1 /* Compute cubic root of float value.
2 Copyright (C) 1997 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Dirk Alboth <dirka@uni-paderborn.de> and
5 Ulrich Drepper <drepper@cygnus.com>, 1997.
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library; if not, write to the Free
19 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
22 #include <machine/asm.h>
31 ASM_TYPE_DIRECTIVE(f3,@object)
32 f3: .double 0.191502161678719066
33 ASM_SIZE_DIRECTIVE(f3)
34 ASM_TYPE_DIRECTIVE(f2,@object)
35 f2: .double 0.697570460207922770
36 ASM_SIZE_DIRECTIVE(f2)
37 ASM_TYPE_DIRECTIVE(f1,@object)
38 f1: .double 0.492659620528969547
39 ASM_SIZE_DIRECTIVE(f1)
41 #define CBRT2 1.2599210498948731648
42 #define ONE_CBRT2 0.793700525984099737355196796584
43 #define SQR_CBRT2 1.5874010519681994748
44 #define ONE_SQR_CBRT2 0.629960524947436582364439673883
46 ASM_TYPE_DIRECTIVE(factor,@object)
48 factor: .double ONE_SQR_CBRT2
53 ASM_SIZE_DIRECTIVE(factor)
55 ASM_TYPE_DIRECTIVE(two25,@object)
56 two25: .byte 0, 0, 0, 0x4c
57 ASM_SIZE_DIRECTIVE(two25)
60 #define MO(op) op##@GOTOFF(%ebx)
61 #define MOX(op,x) op##@GOTOFF(%ebx,x,1)
64 #define MOX(op,x) op(x)
72 andl $0x7fffffff, %eax
74 cmpl $0x7f800000, %eax
81 addl $_GLOBAL_OFFSET_TABLE_+[.-3b], %ebx
84 cmpl $0x00800000, %eax
102 andl $0x7fffffff, %eax
105 andl $0x807fffff, %edx
107 orl $0x3f000000, %edx
112 flds 8(%esp) /* xm */
116 flds 4(%esp) /* xm */
120 /* The following code has two tracks:
121 a) compute the normalized cbrt value
122 b) compute xe/3 and xe%3
123 The right track computes the value for b) and this is done
124 in an optimized way by avoiding division.
126 But why two tracks at all? Very easy: efficiency. Some FP
127 instruction can overlap with a certain amount of integer (and
128 FP) instructions. So we get (except for the imull) all
129 instructions for free. */
131 fld %st(0) /* xm : xm */
132 fmull MO(f3) /* f3*xm : xm */
133 movl $1431655766, %eax
134 fsubrl MO(f2) /* f2-f3*xm : xm */
136 fmul %st(1) /* (f2-f3*xm)*xm : xm */
138 faddl MO(f1) /* u:=f1+(f2-f3*xm)*xm : xm */
140 fld %st /* u : u : xm */
142 fmul %st(1) /* u*u : u : xm */
143 fld %st(2) /* xm : u*u : u : xm */
144 fadd %st /* 2*xm : u*u : u : xm */
145 fxch %st(1) /* u*u : 2*xm : u : xm */
146 fmul %st(2) /* t2:=u*u*u : 2*xm : u : xm */
148 fadd %st, %st(1) /* t2 : t2+2*xm : u : xm */
149 leal (%edx,%edx,2),%edx
150 fadd %st(0) /* 2*t2 : t2+2*xm : u : xm */
152 faddp %st, %st(3) /* t2+2*xm : u : 2*t2+xm */
154 fmulp /* u*(t2+2*xm) : 2*t2+xm */
155 fdivp %st, %st(1) /* u*(t2+2*xm)/(2*t2+xm) */
156 fmull MOX(16+factor,%ecx) /* u*(t2+2*xm)/(2*t2+xm)*FACT */
158 fildl (%esp) /* xe/3 : u*(t2+2*xm)/(2*t2+xm)*FACT */
159 fxch /* u*(t2+2*xm)/(2*t2+xm)*FACT : xe/3 */
160 fscale /* u*(t2+2*xm)/(2*t2+xm)*FACT*2^xe/3 */
174 /* Return the argument. */
178 weak_alias (__cbrtf, cbrtf)