2 * Copyright (c) 2011 Apple Inc. All rights reserved.
4 * @APPLE_APACHE_LICENSE_HEADER_START@
6 * Licensed under the Apache License, Version 2.0 (the "License") ;
7 * you may not use this file except in compliance with the License.
8 * You may obtain a copy of the License at
10 * http://www.apache.org/licenses/LICENSE-2.0
12 * Unless required by applicable law or agreed to in writing, software
13 * distributed under the License is distributed on an "AS IS" BASIS,
14 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15 * See the License for the specific language governing permissions and
16 * limitations under the License.
18 * @APPLE_APACHE_LICENSE_HEADER_END@
24 Contains: Dynamic Predictor decode routines
26 Copyright: (c) 2001-2011 Apple, Inc.
36 #define ALWAYS_INLINE __attribute__ ((always_inline))
43 static inline int32_t ALWAYS_INLINE
44 sign_of_int (int32_t i)
48 negishift = ((uint32_t) - i) >> 31 ;
49 return negishift | (i >> 31) ;
53 unpc_block (const int32_t * pc1, int32_t * out, int32_t num, int16_t * coefs, int32_t numactive, uint32_t chanbits, uint32_t denshift)
55 register int16_t a0, a1, a2, a3 ;
56 register int32_t b0, b1, b2, b3 ;
58 int32_t sum1, sg, sgn, top, dd ;
61 uint32_t chanshift = 32 - chanbits ;
62 int32_t denhalf = 1 << (denshift - 1) ;
67 // just copy if numactive == 0 (but don't bother if in/out pointers the same)
68 if ((num > 1) && (pc1 != out))
69 memcpy (&out [1], &pc1 [1], (num - 1) * sizeof (int32_t)) ;
74 // short-circuit if numactive == 31
77 /* this code is written such that the in/out buffers can be the same
78 to conserve buffer space on embedded devices like the iPod
81 for (j = 1 ; j < num ; j++)
82 del = pc1 [j] + out [j-1] ;
83 out [j] = (del << chanshift) >> chanshift ;
86 for (j = 1 ; j < num ; j++)
88 del = pc1 [j] + prev ;
89 prev = (del << chanshift) >> chanshift ;
95 for (j = 1 ; j <= numactive ; j++)
97 del = pc1 [j] + out [j-1] ;
98 out [j] = arith_shift_left (del, chanshift) >> chanshift ;
101 lim = numactive + 1 ;
105 // optimization for numactive == 4
106 register int16_t ia0, ia1, ia2, ia3 ;
107 register int32_t ib0, ib1, ib2, ib3 ;
114 for (j = lim ; j < num ; j++)
118 top = out [j - lim] ;
121 ib0 = top - pout [0] ;
122 ib1 = top - pout [-1] ;
123 ib2 = top - pout [-2] ;
124 ib3 = top - pout [-3] ;
126 sum1 = (denhalf - ia0 * ib0 - ia1 * ib1 - ia2 * ib2 - ia3 * ib3) >> denshift ;
130 sg = sign_of_int (del) ;
133 out [j] = arith_shift_left (del, chanshift) >> chanshift ;
137 sgn = sign_of_int (ib3) ;
139 del0 -= (4 - 3) * ((sgn * ib3) >> denshift) ;
143 sgn = sign_of_int (ib2) ;
145 del0 -= (4 - 2) * ((sgn * ib2) >> denshift) ;
149 sgn = sign_of_int (ib1) ;
151 del0 -= (4 - 1) * ((sgn * ib1) >> denshift) ;
155 ia0 -= sign_of_int (ib0) ;
159 // note: to avoid unnecessary negations, we flip the value of "sgn"
160 sgn = -sign_of_int (ib3) ;
162 del0 -= (4 - 3) * ((sgn * ib3) >> denshift) ;
166 sgn = -sign_of_int (ib2) ;
168 del0 -= (4 - 2) * ((sgn * ib2) >> denshift) ;
172 sgn = -sign_of_int (ib1) ;
174 del0 -= (4 - 1) * ((sgn * ib1) >> denshift) ;
178 ia0 += sign_of_int (ib0) ;
187 else if (numactive == 8)
189 register int16_t a4, a5, a6, a7 ;
190 register int32_t b4, b5, b6, b7 ;
192 // optimization for numactive == 8
202 for (j = lim ; j < num ; j++)
206 top = out [j - lim] ;
209 b0 = top - (*pout--) ;
210 b1 = top - (*pout--) ;
211 b2 = top - (*pout--) ;
212 b3 = top - (*pout--) ;
213 b4 = top - (*pout--) ;
214 b5 = top - (*pout--) ;
215 b6 = top - (*pout--) ;
219 sum1 = (denhalf - a0 * b0 - a1 * b1 - a2 * b2 - a3 * b3
220 - a4 * b4 - a5 * b5 - a6 * b6 - a7 * b7) >> denshift ;
224 sg = sign_of_int (del) ;
227 out [j] = arith_shift_left (del, chanshift) >> chanshift ;
231 sgn = sign_of_int (b7) ;
233 del0 -= 1 * ((sgn * b7) >> denshift) ;
237 sgn = sign_of_int (b6) ;
239 del0 -= 2 * ((sgn * b6) >> denshift) ;
243 sgn = sign_of_int (b5) ;
245 del0 -= 3 * ((sgn * b5) >> denshift) ;
249 sgn = sign_of_int (b4) ;
251 del0 -= 4 * ((sgn * b4) >> denshift) ;
255 sgn = sign_of_int (b3) ;
257 del0 -= 5 * ((sgn * b3) >> denshift) ;
261 sgn = sign_of_int (b2) ;
263 del0 -= 6 * ((sgn * b2) >> denshift) ;
267 sgn = sign_of_int (b1) ;
269 del0 -= 7 * ((sgn * b1) >> denshift) ;
273 a0 -= sign_of_int (b0) ;
277 // note: to avoid unnecessary negations, we flip the value of "sgn"
278 sgn = -sign_of_int (b7) ;
280 del0 -= 1 * ((sgn * b7) >> denshift) ;
284 sgn = -sign_of_int (b6) ;
286 del0 -= 2 * ((sgn * b6) >> denshift) ;
290 sgn = -sign_of_int (b5) ;
292 del0 -= 3 * ((sgn * b5) >> denshift) ;
296 sgn = -sign_of_int (b4) ;
298 del0 -= 4 * ((sgn * b4) >> denshift) ;
302 sgn = -sign_of_int (b3) ;
304 del0 -= 5 * ((sgn * b3) >> denshift) ;
308 sgn = -sign_of_int (b2) ;
310 del0 -= 6 * ((sgn * b2) >> denshift) ;
314 sgn = -sign_of_int (b1) ;
316 del0 -= 7 * ((sgn * b1) >> denshift) ;
320 a0 += sign_of_int (b0) ;
336 for (j = lim ; j < num ; j++)
344 for (k = 0 ; k < numactive ; k++)
345 sum1 += coefs [k] * (pout [-k] - top) ;
349 sg = sign_of_int (del) ;
350 del += top + ((sum1 + denhalf) >> denshift) ;
351 out [j] = (del << chanshift) >> chanshift ;
355 for (k = (numactive - 1) ; k >= 0 ; k--)
357 dd = top - pout [-k] ;
358 sgn = sign_of_int (dd) ;
360 del0 -= (numactive - k) * ((sgn * dd) >> denshift) ;
367 for (k = (numactive - 1) ; k >= 0 ; k--)
369 dd = top - pout [-k] ;
370 sgn = sign_of_int (dd) ;
372 del0 -= (numactive - k) * ((-sgn * dd) >> denshift) ;