This file is part of the GNU MP Library.
7 dnl The GNU MP Library is free software; you can redistribute it and/or modify
8 dnl it under the terms of the GNU Lesser General Public License as published
9 dnl by the Free Software Foundation; either version 3 of the License, or (at
10 dnl your option) any later version.
12 dnl The GNU MP Library is distributed in the hope that it will be useful, but
13 dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
14 dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
15 dnl License for more details.
17 dnl You should have received a copy of the GNU Lesser General Public License
18 dnl along with the GNU MP Library. If not, see http://www.gnu.org/licenses/.
21 dnl This code runs at 7.25 cycles/limb on PA8000 and 7.75 cycles/limb on
22 dnl PA8500. The cache would saturate at 5 cycles/limb, so there is some room
25 include(`../config.m4')
38 ifdef(`HAVE_ABI_2_0w',
42 PROLOGUE(mpn_sqr_diagonal)
49 xmpyu %fr8l,%fr8r,%fr10
51 xmpyu %fr8r,%fr8r,%fr9
53 xmpyu %fr8l,%fr8l,%fr11
59 fldds,ma 8(up),%fr8 C load next up limb
60 xmpyu %fr4l,%fr4r,%fr6
62 xmpyu %fr4r,%fr4r,%fr5 C multiply in fp regs
64 xmpyu %fr4l,%fr4l,%fr7
67 ldd -16(rp),p00 C accumulate in int regs
79 xmpyu %fr8l,%fr8r,%fr10
81 xmpyu %fr8r,%fr8r,%fr9
83 xmpyu %fr8l,%fr8l,%fr11
98 xmpyu %fr4l,%fr4r,%fr6
100 xmpyu %fr4r,%fr4r,%fr5
102 xmpyu %fr4l,%fr4l,%fr7
127 xmpyu %fr8l,%fr8r,%fr10
128 fstd %fr10,-120(%r30)
129 xmpyu %fr8r,%fr8r,%fr9
131 xmpyu %fr8l,%fr8l,%fr11
160 xmpyu %fr8l,%fr8r,%fr10
161 fstd %fr10,-128(%r30)
162 xmpyu %fr8r,%fr8r,%fr9
164 xmpyu %fr8l,%fr8l,%fr11
180 EPILOGUE(mpn_sqr_diagonal)