1/* ------------------------------------------------------------------
2 * Copyright (C) 1998-2009 PacketVideo
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
13 * express or implied.
14 * See the License for the specific language governing permissions
15 * and limitations under the License.
16 * -------------------------------------------------------------------
17 */
18/*
19------------------------------------------------------------------------------
20
21   PacketVideo Corp.
22   MP3 Decoder Library
23
24   Filename: pvmp3_dct_16.cpp
25
26   Functions:
27    dct_16
28    pv_merge_in_place_N32
29    pv_split
30
31     Date: 09/21/2007
32
33------------------------------------------------------------------------------
34 REVISION HISTORY
35
36
37 Description:
38
39------------------------------------------------------------------------------
40 INPUT AND OUTPUT DEFINITIONS
41
42    dct_16
43
44Input
45    int32 vec[],        input vector length 16
46    Int flag            processing direction: forward (1), backward ( 0)
47 Returns
48
49    int32 vec[],        dct length 16
50
51------------------------------------------------------------------------------
52 INPUT AND OUTPUT DEFINITIONS
53
54    pv_merge_in_place_N32
55
56Input
57    int32 vec[],        input vector length 16
58
59 Returns
60
61    int32 vec[],        merged  output of two dct 16 to create a dct 32
62
63------------------------------------------------------------------------------
64 INPUT AND OUTPUT DEFINITIONS
65
66    pv_split
67
68Input
69    int32 vec[],        input vector length 16
70
71 Returns
72
73    int32 vec[],        splitted even/odd and pre processing rotation
74
75------------------------------------------------------------------------------
76 FUNCTION DESCRIPTION
77
78    dct 16 and tools to assemble a dct32 output
79
80------------------------------------------------------------------------------
81 REQUIREMENTS
82
83
84------------------------------------------------------------------------------
85 REFERENCES
86
87------------------------------------------------------------------------------
88 PSEUDO-CODE
89
90------------------------------------------------------------------------------
91*/
92
93#if ( !defined(PV_ARM_GCC_V5) && !defined(PV_ARM_GCC_V4) )
94/*----------------------------------------------------------------------------
95; INCLUDES
96----------------------------------------------------------------------------*/
97
98#include "pvmp3_dct_16.h"
99#include "pv_mp3dec_fxd_op.h"
100
101/*----------------------------------------------------------------------------
102; MACROS
103; Define module specific macros here
104----------------------------------------------------------------------------*/
105
106
107/*----------------------------------------------------------------------------
108; DEFINES
109; Include all pre-processor statements here. Include conditional
110; compile variables also.
111----------------------------------------------------------------------------*/
112#define Qfmt(a)   (int32)(a*((int32)1<<27))
113
114/*----------------------------------------------------------------------------
115; LOCAL FUNCTION DEFINITIONS
116; Function Prototype declaration
117----------------------------------------------------------------------------*/
118
119/*----------------------------------------------------------------------------
120; LOCAL STORE/BUFFER/POINTER DEFINITIONS
121; Variable declaration - defined here and used outside this module
122----------------------------------------------------------------------------*/
123const int32 CosTable_dct32[16] =
124{
125    Qfmt_31(0.50060299823520F) ,  Qfmt_31(0.50547095989754F) ,
126    Qfmt_31(0.51544730992262F) ,  Qfmt_31(0.53104259108978F) ,
127    Qfmt_31(0.55310389603444F) ,  Qfmt_31(0.58293496820613F) ,
128    Qfmt_31(0.62250412303566F) ,  Qfmt_31(0.67480834145501F) ,
129    Qfmt_31(0.74453627100230F) ,  Qfmt_31(0.83934964541553F) ,
130
131    Qfmt(0.97256823786196F) ,  Qfmt(1.16943993343288F) ,
132    Qfmt(1.48416461631417F) ,  Qfmt(2.05778100995341F) ,
133    Qfmt(3.40760841846872F) ,  Qfmt(10.19000812354803F)
134};
135
136
137/*----------------------------------------------------------------------------
138; EXTERNAL FUNCTION REFERENCES
139; Declare functions defined elsewhere and referenced in this module
140----------------------------------------------------------------------------*/
141
142/*----------------------------------------------------------------------------
143; EXTERNAL GLOBAL STORE/BUFFER/POINTER REFERENCES
144; Declare variables used in this module but defined elsewhere
145----------------------------------------------------------------------------*/
146
147
148/*----------------------------------------------------------------------------
149; FUNCTION CODE
150----------------------------------------------------------------------------*/
151
152void pvmp3_dct_16(int32 vec[], int32 flag)
153{
154    int32 tmp0;
155    int32 tmp1;
156    int32 tmp2;
157    int32 tmp3;
158    int32 tmp4;
159    int32 tmp5;
160    int32 tmp6;
161    int32 tmp7;
162    int32 tmp_o0;
163    int32 tmp_o1;
164    int32 tmp_o2;
165    int32 tmp_o3;
166    int32 tmp_o4;
167    int32 tmp_o5;
168    int32 tmp_o6;
169    int32 tmp_o7;
170    int32 itmp_e0;
171    int32 itmp_e1;
172    int32 itmp_e2;
173
174    /*  split input vector */
175
176    tmp_o0 = fxp_mul32_Q32((vec[ 0] - vec[15]), Qfmt_31(0.50241928618816F));
177    tmp0   =  vec[ 0] + vec[15];
178
179    tmp_o7 = fxp_mul32_Q32((vec[ 7] - vec[ 8]) << 3, Qfmt_31(0.63764357733614F));
180    tmp7   =  vec[ 7] + vec[ 8];
181
182    itmp_e0    = fxp_mul32_Q32((tmp0 - tmp7), Qfmt_31(0.50979557910416F));
183    tmp7 = (tmp0 + tmp7);
184
185    tmp_o1 = fxp_mul32_Q32((vec[ 1] - vec[14]), Qfmt_31(0.52249861493969F));
186    tmp1   =  vec[ 1] + vec[14];
187
188    tmp_o6 = fxp_mul32_Q32((vec[ 6] - vec[ 9]) << 1, Qfmt_31(0.86122354911916F));
189    tmp6   =  vec[ 6] + vec[ 9];
190
191
192
193    itmp_e1 = (tmp1 + tmp6);
194    tmp6    = fxp_mul32_Q32((tmp1 - tmp6), Qfmt_31(0.60134488693505F));
195
196
197
198    tmp_o2 = fxp_mul32_Q32((vec[ 2] - vec[13]), Qfmt_31(0.56694403481636F));
199    tmp2   =  vec[ 2] + vec[13];
200    tmp_o5 = fxp_mul32_Q32((vec[ 5] - vec[10]) << 1, Qfmt_31(0.53033884299517F));
201    tmp5   =  vec[ 5] + vec[10];
202
203    itmp_e2 = (tmp2 + tmp5);
204    tmp5    = fxp_mul32_Q32((tmp2 - tmp5), Qfmt_31(0.89997622313642F));
205
206    tmp_o3 = fxp_mul32_Q32((vec[ 3] - vec[12]), Qfmt_31(0.64682178335999F));
207    tmp3   =  vec[ 3] + vec[12];
208    tmp_o4 = fxp_mul32_Q32((vec[ 4] - vec[11]), Qfmt_31(0.78815462345125F));
209    tmp4   =  vec[ 4] + vec[11];
210
211    tmp1   = (tmp3 + tmp4);
212    tmp4   =  fxp_mul32_Q32((tmp3 - tmp4) << 2, Qfmt_31(0.64072886193538F));
213
214    /*  split even part of tmp_e */
215
216    tmp0 = (tmp7 + tmp1);
217    tmp1 = fxp_mul32_Q32((tmp7 - tmp1), Qfmt_31(0.54119610014620F));
218
219    tmp3 = fxp_mul32_Q32((itmp_e1 - itmp_e2) << 1, Qfmt_31(0.65328148243819F));
220    tmp7 = (itmp_e1 + itmp_e2);
221
222    vec[ 0]  = (tmp0 + tmp7) >> 1;
223    vec[ 8]  = fxp_mul32_Q32((tmp0 - tmp7), Qfmt_31(0.70710678118655F));
224    tmp0     = fxp_mul32_Q32((tmp1 - tmp3) << 1, Qfmt_31(0.70710678118655F));
225    vec[ 4]  =  tmp1 + tmp3 + tmp0;
226    vec[12]  =  tmp0;
227
228    /*  split odd part of tmp_e */
229
230    tmp1 = fxp_mul32_Q32((itmp_e0 - tmp4) << 1, Qfmt_31(0.54119610014620F));
231    tmp7 = itmp_e0 + tmp4;
232
233    tmp3  = fxp_mul32_Q32((tmp6 - tmp5) << 2, Qfmt_31(0.65328148243819F));
234    tmp6 += tmp5;
235
236    tmp4  = fxp_mul32_Q32((tmp7 - tmp6) << 1, Qfmt_31(0.70710678118655F));
237    tmp6 += tmp7;
238    tmp7  = fxp_mul32_Q32((tmp1 - tmp3) << 1, Qfmt_31(0.70710678118655F));
239
240    tmp1    +=  tmp3 + tmp7;
241    vec[ 2]  =  tmp1 + tmp6;
242    vec[ 6]  =  tmp1 + tmp4;
243    vec[10]  =  tmp7 + tmp4;
244    vec[14]  =  tmp7;
245
246
247    // dct8;
248
249    tmp1 = fxp_mul32_Q32((tmp_o0 - tmp_o7) << 1, Qfmt_31(0.50979557910416F));
250    tmp7 = tmp_o0 + tmp_o7;
251
252    tmp6   = tmp_o1 + tmp_o6;
253    tmp_o1 = fxp_mul32_Q32((tmp_o1 - tmp_o6) << 1, Qfmt_31(0.60134488693505F));
254
255    tmp5   = tmp_o2 + tmp_o5;
256    tmp_o5 = fxp_mul32_Q32((tmp_o2 - tmp_o5) << 1, Qfmt_31(0.89997622313642F));
257
258    tmp0 = fxp_mul32_Q32((tmp_o3 - tmp_o4) << 3, Qfmt_31(0.6407288619354F));
259    tmp4 = tmp_o3 + tmp_o4;
260
261    if (!flag)
262    {
263        tmp7   = -tmp7;
264        tmp1   = -tmp1;
265        tmp6   = -tmp6;
266        tmp_o1 = -tmp_o1;
267        tmp5   = -tmp5;
268        tmp_o5 = -tmp_o5;
269        tmp4   = -tmp4;
270        tmp0   = -tmp0;
271    }
272
273
274    tmp2     =  fxp_mul32_Q32((tmp1 -   tmp0) << 1, Qfmt_31(0.54119610014620F));
275    tmp0    +=  tmp1;
276    tmp1     =  fxp_mul32_Q32((tmp7 -   tmp4) << 1, Qfmt_31(0.54119610014620F));
277    tmp7    +=  tmp4;
278    tmp4     =  fxp_mul32_Q32((tmp6 -   tmp5) << 2, Qfmt_31(0.65328148243819F));
279    tmp6    +=  tmp5;
280    tmp5     =  fxp_mul32_Q32((tmp_o1 - tmp_o5) << 2, Qfmt_31(0.65328148243819F));
281    tmp_o1  += tmp_o5;
282
283
284    vec[13]  =  fxp_mul32_Q32((tmp1 -   tmp4) << 1, Qfmt_31(0.70710678118655F));
285    vec[ 5]  =  tmp1 + tmp4 + vec[13];
286
287    vec[ 9]  =  fxp_mul32_Q32((tmp7 -   tmp6) << 1, Qfmt_31(0.70710678118655F));
288    vec[ 1]  =  tmp7 + tmp6;
289
290    tmp4     =  fxp_mul32_Q32((tmp0 - tmp_o1) << 1, Qfmt_31(0.70710678118655F));
291    tmp0    +=  tmp_o1;
292    tmp6     =  fxp_mul32_Q32((tmp2 -   tmp5) << 1, Qfmt_31(0.70710678118655F));
293    tmp2    +=  tmp5 + tmp6;
294    tmp0    +=  tmp2;
295
296    vec[ 1] += tmp0;
297    vec[ 3]  = tmp0 + vec[ 5];
298    tmp2    += tmp4;
299    vec[ 5]  = tmp2 + vec[ 5];
300    vec[ 7]  = tmp2 + vec[ 9];
301    tmp4    += tmp6;
302    vec[ 9]  = tmp4 + vec[ 9];
303    vec[11]  = tmp4 + vec[13];
304    vec[13]  = tmp6 + vec[13];
305    vec[15]  = tmp6;
306
307}
308/*----------------------------------------------------------------------------
309; FUNCTION CODE
310----------------------------------------------------------------------------*/
311void pvmp3_merge_in_place_N32(int32 vec[])
312{
313
314
315    int32 temp0;
316    int32 temp1;
317    int32 temp2;
318    int32 temp3;
319
320    temp0   = vec[14];
321    vec[14] = vec[ 7];
322    temp1   = vec[12];
323    vec[12] = vec[ 6];
324    temp2   = vec[10];
325    vec[10] = vec[ 5];
326    temp3   = vec[ 8];
327    vec[ 8] = vec[ 4];
328    vec[ 6] = vec[ 3];
329    vec[ 4] = vec[ 2];
330    vec[ 2] = vec[ 1];
331
332    vec[ 1] = (vec[16] + vec[17]);
333    vec[16] = temp3;
334    vec[ 3] = (vec[18] + vec[17]);
335    vec[ 5] = (vec[19] + vec[18]);
336    vec[18] = vec[9];
337
338    vec[ 7] = (vec[20] + vec[19]);
339    vec[ 9] = (vec[21] + vec[20]);
340    vec[20] = temp2;
341    temp2   = vec[13];
342    temp3   = vec[11];
343    vec[11] = (vec[22] + vec[21]);
344    vec[13] = (vec[23] + vec[22]);
345    vec[22] = temp3;
346    temp3   = vec[15];
347
348    vec[15] = (vec[24] + vec[23]);
349    vec[17] = (vec[25] + vec[24]);
350    vec[19] = (vec[26] + vec[25]);
351    vec[21] = (vec[27] + vec[26]);
352    vec[23] = (vec[28] + vec[27]);
353    vec[24] = temp1;
354    vec[25] = (vec[29] + vec[28]);
355    vec[26] = temp2;
356    vec[27] = (vec[30] + vec[29]);
357    vec[28] = temp0;
358    vec[29] = (vec[30] + vec[31]);
359    vec[30] = temp3;
360}
361
362
363/*----------------------------------------------------------------------------
364; FUNCTION CODE
365----------------------------------------------------------------------------*/
366
367
368
369void pvmp3_split(int32 *vect)
370{
371
372    int32 i;
373    const int32 *pt_cosTerms = &CosTable_dct32[15];
374    int32 *pt_vect   = vect;
375    int32 *pt_vect_2 = pt_vect - 1;
376
377    for (i = 3; i != 0; i--)
378    {
379        int32 tmp2 = *(pt_vect);
380        int32 tmp1 = *(pt_vect_2);
381        int32 cosx = *(pt_cosTerms--);
382        *(pt_vect_2--) = (tmp1  + tmp2);
383        *(pt_vect++)   = fxp_mul32_Q27((tmp1 - tmp2), cosx);
384
385        tmp2 = *(pt_vect);
386        tmp1 = *(pt_vect_2);
387        cosx = *(pt_cosTerms--);
388        *(pt_vect_2--) = (tmp1  + tmp2);
389        *(pt_vect++)   = fxp_mul32_Q27((tmp1 - tmp2), cosx);
390
391    }
392
393    for (i = 5; i != 0; i--)
394    {
395        int32 tmp2 = *(pt_vect);
396        int32 tmp1 = *(pt_vect_2);
397        int32 cosx = *(pt_cosTerms--);
398        *(pt_vect_2--) = (tmp1  + tmp2);
399        *(pt_vect++) = fxp_mul32_Q32((tmp1 - tmp2) << 1, cosx);
400
401        tmp2 = *(pt_vect);
402        tmp1 = *(pt_vect_2);
403        cosx = *(pt_cosTerms--);
404        *(pt_vect_2--) = (tmp1  + tmp2);
405        *(pt_vect++) = fxp_mul32_Q32((tmp1 - tmp2) << 1, cosx);
406    }
407
408}
409
410#endif
411