1 |
207 |
jeremybenn |
/*
|
2 |
|
|
(C) Copyright 2001,2006,
|
3 |
|
|
International Business Machines Corporation,
|
4 |
|
|
Sony Computer Entertainment, Incorporated,
|
5 |
|
|
Toshiba Corporation,
|
6 |
|
|
|
7 |
|
|
All rights reserved.
|
8 |
|
|
|
9 |
|
|
Redistribution and use in source and binary forms, with or without
|
10 |
|
|
modification, are permitted provided that the following conditions are met:
|
11 |
|
|
|
12 |
|
|
* Redistributions of source code must retain the above copyright notice,
|
13 |
|
|
this list of conditions and the following disclaimer.
|
14 |
|
|
* Redistributions in binary form must reproduce the above copyright
|
15 |
|
|
notice, this list of conditions and the following disclaimer in the
|
16 |
|
|
documentation and/or other materials provided with the distribution.
|
17 |
|
|
* Neither the names of the copyright holders nor the names of their
|
18 |
|
|
contributors may be used to endorse or promote products derived from this
|
19 |
|
|
software without specific prior written permission.
|
20 |
|
|
|
21 |
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
|
22 |
|
|
IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
|
23 |
|
|
TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
|
24 |
|
|
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
|
25 |
|
|
OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
26 |
|
|
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
27 |
|
|
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
28 |
|
|
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
|
29 |
|
|
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
|
30 |
|
|
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
|
31 |
|
|
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
32 |
|
|
*/
|
33 |
|
|
#ifndef _REMQUOF_H_
|
34 |
|
|
#define _REMQUOF_H_ 1
|
35 |
|
|
|
36 |
|
|
#include <spu_intrinsics.h>
|
37 |
|
|
#include "headers/vec_literal.h"
|
38 |
|
|
|
39 |
|
|
|
40 |
|
|
static __inline float _remquof(float x, float y, int *quo)
|
41 |
|
|
{
|
42 |
|
|
int n;
|
43 |
|
|
vec_int4 quotient;
|
44 |
|
|
vec_int4 four = { 4, 4, 4, 4 };
|
45 |
|
|
vec_uint4 vx, vy, z, y2, y4;
|
46 |
|
|
vec_uint4 abs_x, abs_y, abs_2x, abs_8y;
|
47 |
|
|
vec_uint4 exp_x, exp_y;
|
48 |
|
|
vec_uint4 zero_x, zero_y;
|
49 |
|
|
vec_uint4 logb_x, logb_y;
|
50 |
|
|
vec_uint4 mant_x, mant_y;
|
51 |
|
|
vec_uint4 not_ge, overflow, quo_pos;
|
52 |
|
|
vec_uint4 result, result0, resultx, cnt, sign, bias;
|
53 |
|
|
vec_uint4 sign_mask = VEC_SPLAT_U32(0x80000000);
|
54 |
|
|
vec_uint4 implied_1 = VEC_SPLAT_U32(0x00800000);
|
55 |
|
|
vec_uint4 mant_mask = VEC_SPLAT_U32(0x007FFFFF);
|
56 |
|
|
|
57 |
|
|
vx = (vec_uint4)spu_promote(x, 0);
|
58 |
|
|
vy = (vec_uint4)spu_promote(y, 0);
|
59 |
|
|
|
60 |
|
|
abs_x = spu_andc(vx, sign_mask);
|
61 |
|
|
abs_y = spu_andc(vy, sign_mask);
|
62 |
|
|
|
63 |
|
|
abs_8y = spu_add(abs_y, VEC_SPLAT_U32(0x01800000)); /* abs_2y = 8 * abs_y */
|
64 |
|
|
|
65 |
|
|
sign = spu_and(vx, sign_mask);
|
66 |
|
|
|
67 |
|
|
quo_pos = spu_cmpgt((vec_int4)spu_and(spu_xor(vx, vy), sign_mask), -1);
|
68 |
|
|
|
69 |
|
|
/* Compute abs_x = fmodf(abs_x, 8*abs_y). If y is greater than 0.125*SMAX
|
70 |
|
|
* (SMAX is the maximum representable float), then return abs_x.
|
71 |
|
|
*/
|
72 |
|
|
{
|
73 |
|
|
/* Determine ilogb of abs_x and abs_8y and
|
74 |
|
|
* extract the mantissas (mant_x, mant_y)
|
75 |
|
|
*/
|
76 |
|
|
exp_x = spu_rlmask(abs_x, -23);
|
77 |
|
|
exp_y = spu_rlmask(abs_8y, -23);
|
78 |
|
|
|
79 |
|
|
resultx = spu_or(spu_cmpgt(abs_8y, abs_x), spu_cmpgt(abs_y, VEC_SPLAT_U32(0x7E7FFFFF)));
|
80 |
|
|
|
81 |
|
|
zero_x = spu_cmpeq(exp_x, 0);
|
82 |
|
|
zero_y = spu_cmpeq(exp_y, 0);
|
83 |
|
|
|
84 |
|
|
logb_x = spu_add(exp_x, -127);
|
85 |
|
|
logb_y = spu_add(exp_y, -127);
|
86 |
|
|
|
87 |
|
|
mant_x = spu_andc(spu_sel(implied_1, abs_x, mant_mask), zero_x);
|
88 |
|
|
mant_y = spu_andc(spu_sel(implied_1, abs_8y, mant_mask), zero_y);
|
89 |
|
|
|
90 |
|
|
/* Compute fixed point fmod of mant_x and mant_y. Set the flag,
|
91 |
|
|
* result0, to all ones if we detect that the final result is
|
92 |
|
|
* ever 0.
|
93 |
|
|
*/
|
94 |
|
|
result0 = spu_or(zero_x, zero_y);
|
95 |
|
|
|
96 |
|
|
n = spu_extract(spu_sub(logb_x, logb_y), 0);
|
97 |
|
|
|
98 |
|
|
|
99 |
|
|
while (n-- > 0) {
|
100 |
|
|
z = spu_sub(mant_x, mant_y);
|
101 |
|
|
|
102 |
|
|
result0 = spu_or(spu_cmpeq(z, 0), result0);
|
103 |
|
|
|
104 |
|
|
mant_x = spu_sel(spu_add(mant_x, mant_x), spu_add(z, z),
|
105 |
|
|
spu_cmpgt((vec_int4)z, -1));
|
106 |
|
|
}
|
107 |
|
|
|
108 |
|
|
z = spu_sub(mant_x, mant_y);
|
109 |
|
|
mant_x = spu_sel(mant_x, z, spu_cmpgt((vec_int4)z, -1));
|
110 |
|
|
|
111 |
|
|
result0 = spu_or(spu_cmpeq(mant_x, 0), result0);
|
112 |
|
|
|
113 |
|
|
/* Convert the result back to floating point and restore
|
114 |
|
|
* the sign. If we flagged the result to be zero (result0),
|
115 |
|
|
* zero it. If we flagged the result to equal its input x,
|
116 |
|
|
* (resultx) then return x.
|
117 |
|
|
*/
|
118 |
|
|
cnt = spu_add(spu_cntlz(mant_x), -8);
|
119 |
|
|
|
120 |
|
|
mant_x = spu_rl(spu_andc(mant_x, implied_1), (vec_int4)cnt);
|
121 |
|
|
|
122 |
|
|
exp_y = spu_sub(exp_y, cnt);
|
123 |
|
|
result0 = spu_orc(result0, spu_cmpgt((vec_int4)exp_y, 0)); /* zero denorm results */
|
124 |
|
|
exp_y = spu_rl(exp_y, 23);
|
125 |
|
|
|
126 |
|
|
result = spu_sel(exp_y, mant_x, mant_mask);
|
127 |
|
|
abs_x = spu_sel(spu_andc(result, spu_rlmask(result0, -1)), abs_x, resultx);
|
128 |
|
|
}
|
129 |
|
|
|
130 |
|
|
/* if (x >= 4*y)
|
131 |
|
|
* x -= 4*y
|
132 |
|
|
* quotient = 4
|
133 |
|
|
* else
|
134 |
|
|
* quotient = 0
|
135 |
|
|
*/
|
136 |
|
|
y4 = spu_andc(spu_add(abs_y, VEC_SPLAT_U32(0x01000000)), zero_y);
|
137 |
|
|
|
138 |
|
|
overflow = spu_cmpgt(abs_y, VEC_SPLAT_U32(0x7EFFFFFF));
|
139 |
|
|
not_ge = spu_or(spu_cmpgt(y4, abs_x), overflow);
|
140 |
|
|
|
141 |
|
|
abs_x = spu_sel((vec_uint4)spu_sub((vec_float4)abs_x, (vec_float4)y4), abs_x, not_ge);
|
142 |
|
|
quotient = spu_andc (four, (vec_int4)not_ge);
|
143 |
|
|
|
144 |
|
|
/* if (x >= 2*y
|
145 |
|
|
* x -= 2*y
|
146 |
|
|
* quotient += 2
|
147 |
|
|
*/
|
148 |
|
|
y2 = spu_andc(spu_add(abs_y, implied_1), zero_y);
|
149 |
|
|
not_ge = spu_cmpgt(y2, abs_x);
|
150 |
|
|
|
151 |
|
|
abs_x = spu_sel((vec_uint4)spu_sub((vec_float4)abs_x, (vec_float4)y2), abs_x, not_ge);
|
152 |
|
|
quotient = spu_sel(spu_add(quotient, 2), quotient, not_ge);
|
153 |
|
|
|
154 |
|
|
/* if (2*x > y)
|
155 |
|
|
* x -= y
|
156 |
|
|
* if (2*x >= y) x -= y
|
157 |
|
|
*/
|
158 |
|
|
abs_2x = spu_add(abs_x, implied_1);
|
159 |
|
|
bias = spu_cmpgt(abs_2x, abs_y);
|
160 |
|
|
abs_x = spu_sel(abs_x, (vec_uint4)spu_sub((vec_float4)abs_x, (vec_float4)abs_y), bias);
|
161 |
|
|
quotient = spu_sub(quotient, (vec_int4)bias);
|
162 |
|
|
|
163 |
|
|
bias = spu_andc(bias, spu_rlmaska((vec_uint4)spu_msub((vec_float4)abs_x, VEC_SPLAT_F32(2.0f), (vec_float4)abs_y), -31));
|
164 |
|
|
abs_x = spu_sel(abs_x, (vec_uint4)spu_sub((vec_float4)abs_x, (vec_float4)abs_y), bias);
|
165 |
|
|
quotient = spu_sub(quotient, (vec_int4)bias);
|
166 |
|
|
|
167 |
|
|
/* Generate a correct final sign
|
168 |
|
|
*/
|
169 |
|
|
result = spu_xor(abs_x, sign);
|
170 |
|
|
|
171 |
|
|
quotient = spu_and(quotient, 7);
|
172 |
|
|
quotient = spu_sel(spu_sub(0, quotient), quotient, quo_pos);
|
173 |
|
|
|
174 |
|
|
*quo = spu_extract(quotient, 0);
|
175 |
|
|
|
176 |
|
|
return (spu_extract((vec_float4)result, 0));
|
177 |
|
|
}
|
178 |
|
|
#endif /* _REMQUOF_H_ */
|