1 /* --------------------------------------------------------------  */
2 /* (C)Copyright 2006,2008,                                         */
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      */
10 /* without modification, are permitted provided that the           */
11 /* following conditions are met:                                   */
12 /*                                                                 */
13 /* - Redistributions of source code must retain the above copyright*/
14 /*   notice, this list of conditions and the following disclaimer. */
15 /*                                                                 */
16 /* - Redistributions in binary form must reproduce the above       */
17 /*   copyright notice, this list of conditions and the following   */
18 /*   disclaimer in the documentation and/or other materials        */
19 /*   provided with the distribution.                               */
20 /*                                                                 */
21 /* - Neither the name of IBM Corporation nor the names of its      */
22 /*   contributors may be used to endorse or promote products       */
23 /*   derived from this software without specific prior written     */
24 /*   permission.                                                   */
25 /*                                                                 */
26 /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND          */
27 /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,     */
28 /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF        */
29 /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE        */
30 /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR            */
31 /* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,    */
32 /* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT    */
33 /* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;    */
34 /* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)        */
35 /* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN       */
36 /* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR    */
37 /* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,  */
38 /* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.              */
39 /* --------------------------------------------------------------  */
40 /* PROLOG END TAG zYx                                              */
41 #ifdef __SPU__
42 #ifndef _SCALBNF4_H_
43 #define _SCALBNF4_H_	1
44 
45 #include <spu_intrinsics.h>
46 
47 /*
48  * FUNCTION
49  *	vector float _scalbnf4(vector float x, vector signed int exp)
50  *
51  * DESCRIPTION
52  *      The _scalbnf4 function returns a vector containing each element of x
53  *      multiplied by 2^n computed efficiently.  This function is computed
54  *      without the assistance of any floating point operations and as such
55  *      does not set any floating point exceptions.
56  *
57  *      Special Cases:
58  *	  - if the exponent is 0, then x is either 0 or a subnormal, and the
59  *          result will be returned as 0.
60  *        - if the result if underflows, it will be returned as 0.
61  *        - if the result overflows, it will be returned as FLT_MAX.
62  *
63  */
_scalbnf4(vector float x,vector signed int exp)64 static __inline vector float _scalbnf4(vector float x, vector signed int exp)
65 {
66   vec_int4 x_exp;
67   vec_uint4 zero;
68   vec_uint4 overflow;
69   vec_uint4 exp_mask = (vec_uint4)spu_splats(0x7F800000);
70   vec_float4 out;
71 
72   /* Extract exponent from x. If the exponent is 0, then
73    * x is either 0 or a denorm and x*2^exp is a zero.
74    */
75   x_exp = spu_and(spu_rlmask((vec_int4)x, -23), 0xFF);
76   zero = spu_cmpeq(x_exp, 0);
77 
78   /* Compute the expected exponent and determine if the
79    * result is within range.
80    */
81   x_exp = spu_add(exp, x_exp);
82 
83   /* Check for zero or overflow of result.
84    * Note: set high bit of flags = 0, since we have to
85    * return -0 when x = -0
86    */
87   zero     = spu_rlmask(spu_orc(zero, spu_cmpgt(x_exp, 0)), -1);
88   overflow = spu_rlmask(spu_cmpgt(x_exp, 255), -1);
89 
90   /* Merge the expect exponent with x's mantissa. Zero the
91    * result if underflow and force to max if overflow.
92    */
93   out = spu_sel(x, (vec_float4)spu_rl(x_exp, 23), exp_mask);
94   out = spu_andc(out, (vec_float4)zero);
95   out = spu_or(out, (vec_float4)overflow);
96 
97   return out;
98 }
99 
100 #endif /* _SCALBNF4_H_ */
101 #endif /* __SPU__ */
102