1 dnl IA-64 mpn_invert_limb -- Invert a normalized limb.
3 dnl Copyright 2000, 2002, 2004 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/.
20 include(`../config.m4')
29 C It should be possible to avoid the xmpy.hu and the following tests by
30 C explicitly chopping in the last fma. That would save about 10 cycles.
35 ifdef(`HAVE_DOUBLE_IEEE_LITTLE_ENDIAN',`
36 .LC0: data4 0x00000000, 0x80000000, 0x0000403f, 0x00000000 C 2^64
37 .LC1: data4 0x00000000, 0x80000000, 0x0000407f, 0x00000000 C 2^128
39 ',`ifdef(`HAVE_DOUBLE_IEEE_BIG_ENDIAN',`
40 .LC0: data4 0x403f8000, 0x00000000, 0x00000000, 0x00000000 C 2^64
41 .LC1: data4 0x407f8000, 0x00000000, 0x00000000, 0x00000000 C 2^128
43 ',`m4_error(`Oops, need to know float endianness
47 PROLOGUE(mpn_invert_limb)
49 addl r14 = @gprel(.LC0), gp
50 addl r15 = @gprel(.LC1), gp
52 add r9 = r32, r32 C check for d = 2^63
54 ldfe f10 = [r14] C 2^64
55 ldfe f8 = [r15] C 2^128
56 cmp.eq p6, p0 = 0, r9 C check for d = 2^63
57 mov r8 = -1 C retval for 2^63
58 (p6) br.ret.spnt.many b0
60 fmpy.s1 f11 = f7, f10 C f11 = d * 2^64
61 fnma.s1 f6 = f7, f10, f8 C f6 = 2^128 - d * 2^64
63 frcpa.s1 f8, p6 = f6, f7
65 (p6) fnma.s1 f9 = f7, f8, f1
66 (p6) fmpy.s1 f10 = f6, f8
68 (p6) fmpy.s1 f11 = f9, f9
69 (p6) fma.s1 f10 = f9, f10, f10
71 (p6) fma.s1 f8 = f9, f8, f8
72 (p6) fma.s1 f9 = f11, f10, f10
74 (p6) fma.s1 f8 = f11, f8, f8
75 (p6) fnma.s1 f10 = f7, f9, f6
77 (p6) fma.s1 f8 = f10, f8, f9
79 fcvt.fxu.trunc.s1 f8 = f8
82 xmpy.hu f10 = f8, f7 C di * d
85 andcm r9 = -1, r32 C one's complement
87 cmp.ltu p6, p0 = r9, r14 C got overflow?
89 (p6) add r8 = -1, r8 C adjust di down