?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
autodetection: convert path to native separators before displaying it.
[Rockbox.git] / apps / codecs / libmusepack / synth_filter.c
blobd48b563a0b4dc4bc275ea13ccbbf33a3e92ae353
1 /*
2 Copyright (c) 2005, The Musepack Development Team
3 All rights reserved.
4
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions are
7 met:
8
9 * Redistributions of source code must retain the above copyright
10 notice, this list of conditions and the following disclaimer.
11
12 * Redistributions in binary form must reproduce the above
13 copyright notice, this list of conditions and the following
14 disclaimer in the documentation and/or other materials provided
15 with the distribution.
16
17 * Neither the name of the The Musepack Development Team nor the
18 names of its contributors may be used to endorse or promote
19 products derived from this software without specific prior
20 written permission.
21
22 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35 /// \file synth_filter.c
36 /// Synthesis functions.
37 /// \todo document me
38
39 #include "musepack.h"
40 #include "internal.h"
41
42 /* C O N S T A N T S */
43 #undef _
44
45 #if defined(MPC_FIXED_POINT)
46 #if defined(OPTIMIZE_FOR_SPEED)
47 // round to +/- 2^14 as pre-shift before 32=32x32-multiply
48 #define _(value) (MPC_SHR_RND(value, 3))
49
50 // round to +/- 2^17 as pre-shift before 32=32x32-multiply
51 #define MPC_V_PRESHIFT(X) MPC_SHR_RND(X, 14)
52 #else
53 // saturate to +/- 2^31 (= value << (31-17)), D-values are +/- 2^17
54 #define _(value) (value << (14))
55
56 // do not perform pre-shift
57 #define MPC_V_PRESHIFT(X) (X)
58 #endif
59 #else
60 // IMPORTANT: internal scaling is somehow strange for floating point, therefore we scale the coefficients Di_opt
61 // by the correct amount to have proper scaled output
62 #define _(value) MAKE_MPC_SAMPLE((double)value*(double)(0x1000))
63
64 // do not perform pre-shift
65 #define MPC_V_PRESHIFT(X) (X)
66 #endif
67
68 // Di_opt coefficients are +/- 2^17
69 static const MPC_SAMPLE_FORMAT Di_opt [512] ICONST_ATTR = {
70 /* 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 */
71 /* 0 */ _( 0), _( -29), _( 213), _( -459), _( 2037), _(-5153), _( 6574), _(-37489), _(75038), _(37489), _(6574), _( 5153), _(2037), _( 459), _(213), _(29),
72 /* 1 */ _( -1), _( -31), _( 218), _( -519), _( 2000), _(-5517), _( 5959), _(-39336), _(74992), _(35640), _(7134), _( 4788), _(2063), _( 401), _(208), _(26),
73 /* 2 */ _( -1), _( -35), _( 222), _( -581), _( 1952), _(-5879), _( 5288), _(-41176), _(74856), _(33791), _(7640), _( 4425), _(2080), _( 347), _(202), _(24),
74 /* 3 */ _( -1), _( -38), _( 225), _( -645), _( 1893), _(-6237), _( 4561), _(-43006), _(74630), _(31947), _(8092), _( 4063), _(2087), _( 294), _(196), _(21),
75 /* 4 */ _( -1), _( -41), _( 227), _( -711), _( 1822), _(-6589), _( 3776), _(-44821), _(74313), _(30112), _(8492), _( 3705), _(2085), _( 244), _(190), _(19),
76 /* 5 */ _( -1), _( -45), _( 228), _( -779), _( 1739), _(-6935), _( 2935), _(-46617), _(73908), _(28289), _(8840), _( 3351), _(2075), _( 197), _(183), _(17),
77 /* 6 */ _( -1), _( -49), _( 228), _( -848), _( 1644), _(-7271), _( 2037), _(-48390), _(73415), _(26482), _(9139), _( 3004), _(2057), _( 153), _(176), _(16),
78 /* 7 */ _( -2), _( -53), _( 227), _( -919), _( 1535), _(-7597), _( 1082), _(-50137), _(72835), _(24694), _(9389), _( 2663), _(2032), _( 111), _(169), _(14),
79 /* 8 */ _( -2), _( -58), _( 224), _( -991), _( 1414), _(-7910), _( 70), _(-51853), _(72169), _(22929), _(9592), _( 2330), _(2001), _( 72), _(161), _(13),
80 /* 9 */ _( -2), _( -63), _( 221), _(-1064), _( 1280), _(-8209), _( -998), _(-53534), _(71420), _(21189), _(9750), _( 2006), _(1962), _( 36), _(154), _(11),
81 /* 10 */ _( -2), _( -68), _( 215), _(-1137), _( 1131), _(-8491), _( -2122), _(-55178), _(70590), _(19478), _(9863), _( 1692), _(1919), _( 2), _(147), _(10),
82 /* 11 */ _( -3), _( -73), _( 208), _(-1210), _( 970), _(-8755), _( -3300), _(-56778), _(69679), _(17799), _(9935), _( 1388), _(1870), _( -29), _(139), _( 9),
83 /* 12 */ _( -3), _( -79), _( 200), _(-1283), _( 794), _(-8998), _( -4533), _(-58333), _(68692), _(16155), _(9966), _( 1095), _(1817), _( -57), _(132), _( 8),
84 /* 13 */ _( -4), _( -85), _( 189), _(-1356), _( 605), _(-9219), _( -5818), _(-59838), _(67629), _(14548), _(9959), _( 814), _(1759), _( -83), _(125), _( 7),
85 /* 14 */ _( -4), _( -91), _( 177), _(-1428), _( 402), _(-9416), _( -7154), _(-61289), _(66494), _(12980), _(9916), _( 545), _(1698), _(-106), _(117), _( 7),
86 /* 15 */ _( -5), _( -97), _( 163), _(-1498), _( 185), _(-9585), _( -8540), _(-62684), _(65290), _(11455), _(9838), _( 288), _(1634), _(-127), _(111), _( 6),
87 /* 16 */ _( -5), _(-104), _( 146), _(-1567), _( -45), _(-9727), _( -9975), _(-64019), _(64019), _( 9975), _(9727), _( 45), _(1567), _(-146), _(104), _( 5),
88 /* 17 */ _( -6), _(-111), _( 127), _(-1634), _( -288), _(-9838), _(-11455), _(-65290), _(62684), _( 8540), _(9585), _( -185), _(1498), _(-163), _( 97), _( 5),
89 /* 18 */ _( -7), _(-117), _( 106), _(-1698), _( -545), _(-9916), _(-12980), _(-66494), _(61289), _( 7154), _(9416), _( -402), _(1428), _(-177), _( 91), _( 4),
90 /* 19 */ _( -7), _(-125), _( 83), _(-1759), _( -814), _(-9959), _(-14548), _(-67629), _(59838), _( 5818), _(9219), _( -605), _(1356), _(-189), _( 85), _( 4),
91 /* 20 */ _( -8), _(-132), _( 57), _(-1817), _(-1095), _(-9966), _(-16155), _(-68692), _(58333), _( 4533), _(8998), _( -794), _(1283), _(-200), _( 79), _( 3),
92 /* 21 */ _( -9), _(-139), _( 29), _(-1870), _(-1388), _(-9935), _(-17799), _(-69679), _(56778), _( 3300), _(8755), _( -970), _(1210), _(-208), _( 73), _( 3),
93 /* 22 */ _(-10), _(-147), _( -2), _(-1919), _(-1692), _(-9863), _(-19478), _(-70590), _(55178), _( 2122), _(8491), _(-1131), _(1137), _(-215), _( 68), _( 2),
94 /* 23 */ _(-11), _(-154), _( -36), _(-1962), _(-2006), _(-9750), _(-21189), _(-71420), _(53534), _( 998), _(8209), _(-1280), _(1064), _(-221), _( 63), _( 2),
95 /* 24 */ _(-13), _(-161), _( -72), _(-2001), _(-2330), _(-9592), _(-22929), _(-72169), _(51853), _( -70), _(7910), _(-1414), _( 991), _(-224), _( 58), _( 2),
96 /* 25 */ _(-14), _(-169), _(-111), _(-2032), _(-2663), _(-9389), _(-24694), _(-72835), _(50137), _(-1082), _(7597), _(-1535), _( 919), _(-227), _( 53), _( 2),
97 /* 26 */ _(-16), _(-176), _(-153), _(-2057), _(-3004), _(-9139), _(-26482), _(-73415), _(48390), _(-2037), _(7271), _(-1644), _( 848), _(-228), _( 49), _( 1),
98 /* 27 */ _(-17), _(-183), _(-197), _(-2075), _(-3351), _(-8840), _(-28289), _(-73908), _(46617), _(-2935), _(6935), _(-1739), _( 779), _(-228), _( 45), _( 1),
99 /* 28 */ _(-19), _(-190), _(-244), _(-2085), _(-3705), _(-8492), _(-30112), _(-74313), _(44821), _(-3776), _(6589), _(-1822), _( 711), _(-227), _( 41), _( 1),
100 /* 29 */ _(-21), _(-196), _(-294), _(-2087), _(-4063), _(-8092), _(-31947), _(-74630), _(43006), _(-4561), _(6237), _(-1893), _( 645), _(-225), _( 38), _( 1),
101 /* 30 */ _(-24), _(-202), _(-347), _(-2080), _(-4425), _(-7640), _(-33791), _(-74856), _(41176), _(-5288), _(5879), _(-1952), _( 581), _(-222), _( 35), _( 1),
102 /* 31 */ _(-26), _(-208), _(-401), _(-2063), _(-4788), _(-7134), _(-35640), _(-74992), _(39336), _(-5959), _(5517), _(-2000), _( 519), _(-218), _( 31), _( 1)
103 };
104
105 #undef _
106
107 // needed to prevent from internal overflow in calculate_V
108 #define OVERFLOW_FIX 1
109
110 // V-coefficients were expanded (<<) by V_COEFFICIENT_EXPAND
111 #define V_COEFFICIENT_EXPAND 27
112
113 #if defined(MPC_FIXED_POINT)
114 #if defined(OPTIMIZE_FOR_SPEED)
115 // define 32=32x32-multiplication for DCT-coefficients with samples, vcoef will be pre-shifted on creation
116 // samples are rounded to +/- 2^19 as pre-shift before 32=32x32-multiply
117 #define MPC_MULTIPLY_V(sample, vcoef) ( MPC_SHR_RND(sample, 12) * vcoef )
118
119 // pre- and postscale are used to avoid internal overflow in synthesis calculation
120 #define MPC_MULTIPLY_V_PRESCALE(sample, vcoef) ( MPC_SHR_RND(sample, (12+OVERFLOW_FIX)) * vcoef )
121 #define MPC_MULTIPLY_V_POSTSCALE(sample, vcoef) ( MPC_SHR_RND(sample, (12-OVERFLOW_FIX)) * vcoef )
122 #define MPC_V_POSTSCALE(sample) (sample<<OVERFLOW_FIX)
123
124 // round to +/- 2^16 as pre-shift before 32=32x32-multiply
125 #define MPC_MAKE_INVCOS(value) (MPC_SHR_RND(value, 15))
126 #else
127 // define 64=32x32-multiplication for DCT-coefficients with samples. Via usage of MPC_FRACT highly optimized assembler might be used
128 // MULTIPLY_FRACT will do >>32 after multiplication, as V-coef were expanded by V_COEFFICIENT_EXPAND we'll correct this on the result.
129 // Will loose 5bit accuracy on result in fract part without effect on final audio result
130 #define MPC_MULTIPLY_V(sample, vcoef) ( (MPC_MULTIPLY_FRACT(sample, vcoef)) << (32-V_COEFFICIENT_EXPAND) )
131
132 // pre- and postscale are used to avoid internal overflow in synthesis calculation
133 #define MPC_MULTIPLY_V_PRESCALE(sample, vcoef) ( (MPC_MULTIPLY_FRACT(sample, vcoef)) << (32-V_COEFFICIENT_EXPAND-OVERFLOW_FIX) )
134 #define MPC_MULTIPLY_V_POSTSCALE(sample, vcoef) ( (MPC_MULTIPLY_FRACT(sample, vcoef)) << (32-V_COEFFICIENT_EXPAND+OVERFLOW_FIX) )
135 #define MPC_V_POSTSCALE(sample) (sample<<OVERFLOW_FIX)
136
137 // directly use accurate 32bit-coefficients
138 #define MPC_MAKE_INVCOS(value) (value)
139 #endif
140 #else
141 // for floating point use the standard multiplication macro
142 #define MPC_MULTIPLY_V (sample, vcoef) ( MPC_MULTIPLY(sample, vcoef) )
143 #define MPC_MULTIPLY_V_PRESCALE (sample, vcoef) ( MPC_MULTIPLY(sample, vcoef) )
144 #define MPC_MULTIPLY_V_POSTSCALE(sample, vcoef) ( MPC_MULTIPLY(sample, vcoef) )
145 #define MPC_V_POSTSCALE(sample) (sample)
146
147 // downscale the accurate 32bit-coefficients and convert to float
148 #define MPC_MAKE_INVCOS(value) MAKE_MPC_SAMPLE((double)value/(double)(1<<V_COEFFICIENT_EXPAND))
149 #endif
150
151 // define constants for DCT-synthesis
152 // INVCOSxx = (0.5 / cos(xx*PI/64)) << 27, <<27 to saturate to +/- 2^31
153 #define INVCOS01 MPC_MAKE_INVCOS( 67189797)
154 #define INVCOS02 MPC_MAKE_INVCOS( 67433575)
155 #define INVCOS03 MPC_MAKE_INVCOS( 67843164)
156 #define INVCOS04 MPC_MAKE_INVCOS( 68423604)
157 #define INVCOS05 MPC_MAKE_INVCOS( 69182167)
158 #define INVCOS06 MPC_MAKE_INVCOS( 70128577)
159 #define INVCOS07 MPC_MAKE_INVCOS( 71275330)
160 #define INVCOS08 MPC_MAKE_INVCOS( 72638111)
161 #define INVCOS09 MPC_MAKE_INVCOS( 74236348)
162 #define INVCOS10 MPC_MAKE_INVCOS( 76093940)
163 #define INVCOS11 MPC_MAKE_INVCOS( 78240207)
164 #define INVCOS12 MPC_MAKE_INVCOS( 80711144)
165 #define INVCOS13 MPC_MAKE_INVCOS( 83551089)
166 #define INVCOS14 MPC_MAKE_INVCOS( 86814950)
167 #define INVCOS15 MPC_MAKE_INVCOS( 90571242)
168 #define INVCOS16 MPC_MAKE_INVCOS( 94906266)
169 #define INVCOS17 MPC_MAKE_INVCOS( 99929967)
170 #define INVCOS18 MPC_MAKE_INVCOS( 105784321)
171 #define INVCOS19 MPC_MAKE_INVCOS( 112655602)
172 #define INVCOS20 MPC_MAKE_INVCOS( 120792764)
173 #define INVCOS21 MPC_MAKE_INVCOS( 130535899)
174 #define INVCOS22 MPC_MAKE_INVCOS( 142361749)
175 #define INVCOS23 MPC_MAKE_INVCOS( 156959571)
176 #define INVCOS24 MPC_MAKE_INVCOS( 175363913)
177 #define INVCOS25 MPC_MAKE_INVCOS( 199201203)
178 #define INVCOS26 MPC_MAKE_INVCOS( 231182936)
179 #define INVCOS27 MPC_MAKE_INVCOS( 276190692)
180 #define INVCOS28 MPC_MAKE_INVCOS( 343988688)
181 #define INVCOS29 MPC_MAKE_INVCOS( 457361460)
182 #define INVCOS30 MPC_MAKE_INVCOS( 684664578)
183 #define INVCOS31 MPC_MAKE_INVCOS(1367679739)
184
185 void
186 mpc_calculate_new_V ( const MPC_SAMPLE_FORMAT * Sample, MPC_SAMPLE_FORMAT * V )
187 ICODE_ATTR_MPC_LARGE_IRAM;
188
189 void
190 mpc_calculate_new_V ( const MPC_SAMPLE_FORMAT * Sample, MPC_SAMPLE_FORMAT * V )
191 {
192 // Calculating new V-buffer values for left channel
193 // calculate new V-values (ISO-11172-3, p. 39)
194 // based upon fast-MDCT algorithm by Byeong Gi Lee
195 MPC_SAMPLE_FORMAT A[16];
196 MPC_SAMPLE_FORMAT B[16];
197 MPC_SAMPLE_FORMAT tmp;
198
199 A[ 0] = Sample[ 0] + Sample[31];
200 A[ 1] = Sample[ 1] + Sample[30];
201 A[ 2] = Sample[ 2] + Sample[29];
202 A[ 3] = Sample[ 3] + Sample[28];
203 A[ 4] = Sample[ 4] + Sample[27];
204 A[ 5] = Sample[ 5] + Sample[26];
205 A[ 6] = Sample[ 6] + Sample[25];
206 A[ 7] = Sample[ 7] + Sample[24];
207 A[ 8] = Sample[ 8] + Sample[23];
208 A[ 9] = Sample[ 9] + Sample[22];
209 A[10] = Sample[10] + Sample[21];
210 A[11] = Sample[11] + Sample[20];
211 A[12] = Sample[12] + Sample[19];
212 A[13] = Sample[13] + Sample[18];
213 A[14] = Sample[14] + Sample[17];
214 A[15] = Sample[15] + Sample[16];
215 // 16 adds
216
217 B[ 0] = A[ 0] + A[15];
218 B[ 1] = A[ 1] + A[14];
219 B[ 2] = A[ 2] + A[13];
220 B[ 3] = A[ 3] + A[12];
221 B[ 4] = A[ 4] + A[11];
222 B[ 5] = A[ 5] + A[10];
223 B[ 6] = A[ 6] + A[ 9];
224 B[ 7] = A[ 7] + A[ 8];;
225 B[ 8] = MPC_MULTIPLY_V((A[ 0] - A[15]), INVCOS02);
226 B[ 9] = MPC_MULTIPLY_V((A[ 1] - A[14]), INVCOS06);
227 B[10] = MPC_MULTIPLY_V((A[ 2] - A[13]), INVCOS10);
228 B[11] = MPC_MULTIPLY_V((A[ 3] - A[12]), INVCOS14);
229 B[12] = MPC_MULTIPLY_V((A[ 4] - A[11]), INVCOS18);
230 B[13] = MPC_MULTIPLY_V((A[ 5] - A[10]), INVCOS22);
231 B[14] = MPC_MULTIPLY_V((A[ 6] - A[ 9]), INVCOS26);
232 B[15] = MPC_MULTIPLY_V((A[ 7] - A[ 8]), INVCOS30);
233 // 8 adds, 8 subs, 8 muls, 8 shifts
234
235 A[ 0] = B[ 0] + B[ 7];
236 A[ 1] = B[ 1] + B[ 6];
237 A[ 2] = B[ 2] + B[ 5];
238 A[ 3] = B[ 3] + B[ 4];
239 A[ 4] = MPC_MULTIPLY_V((B[ 0] - B[ 7]), INVCOS04);
240 A[ 5] = MPC_MULTIPLY_V((B[ 1] - B[ 6]), INVCOS12);
241 A[ 6] = MPC_MULTIPLY_V((B[ 2] - B[ 5]), INVCOS20);
242 A[ 7] = MPC_MULTIPLY_V((B[ 3] - B[ 4]), INVCOS28);
243 A[ 8] = B[ 8] + B[15];
244 A[ 9] = B[ 9] + B[14];
245 A[10] = B[10] + B[13];
246 A[11] = B[11] + B[12];
247 A[12] = MPC_MULTIPLY_V((B[ 8] - B[15]), INVCOS04);
248 A[13] = MPC_MULTIPLY_V((B[ 9] - B[14]), INVCOS12);
249 A[14] = MPC_MULTIPLY_V((B[10] - B[13]), INVCOS20);
250 A[15] = MPC_MULTIPLY_V((B[11] - B[12]), INVCOS28);
251 // 8 adds, 8 subs, 8 muls, 8 shifts
252
253 B[ 0] = A[ 0] + A[ 3];
254 B[ 1] = A[ 1] + A[ 2];
255 B[ 2] = MPC_MULTIPLY_V((A[ 0] - A[ 3]), INVCOS08);
256 B[ 3] = MPC_MULTIPLY_V((A[ 1] - A[ 2]), INVCOS24);
257 B[ 4] = A[ 4] + A[ 7];
258 B[ 5] = A[ 5] + A[ 6];
259 B[ 6] = MPC_MULTIPLY_V((A[ 4] - A[ 7]), INVCOS08);
260 B[ 7] = MPC_MULTIPLY_V((A[ 5] - A[ 6]), INVCOS24);
261 B[ 8] = A[ 8] + A[11];
262 B[ 9] = A[ 9] + A[10];
263 B[10] = MPC_MULTIPLY_V((A[ 8] - A[11]), INVCOS08);
264 B[11] = MPC_MULTIPLY_V((A[ 9] - A[10]), INVCOS24);
265 B[12] = A[12] + A[15];
266 B[13] = A[13] + A[14];
267 B[14] = MPC_MULTIPLY_V((A[12] - A[15]), INVCOS08);
268 B[15] = MPC_MULTIPLY_V((A[13] - A[14]), INVCOS24);
269 // 8 adds, 8 subs, 8 muls, 8 shifts
270
271 A[ 0] = B[ 0] + B[ 1];
272 A[ 1] = MPC_MULTIPLY_V((B[ 0] - B[ 1]), INVCOS16);
273 A[ 2] = B[ 2] + B[ 3];
274 A[ 3] = MPC_MULTIPLY_V((B[ 2] - B[ 3]), INVCOS16);
275 A[ 4] = B[ 4] + B[ 5];
276 A[ 5] = MPC_MULTIPLY_V((B[ 4] - B[ 5]), INVCOS16);
277 A[ 6] = B[ 6] + B[ 7];
278 A[ 7] = MPC_MULTIPLY_V((B[ 6] - B[ 7]), INVCOS16);
279 A[ 8] = B[ 8] + B[ 9];
280 A[ 9] = MPC_MULTIPLY_V((B[ 8] - B[ 9]), INVCOS16);
281 A[10] = B[10] + B[11];
282 A[11] = MPC_MULTIPLY_V((B[10] - B[11]), INVCOS16);
283 A[12] = B[12] + B[13];
284 A[13] = MPC_MULTIPLY_V((B[12] - B[13]), INVCOS16);
285 A[14] = B[14] + B[15];
286 A[15] = MPC_MULTIPLY_V((B[14] - B[15]), INVCOS16);
287 // 8 adds, 8 subs, 8 muls, 8 shifts
288
289 // multiple used expressions: -(A[12] + A[14] + A[15])
290 V[48] = -A[ 0];
291 V[ 0] = A[ 1];
292 V[40] = -A[ 2] - (V[ 8] = A[ 3]);
293 V[36] = -((V[ 4] = A[ 5] + (V[12] = A[ 7])) + A[ 6]);
294 V[44] = - A[ 4] - A[ 6] - A[ 7];
295 V[ 6] = (V[10] = A[11] + (V[14] = A[15])) + A[13];
296 V[38] = (V[34] = -(V[ 2] = A[ 9] + A[13] + A[15]) - A[14]) + A[ 9] - A[10] - A[11];
297 V[46] = (tmp = -(A[12] + A[14] + A[15])) - A[ 8];
298 V[42] = tmp - A[10] - A[11];
299 // 9 adds, 9 subs
300
301 A[ 0] = MPC_MULTIPLY_V_PRESCALE((Sample[ 0] - Sample[31]), INVCOS01);
302 A[ 1] = MPC_MULTIPLY_V_PRESCALE((Sample[ 1] - Sample[30]), INVCOS03);
303 A[ 2] = MPC_MULTIPLY_V_PRESCALE((Sample[ 2] - Sample[29]), INVCOS05);
304 A[ 3] = MPC_MULTIPLY_V_PRESCALE((Sample[ 3] - Sample[28]), INVCOS07);
305 A[ 4] = MPC_MULTIPLY_V_PRESCALE((Sample[ 4] - Sample[27]), INVCOS09);
306 A[ 5] = MPC_MULTIPLY_V_PRESCALE((Sample[ 5] - Sample[26]), INVCOS11);
307 A[ 6] = MPC_MULTIPLY_V_PRESCALE((Sample[ 6] - Sample[25]), INVCOS13);
308 A[ 7] = MPC_MULTIPLY_V_PRESCALE((Sample[ 7] - Sample[24]), INVCOS15);
309 A[ 8] = MPC_MULTIPLY_V_PRESCALE((Sample[ 8] - Sample[23]), INVCOS17);
310 A[ 9] = MPC_MULTIPLY_V_PRESCALE((Sample[ 9] - Sample[22]), INVCOS19);
311 A[10] = MPC_MULTIPLY_V_PRESCALE((Sample[10] - Sample[21]), INVCOS21);
312 A[11] = MPC_MULTIPLY_V_PRESCALE((Sample[11] - Sample[20]), INVCOS23);
313 A[12] = MPC_MULTIPLY_V_PRESCALE((Sample[12] - Sample[19]), INVCOS25);
314 A[13] = MPC_MULTIPLY_V_PRESCALE((Sample[13] - Sample[18]), INVCOS27);
315 A[14] = MPC_MULTIPLY_V_PRESCALE((Sample[14] - Sample[17]), INVCOS29);
316 A[15] = MPC_MULTIPLY_V_PRESCALE((Sample[15] - Sample[16]), INVCOS31);
317 // 16 subs, 16 muls, 16 shifts
318
319 B[ 0] = A[ 0] + A[15];
320 B[ 1] = A[ 1] + A[14];
321 B[ 2] = A[ 2] + A[13];
322 B[ 3] = A[ 3] + A[12];
323 B[ 4] = A[ 4] + A[11];
324 B[ 5] = A[ 5] + A[10];
325 B[ 6] = A[ 6] + A[ 9];
326 B[ 7] = A[ 7] + A[ 8];
327 B[ 8] = MPC_MULTIPLY_V((A[ 0] - A[15]), INVCOS02);
328 B[ 9] = MPC_MULTIPLY_V((A[ 1] - A[14]), INVCOS06);
329 B[10] = MPC_MULTIPLY_V((A[ 2] - A[13]), INVCOS10);
330 B[11] = MPC_MULTIPLY_V((A[ 3] - A[12]), INVCOS14);
331 B[12] = MPC_MULTIPLY_V((A[ 4] - A[11]), INVCOS18);
332 B[13] = MPC_MULTIPLY_V((A[ 5] - A[10]), INVCOS22);
333 B[14] = MPC_MULTIPLY_V((A[ 6] - A[ 9]), INVCOS26);
334 B[15] = MPC_MULTIPLY_V((A[ 7] - A[ 8]), INVCOS30);
335 // 8 adds, 8 subs, 8 muls, 8 shift
336
337 A[ 0] = B[ 0] + B[ 7];
338 A[ 1] = B[ 1] + B[ 6];
339 A[ 2] = B[ 2] + B[ 5];
340 A[ 3] = B[ 3] + B[ 4];
341 A[ 4] = MPC_MULTIPLY_V((B[ 0] - B[ 7]), INVCOS04);
342 A[ 5] = MPC_MULTIPLY_V((B[ 1] - B[ 6]), INVCOS12);
343 A[ 6] = MPC_MULTIPLY_V((B[ 2] - B[ 5]), INVCOS20);
344 A[ 7] = MPC_MULTIPLY_V((B[ 3] - B[ 4]), INVCOS28);
345 A[ 8] = B[ 8] + B[15];
346 A[ 9] = B[ 9] + B[14];
347 A[10] = B[10] + B[13];
348 A[11] = B[11] + B[12];
349 A[12] = MPC_MULTIPLY_V((B[ 8] - B[15]), INVCOS04);
350 A[13] = MPC_MULTIPLY_V((B[ 9] - B[14]), INVCOS12);
351 A[14] = MPC_MULTIPLY_V((B[10] - B[13]), INVCOS20);
352 A[15] = MPC_MULTIPLY_V((B[11] - B[12]), INVCOS28);
353 // 8 adds, 8 subs, 8 muls, 8 shift
354
355 B[ 0] = A[ 0] + A[ 3];
356 B[ 1] = A[ 1] + A[ 2];
357 B[ 2] = MPC_MULTIPLY_V((A[ 0] - A[ 3]), INVCOS08);
358 B[ 3] = MPC_MULTIPLY_V((A[ 1] - A[ 2]), INVCOS24);
359 B[ 4] = A[ 4] + A[ 7];
360 B[ 5] = A[ 5] + A[ 6];
361 B[ 6] = MPC_MULTIPLY_V((A[ 4] - A[ 7]), INVCOS08);
362 B[ 7] = MPC_MULTIPLY_V((A[ 5] - A[ 6]), INVCOS24);
363 B[ 8] = A[ 8] + A[11];
364 B[ 9] = A[ 9] + A[10];
365 B[10] = MPC_MULTIPLY_V((A[ 8] - A[11]), INVCOS08);
366 B[11] = MPC_MULTIPLY_V((A[ 9] - A[10]), INVCOS24);
367 B[12] = A[12] + A[15];
368 B[13] = A[13] + A[14];
369 B[14] = MPC_MULTIPLY_V((A[12] - A[15]), INVCOS08);
370 B[15] = MPC_MULTIPLY_V((A[13] - A[14]), INVCOS24);
371 // 8 adds, 8 subs, 8 muls, 8 shift
372
373 A[ 0] = MPC_V_POSTSCALE((B[ 0] + B[ 1]));
374 A[ 1] = MPC_MULTIPLY_V_POSTSCALE((B[ 0] - B[ 1]), INVCOS16);
375 A[ 2] = MPC_V_POSTSCALE((B[ 2] + B[ 3]));
376 A[ 3] = MPC_MULTIPLY_V_POSTSCALE((B[ 2] - B[ 3]), INVCOS16);
377 A[ 4] = MPC_V_POSTSCALE((B[ 4] + B[ 5]));
378 A[ 5] = MPC_MULTIPLY_V_POSTSCALE((B[ 4] - B[ 5]), INVCOS16);
379 A[ 6] = MPC_V_POSTSCALE((B[ 6] + B[ 7]));
380 A[ 7] = MPC_MULTIPLY_V_POSTSCALE((B[ 6] - B[ 7]), INVCOS16);
381 A[ 8] = MPC_V_POSTSCALE((B[ 8] + B[ 9]));
382 A[ 9] = MPC_MULTIPLY_V_POSTSCALE((B[ 8] - B[ 9]), INVCOS16);
383 A[10] = MPC_V_POSTSCALE((B[10] + B[11]));
384 A[11] = MPC_MULTIPLY_V_POSTSCALE((B[10] - B[11]), INVCOS16);
385 A[12] = MPC_V_POSTSCALE((B[12] + B[13]));
386 A[13] = MPC_MULTIPLY_V_POSTSCALE((B[12] - B[13]), INVCOS16);
387 A[14] = MPC_V_POSTSCALE((B[14] + B[15]));
388 A[15] = MPC_MULTIPLY_V_POSTSCALE((B[14] - B[15]), INVCOS16);
389 // 8 adds, 8 subs, 8 muls, 8 shift
390
391 // multiple used expressions: A[ 4]+A[ 6]+A[ 7], A[ 9]+A[13]+A[15]
392 V[ 5] = (V[11] = (V[13] = A[ 7] + (V[15] = A[15])) + A[11]) + A[ 5] + A[13];
393 V[ 7] = (V[ 9] = A[ 3] + A[11] + A[15]) + A[13];
394 V[33] = -(V[ 1] = A[ 1] + A[ 9] + A[13] + A[15]) - A[14];
395 V[35] = -(V[ 3] = A[ 5] + A[ 7] + A[ 9] + A[13] + A[15]) - A[ 6] - A[14];
396 V[37] = (tmp = -(A[10] + A[11] + A[13] + A[14] + A[15])) - A[ 5] - A[ 6] - A[ 7];
397 V[39] = tmp - A[ 2] - A[ 3];
398 V[41] = (tmp += A[13] - A[12]) - A[ 2] - A[ 3];
399 V[43] = tmp - A[ 4] - A[ 6] - A[ 7];
400 V[47] = (tmp = -(A[ 8] + A[12] + A[14] + A[15])) - A[ 0];
401 V[45] = tmp - A[ 4] - A[ 6] - A[ 7];
402 // 22 adds, 18 subs
403
404 V[32] = -(V[ 0] = MPC_V_PRESHIFT(V[ 0]));
405 V[31] = -(V[ 1] = MPC_V_PRESHIFT(V[ 1]));
406 V[30] = -(V[ 2] = MPC_V_PRESHIFT(V[ 2]));
407 V[29] = -(V[ 3] = MPC_V_PRESHIFT(V[ 3]));
408 V[28] = -(V[ 4] = MPC_V_PRESHIFT(V[ 4]));
409 V[27] = -(V[ 5] = MPC_V_PRESHIFT(V[ 5]));
410 V[26] = -(V[ 6] = MPC_V_PRESHIFT(V[ 6]));
411 V[25] = -(V[ 7] = MPC_V_PRESHIFT(V[ 7]));
412 V[24] = -(V[ 8] = MPC_V_PRESHIFT(V[ 8]));
413 V[23] = -(V[ 9] = MPC_V_PRESHIFT(V[ 9]));
414 V[22] = -(V[10] = MPC_V_PRESHIFT(V[10]));
415 V[21] = -(V[11] = MPC_V_PRESHIFT(V[11]));
416 V[20] = -(V[12] = MPC_V_PRESHIFT(V[12]));
417 V[19] = -(V[13] = MPC_V_PRESHIFT(V[13]));
418 V[18] = -(V[14] = MPC_V_PRESHIFT(V[14]));
419 V[17] = -(V[15] = MPC_V_PRESHIFT(V[15]));
420 // 16 adds, 16 shifts (OPTIMIZE_FOR_SPEED only)
421
422 V[63] = (V[33] = MPC_V_PRESHIFT(V[33]));
423 V[62] = (V[34] = MPC_V_PRESHIFT(V[34]));
424 V[61] = (V[35] = MPC_V_PRESHIFT(V[35]));
425 V[60] = (V[36] = MPC_V_PRESHIFT(V[36]));
426 V[59] = (V[37] = MPC_V_PRESHIFT(V[37]));
427 V[58] = (V[38] = MPC_V_PRESHIFT(V[38]));
428 V[57] = (V[39] = MPC_V_PRESHIFT(V[39]));
429 V[56] = (V[40] = MPC_V_PRESHIFT(V[40]));
430 V[55] = (V[41] = MPC_V_PRESHIFT(V[41]));
431 V[54] = (V[42] = MPC_V_PRESHIFT(V[42]));
432 V[53] = (V[43] = MPC_V_PRESHIFT(V[43]));
433 V[52] = (V[44] = MPC_V_PRESHIFT(V[44]));
434 V[51] = (V[45] = MPC_V_PRESHIFT(V[45]));
435 V[50] = (V[46] = MPC_V_PRESHIFT(V[46]));
436 V[49] = (V[47] = MPC_V_PRESHIFT(V[47]));
437 V[48] = (V[48] = MPC_V_PRESHIFT(V[48]));
438 // 16 adds, 16 shifts (OPTIMIZE_FOR_SPEED only)
439
440 // OPTIMIZE_FOR_SPEED total: 143 adds, 107 subs, 80 muls, 112 shifts
441 // total: 111 adds, 107 subs, 80 muls, 80 shifts
442 }
443
444 #if defined(CPU_ARM)
445 extern void
446 mpc_decoder_windowing_D(MPC_SAMPLE_FORMAT * Data,
447 const MPC_SAMPLE_FORMAT * V,
448 const MPC_SAMPLE_FORMAT * D);
449 #else
450 static void
451 mpc_decoder_windowing_D(MPC_SAMPLE_FORMAT * Data,
452 const MPC_SAMPLE_FORMAT * V,
453 const MPC_SAMPLE_FORMAT * D)
454 {
455 mpc_int32_t k;
456
457 #if defined(OPTIMIZE_FOR_SPEED)
458 // 32=32x32-multiply (FIXED_POINT)
459 for ( k = 0; k < 32; k++, D += 16, V++ )
460 {
461 *Data = V[ 0]*D[ 0] + V[ 96]*D[ 1] + V[128]*D[ 2] + V[224]*D[ 3]
462 + V[256]*D[ 4] + V[352]*D[ 5] + V[384]*D[ 6] + V[480]*D[ 7]
463 + V[512]*D[ 8] + V[608]*D[ 9] + V[640]*D[10] + V[736]*D[11]
464 + V[768]*D[12] + V[864]*D[13] + V[896]*D[14] + V[992]*D[15];
465 Data += 1;
466 // total: 32 * (16 muls, 15 adds)
467 }
468 #else
469 #if defined(CPU_COLDFIRE)
470 // 64=32x32-multiply assembler for Coldfire
471 for ( k = 0; k < 32; k++, D += 16, V++ )
472 {
473 asm volatile (
474 "movem.l (%[D]), %%d0-%%d3 \n\t"
475 "move.l (%[V]), %%a5 \n\t"
476 "mac.l %%d0, %%a5, (96*4, %[V]), %%a5, %%acc0 \n\t"
477 "mac.l %%d1, %%a5, (128*4, %[V]), %%a5, %%acc0\n\t"
478 "mac.l %%d2, %%a5, (224*4, %[V]), %%a5, %%acc0\n\t"
479 "mac.l %%d3, %%a5, (256*4, %[V]), %%a5, %%acc0\n\t"
480 "movem.l (4*4, %[D]), %%d0-%%d3 \n\t"
481 "mac.l %%d0, %%a5, (352*4, %[V]), %%a5, %%acc0\n\t"
482 "mac.l %%d1, %%a5, (384*4, %[V]), %%a5, %%acc0\n\t"
483 "mac.l %%d2, %%a5, (480*4, %[V]), %%a5, %%acc0\n\t"
484 "mac.l %%d3, %%a5, (512*4, %[V]), %%a5, %%acc0\n\t"
485 "movem.l (8*4, %[D]), %%d0-%%d3 \n\t"
486 "mac.l %%d0, %%a5, (608*4, %[V]), %%a5, %%acc0\n\t"
487 "mac.l %%d1, %%a5, (640*4, %[V]), %%a5, %%acc0\n\t"
488 "mac.l %%d2, %%a5, (736*4, %[V]), %%a5, %%acc0\n\t"
489 "mac.l %%d3, %%a5, (768*4, %[V]), %%a5, %%acc0\n\t"
490 "movem.l (12*4, %[D]), %%d0-%%d3 \n\t"
491 "mac.l %%d0, %%a5, (864*4, %[V]), %%a5, %%acc0\n\t"
492 "mac.l %%d1, %%a5, (896*4, %[V]), %%a5, %%acc0\n\t"
493 "mac.l %%d2, %%a5, (992*4, %[V]), %%a5, %%acc0\n\t"
494 "mac.l %%d3, %%a5, %%acc0 \n\t"
495 "movclr.l %%acc0, %%d0 \n\t"
496 "move.l %%d0, (%[Data])+ \n"
497 : [Data] "+a" (Data)
498 : [V] "a" (V), [D] "a" (D)
499 : "d0", "d1", "d2", "d3", "a5");
500 }
501 #else
502 // 64=64x64-multiply (FIXED_POINT) or float=float*float (!FIXED_POINT) in C
503 for ( k = 0; k < 32; k++, D += 16, V++ )
504 {
505 *Data = MPC_MULTIPLY_EX(V[ 0],D[ 0],31) + MPC_MULTIPLY_EX(V[ 96],D[ 1],31)
506 + MPC_MULTIPLY_EX(V[128],D[ 2],31) + MPC_MULTIPLY_EX(V[224],D[ 3],31)
507 + MPC_MULTIPLY_EX(V[256],D[ 4],31) + MPC_MULTIPLY_EX(V[352],D[ 5],31)
508 + MPC_MULTIPLY_EX(V[384],D[ 6],31) + MPC_MULTIPLY_EX(V[480],D[ 7],31)
509 + MPC_MULTIPLY_EX(V[512],D[ 8],31) + MPC_MULTIPLY_EX(V[608],D[ 9],31)
510 + MPC_MULTIPLY_EX(V[640],D[10],31) + MPC_MULTIPLY_EX(V[736],D[11],31)
511 + MPC_MULTIPLY_EX(V[768],D[12],31) + MPC_MULTIPLY_EX(V[864],D[13],31)
512 + MPC_MULTIPLY_EX(V[896],D[14],31) + MPC_MULTIPLY_EX(V[992],D[15],31);
513 Data += 1;
514 // total: 16 muls, 15 adds, 16 shifts
515 }
516 #endif
517 #endif
518 }
519 #endif /* CPU_ARM */
520
521 static void
522 mpc_full_synthesis_filter(MPC_SAMPLE_FORMAT *OutData, MPC_SAMPLE_FORMAT *V, const MPC_SAMPLE_FORMAT *Y)
523 {
524 mpc_uint32_t n;
525
526 if (NULL != OutData)
527 {
528 for ( n = 0; n < 36; n++, Y += 32, OutData += 32 )
529 {
530 V -= 64;
531 mpc_calculate_new_V ( Y, V );
532 mpc_decoder_windowing_D( OutData, V, Di_opt );
533 }
534 }
535 }
536
537 void
538 mpc_decoder_synthese_filter_float(mpc_decoder *d, MPC_SAMPLE_FORMAT *OutData)
539 {
540 /********* left channel ********/
541 memmove(d->V_L + MPC_V_MEM, d->V_L, 960 * sizeof(MPC_SAMPLE_FORMAT) );
542
543 mpc_full_synthesis_filter(
544 OutData,
545 (MPC_SAMPLE_FORMAT *)(d->V_L + MPC_V_MEM),
546 (MPC_SAMPLE_FORMAT *)(d->Y_L [0]));
547
548 /******** right channel ********/
549 memmove(d->V_R + MPC_V_MEM, d->V_R, 960 * sizeof(MPC_SAMPLE_FORMAT) );
550
551 mpc_full_synthesis_filter(
552 (OutData == NULL ? NULL : OutData + MPC_FRAME_LENGTH),
553 (MPC_SAMPLE_FORMAT *)(d->V_R + MPC_V_MEM),
554 (MPC_SAMPLE_FORMAT *)(d->Y_R [0]));
555 }
556
557 /*******************************************/
558 /* */
559 /* dithered synthesis */
560 /* */
561 /*******************************************/
562
563 static const unsigned char Parity [256] ICONST_ATTR = { // parity
564 0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
565 1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
566 1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
567 0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
568 1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,
569 0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
570 0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0,1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,
571 1,0,0,1,0,1,1,0,0,1,1,0,1,0,0,1,0,1,1,0,1,0,0,1,1,0,0,1,0,1,1,0
572 };
573
574 /*
575 * This is a simple random number generator with good quality for audio purposes.
576 * It consists of two polycounters with opposite rotation direction and different
577 * periods. The periods are coprime, so the total period is the product of both.
578 *
579 * -------------------------------------------------------------------------------------------------
580 * +-> |31:30:29:28:27:26:25:24:23:22:21:20:19:18:17:16:15:14:13:12:11:10: 9: 8: 7: 6: 5: 4: 3: 2: 1: 0|
581 * | -------------------------------------------------------------------------------------------------
582 * | | | | | | |
583 * | +--+--+--+-XOR-+--------+
584 * | |
585 * +--------------------------------------------------------------------------------------+
586 *
587 * -------------------------------------------------------------------------------------------------
588 * |31:30:29:28:27:26:25:24:23:22:21:20:19:18:17:16:15:14:13:12:11:10: 9: 8: 7: 6: 5: 4: 3: 2: 1: 0| <-+
589 * ------------------------------------------------------------------------------------------------- |
590 * | | | | |
591 * +--+----XOR----+--+ |
592 * | |
593 * +----------------------------------------------------------------------------------------+
594 *
595 *
596 * The first has an period of 3*5*17*257*65537, the second of 7*47*73*178481,
597 * which gives a period of 18.410.713.077.675.721.215. The result is the
598 * XORed values of both generators.
599 */
600 mpc_uint32_t
601 mpc_random_int(mpc_decoder *d)
602 {
603 #if 1
604 mpc_uint32_t t1, t2, t3, t4;
605
606 t3 = t1 = d->__r1; t4 = t2 = d->__r2; // Parity calculation is done via table lookup, this is also available
607 t1 &= 0xF5; t2 >>= 25; // on CPUs without parity, can be implemented in C and avoid unpredictable
608 t1 = Parity [t1]; t2 &= 0x63; // jumps and slow rotate through the carry flag operations.
609 t1 <<= 31; t2 = Parity [t2];
610
611 return (d->__r1 = (t3 >> 1) | t1 ) ^ (d->__r2 = (t4 + t4) | t2 );
612 #else
613 return (d->__r1 = (d->__r1 >> 1) | ((mpc_uint32_t)Parity [d->__r1 & 0xF5] << 31) ) ^
614 (d->__r2 = (d->__r2 << 1) | (mpc_uint32_t)Parity [(d->__r2 >> 25) & 0x63] );
615 #endif
616 }
Open Source Jukebox Firmware