1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
|
/******************************************************************************
*
* Copyright (C) 2015 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*****************************************************************************
* Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
*/
/**
*******************************************************************************
* @file
* ih264_iquant_itrans_recon_sse42.c
*
* @brief
* Contains function definitions for inverse quantization, inverse
* transform and reconstruction
*
* @author
* Mohit [100664]
*
* @par List of Functions:
* - ih264_iquant_itrans_recon_4x4_sse42()
* - ih264_iquant_itrans_recon_chroma_4x4_sse42()
*
* @remarks
* None
*
*******************************************************************************
*/
/* User include files */
#include "ih264_typedefs.h"
#include "ih264_defs.h"
#include "ih264_trans_macros.h"
#include "ih264_macros.h"
#include "ih264_platform_macros.h"
#include "ih264_trans_data.h"
#include "ih264_size_defs.h"
#include "ih264_structs.h"
#include "ih264_trans_quant_itrans_iquant.h"
#include <immintrin.h>
/*
********************************************************************************
*
* @brief This function reconstructs a 4x4 sub block from quantized resiude and
* prediction buffer
*
* @par Description:
* The quantized residue is first inverse quantized, then inverse transformed.
* This inverse transformed content is added to the prediction buffer to recon-
* struct the end output
*
* @param[in] pi2_src
* quantized 4x4 block
*
* @param[in] pu1_pred
* prediction 4x4 block
*
* @param[out] pu1_out
* reconstructed 4x4 block
*
* @param[in] src_strd
* quantization buffer stride
*
* @param[in] pred_strd,
* Prediction buffer stride
*
* @param[in] out_strd
* recon buffer Stride
*
* @param[in] pu2_scaling_list
* pointer to scaling list
*
* @param[in] pu2_norm_adjust
* pointer to inverse scale matrix
*
* @param[in] u4_qp_div_6
* Floor (qp/6)
*
* @param[in] pi4_tmp
* temporary buffer of size 1*16
*
* @returns none
*
* @remarks none
*
*******************************************************************************
*/
void ih264_iquant_itrans_recon_4x4_sse42(WORD16 *pi2_src,
UWORD8 *pu1_pred,
UWORD8 *pu1_out,
WORD32 pred_strd,
WORD32 out_strd,
const UWORD16 *pu2_iscal_mat,
const UWORD16 *pu2_weigh_mat,
UWORD32 u4_qp_div_6,
WORD16 *pi2_tmp,
WORD32 iq_start_idx,
WORD16 *pi2_dc_ld_addr)
{
UWORD32 *pu4_out = (UWORD32 *) pu1_out;
__m128i src_r0_r1, src_r2_r3;
__m128i src_r0, src_r1, src_r2, src_r3;
__m128i scalemat_r0_r1, scalemat_r2_r3;
__m128i pred_r0, pred_r1, pred_r2, pred_r3;
__m128i sign_reg, dequant_r0_r1, dequant_r2_r3;
__m128i zero_8x16b = _mm_setzero_si128(); // all bits reset to zero
__m128i temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
__m128i resq_r0, resq_r1, resq_r2, resq_r3;
__m128i add_rshift = _mm_set1_epi32((1 << (3 - u4_qp_div_6)));
__m128i value_32 = _mm_set1_epi32(32);
UNUSED (pi2_tmp);
/*************************************************************/
/* Dequantization of coefficients. Will be replaced by SIMD */
/* operations on platform */
/*************************************************************/
src_r0_r1 = _mm_loadu_si128((__m128i *) (pi2_src)); //a00 a01 a02 a03 a10 a11 a12 a13 -- the source matrix 0th,1st row
src_r2_r3 = _mm_loadu_si128((__m128i *) (pi2_src + 8)); //a20 a21 a22 a23 a30 a31 a32 a33 -- the source matrix 2nd,3rd row
scalemat_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat)); //b00 b01 b02 b03 b10 b11 b12 b13 -- the scaling matrix 0th,1st row
scalemat_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat + 8)); //b20 b21 b22 b23 b30 b31 b32 b33 -- the scaling matrix 2nd,3rd row
dequant_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat)); //q00 q01 q02 q03 q10 q11 q12 q13 -- all 16 bits
dequant_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat + 8)); //q20 q21 q22 q23 q30 q31 q32 q33 -- all 16 bits
temp0 = _mm_mullo_epi16(scalemat_r0_r1, dequant_r0_r1); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result
temp1 = _mm_mullo_epi16(scalemat_r2_r3, dequant_r2_r3); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result
temp4 = _mm_unpacklo_epi16(temp0, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long
temp5 = _mm_unpackhi_epi16(temp0, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long
temp6 = _mm_unpacklo_epi16(temp1, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long
temp7 = _mm_unpackhi_epi16(temp1, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long
src_r0 = _mm_unpacklo_epi16(src_r0_r1, zero_8x16b); // a00 0 a01 0 a02 0 a03 0 -- 16 bit long
src_r1 = _mm_unpackhi_epi16(src_r0_r1, zero_8x16b); // a10 0 a11 0 a12 0 a13 0 -- 16 bit long
src_r2 = _mm_unpacklo_epi16(src_r2_r3, zero_8x16b); // a20 0 a21 0 a22 0 a23 0 -- 16 bit long
src_r3 = _mm_unpackhi_epi16(src_r2_r3, zero_8x16b); // a30 0 a31 0 a32 0 a33 0 -- 16 bit long
temp4 = _mm_madd_epi16(src_r0, temp4); //a00*b00*q00 a10*b10*q10 a20*b20*q20 a30*b30 q30 -- 32 bits long
temp5 = _mm_madd_epi16(src_r1, temp5);
temp6 = _mm_madd_epi16(src_r2, temp6);
temp7 = _mm_madd_epi16(src_r3, temp7);
if (u4_qp_div_6 >= 4) {
resq_r0 = _mm_slli_epi32(temp4, u4_qp_div_6 - 4);
resq_r1 = _mm_slli_epi32(temp5, u4_qp_div_6 - 4);
resq_r2 = _mm_slli_epi32(temp6, u4_qp_div_6 - 4);
resq_r3 = _mm_slli_epi32(temp7, u4_qp_div_6 - 4);
} else {
temp4 = _mm_add_epi32(temp4, add_rshift);
temp5 = _mm_add_epi32(temp5, add_rshift);
temp6 = _mm_add_epi32(temp6, add_rshift);
temp7 = _mm_add_epi32(temp7, add_rshift);
resq_r0 = _mm_srai_epi32(temp4, 4 - u4_qp_div_6);
resq_r1 = _mm_srai_epi32(temp5, 4 - u4_qp_div_6);
resq_r2 = _mm_srai_epi32(temp6, 4 - u4_qp_div_6);
resq_r3 = _mm_srai_epi32(temp7, 4 - u4_qp_div_6);
}
if (iq_start_idx == 1)
resq_r0 = _mm_insert_epi32(resq_r0,(WORD32)pi2_dc_ld_addr[0],0);
/* Perform Inverse transform */
/*-------------------------------------------------------------*/
/* IDCT [ Horizontal transformation ] */
/*-------------------------------------------------------------*/
// Matrix transpose
/*
* a0 a1 a2 a3
* b0 b1 b2 b3
* c0 c1 c2 c3
* d0 d1 d2 d3
*/
temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 b0 a1 b1
temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //c0 d0 c1 d1
temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //a2 b2 a3 b3
temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 d2 c3 d3
resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 b0 c0 d0
resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //a1 b1 c1 d1
resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //a2 b2 c2 d2
resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //a3 b3 c3 d3
//Transform starts -- horizontal transform
/*------------------------------------------------------------------*/
/* z0 = w0 + w2 */
temp0 = _mm_add_epi32(resq_r0, resq_r2);
/* z1 = w0 - w2 */
temp1 = _mm_sub_epi32(resq_r0, resq_r2);
/* z2 = (w1 >> 1) - w3 */
temp2 = _mm_srai_epi32(resq_r1, 1); //(w1>>1)
temp2 = _mm_sub_epi32(temp2, resq_r3); //(w1>>1) - w3
/* z3 = w1 + (w3 >> 1) */
temp3 = _mm_srai_epi32(resq_r3, 1); //(w3>>1) + w1
temp3 = _mm_add_epi32(temp3, resq_r1);
/*----------------------------------------------------------*/
/* x0 = z0 + z3 */
resq_r0 = _mm_add_epi32(temp0, temp3);
/* x1 = z1 + z2 */
resq_r1 = _mm_add_epi32(temp1, temp2);
/* x2 = z1 - z2 */
resq_r2 = _mm_sub_epi32(temp1, temp2);
/* x3 = z0 - z3 */
resq_r3 = _mm_sub_epi32(temp0, temp3);
// Matrix transpose
/*
* a0 b0 c0 d0
* a1 b1 c1 d1
* a2 b2 c2 d2
* a3 b3 c3 d3
*/
temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 a1 b0 b1
temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //a2 a3 b2 b3
temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //c0 c1 d0 d1
temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 c3 d2 d3
resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 a1 a2 a3
resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //b0 b1 b2 b3
resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //c0 c1 c2 c3
resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //d0 d1 d2 d3
//Transform ends -- horizontal transform
//Load pred buffer
pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); //p00 p01 p02 p03 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[pred_strd])); //p10 p11 p12 p13 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * pred_strd])); //p20 p21 p22 p23 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * pred_strd])); //p30 p31 p32 p33 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r0 = _mm_cvtepu8_epi32(pred_r0); //p00 p01 p02 p03 -- all 32 bits
pred_r1 = _mm_cvtepu8_epi32(pred_r1); //p10 p11 p12 p13 -- all 32 bits
pred_r2 = _mm_cvtepu8_epi32(pred_r2); //p20 p21 p22 p23 -- all 32 bits
pred_r3 = _mm_cvtepu8_epi32(pred_r3); //p30 p31 p32 p33 -- all 32 bits
/*--------------------------------------------------------------*/
/* IDCT [ Vertical transformation] and Xij = (xij + 32)>>6 */
/* */
/* Add the prediction and store it back to same buffer */
/*--------------------------------------------------------------*/
/* z0j = y0j + y2j */
temp0 = _mm_add_epi32(resq_r0, resq_r2);
/* z1j = y0j - y2j */
temp1 = _mm_sub_epi32(resq_r0, resq_r2);
/* z2j = (y1j>>1) - y3j */
temp2 = _mm_srai_epi32(resq_r1, 1); //(y1j>>1)
temp2 = _mm_sub_epi32(temp2, resq_r3);
/* z3j = y1j + (y3j>>1) */
temp3 = _mm_srai_epi32(resq_r3, 1); //(y3j>>1)
temp3 = _mm_add_epi32(temp3, resq_r1);
/* x0j = z0j + z3j */
temp4 = _mm_add_epi32(temp0, temp3);
temp4 = _mm_add_epi32(temp4, value_32);
temp4 = _mm_srai_epi32(temp4, 6);
temp4 = _mm_add_epi32(temp4, pred_r0);
/* x1j = z1j + z2j */
temp5 = _mm_add_epi32(temp1, temp2);
temp5 = _mm_add_epi32(temp5, value_32);
temp5 = _mm_srai_epi32(temp5, 6);
temp5 = _mm_add_epi32(temp5, pred_r1);
/* x2j = z1j - z2j */
temp6 = _mm_sub_epi32(temp1, temp2);
temp6 = _mm_add_epi32(temp6, value_32);
temp6 = _mm_srai_epi32(temp6, 6);
temp6 = _mm_add_epi32(temp6, pred_r2);
/* x3j = z0j - z3j */
temp7 = _mm_sub_epi32(temp0, temp3);
temp7 = _mm_add_epi32(temp7, value_32);
temp7 = _mm_srai_epi32(temp7, 6);
temp7 = _mm_add_epi32(temp7, pred_r3);
// 32-bit to 16-bit conversion
temp0 = _mm_packs_epi32(temp4, temp5);
temp1 = _mm_packs_epi32(temp6, temp7);
/*------------------------------------------------------------------*/
//Clipping the results to 8 bits
sign_reg = _mm_cmpgt_epi16(temp0, zero_8x16b); // sign check
temp0 = _mm_and_si128(temp0, sign_reg);
sign_reg = _mm_cmpgt_epi16(temp1, zero_8x16b);
temp1 = _mm_and_si128(temp1, sign_reg);
resq_r0 = _mm_packus_epi16(temp0, temp1);
resq_r1 = _mm_srli_si128(resq_r0, 4);
resq_r2 = _mm_srli_si128(resq_r1, 4);
resq_r3 = _mm_srli_si128(resq_r2, 4);
*pu4_out = _mm_cvtsi128_si32(resq_r0);
pu1_out += out_strd;
pu4_out = (UWORD32 *) (pu1_out);
*(pu4_out) = _mm_cvtsi128_si32(resq_r1);
pu1_out += out_strd;
pu4_out = (UWORD32 *) (pu1_out);
*(pu4_out) = _mm_cvtsi128_si32(resq_r2);
pu1_out += out_strd;
pu4_out = (UWORD32 *) (pu1_out);
*(pu4_out) = _mm_cvtsi128_si32(resq_r3);
}
/*
********************************************************************************
*
* @brief This function reconstructs a 4x4 sub block from quantized chroma resiude and
* prediction buffer
*
* @par Description:
* The quantized residue is first inverse quantized, then inverse transformed.
* This inverse transformed content is added to the prediction buffer to recon-
* struct the end output
*
* @param[in] pi2_src
* quantized 4x4 block
*
* @param[in] pu1_pred
* prediction 4x4 block
*
* @param[out] pu1_out
* reconstructed 4x4 block
*
* @param[in] src_strd
* quantization buffer stride
*
* @param[in] pred_strd,
* Prediction buffer stride
*
* @param[in] out_strd
* recon buffer Stride
*
* @param[in] pu2_scaling_list
* pointer to scaling list
*
* @param[in] pu2_norm_adjust
* pointer to inverse scale matrix
*
* @param[in] u4_qp_div_6
* Floor (qp/6)
*
* @param[in] pi4_tmp
* temporary buffer of size 1*16
*
* @returns none
*
* @remarks none
*
*******************************************************************************
*/
void ih264_iquant_itrans_recon_chroma_4x4_sse42(WORD16 *pi2_src,
UWORD8 *pu1_pred,
UWORD8 *pu1_out,
WORD32 pred_strd,
WORD32 out_strd,
const UWORD16 *pu2_iscal_mat,
const UWORD16 *pu2_weigh_mat,
UWORD32 u4_qp_div_6,
WORD16 *pi2_tmp,
WORD16 *pi2_dc_ld_addr)
{
__m128i src_r0_r1, src_r2_r3;
__m128i src_r0, src_r1, src_r2, src_r3;
__m128i scalemat_r0_r1, scalemat_r2_r3;
__m128i pred_r0, pred_r1, pred_r2, pred_r3;
__m128i sign_reg, dequant_r0_r1, dequant_r2_r3;
__m128i zero_8x16b = _mm_setzero_si128(); // all bits reset to zero
__m128i temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
__m128i resq_r0, resq_r1, resq_r2, resq_r3;
__m128i add_rshift = _mm_set1_epi32((1 << (3 - u4_qp_div_6)));
__m128i value_32 = _mm_set1_epi32(32);
__m128i chroma_mask = _mm_set1_epi16 (0xFF);
__m128i out_r0, out_r1, out_r2, out_r3;
UNUSED (pi2_tmp);
/*************************************************************/
/* Dequantization of coefficients. Will be replaced by SIMD */
/* operations on platform */
/*************************************************************/
src_r0_r1 = _mm_loadu_si128((__m128i *) (pi2_src)); //a00 a01 a02 a03 a10 a11 a12 a13 -- the source matrix 0th,1st row
src_r2_r3 = _mm_loadu_si128((__m128i *) (pi2_src + 8)); //a20 a21 a22 a23 a30 a31 a32 a33 -- the source matrix 2nd,3rd row
scalemat_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat)); //b00 b01 b02 b03 b10 b11 b12 b13 -- the scaling matrix 0th,1st row
scalemat_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat + 8)); //b20 b21 b22 b23 b30 b31 b32 b33 -- the scaling matrix 2nd,3rd row
dequant_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat)); //q00 q01 q02 q03 q10 q11 q12 q13 -- all 16 bits
dequant_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat + 8)); //q20 q21 q22 q23 q30 q31 q32 q33 -- all 16 bits
temp0 = _mm_mullo_epi16(scalemat_r0_r1, dequant_r0_r1); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result
temp1 = _mm_mullo_epi16(scalemat_r2_r3, dequant_r2_r3); //b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 b12*q12 b13*q13 -- 16 bit result
temp4 = _mm_unpacklo_epi16(temp0, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long
temp5 = _mm_unpackhi_epi16(temp0, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long
temp6 = _mm_unpacklo_epi16(temp1, zero_8x16b); // b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long
temp7 = _mm_unpackhi_epi16(temp1, zero_8x16b); // b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long
src_r0 = _mm_unpacklo_epi16(src_r0_r1, zero_8x16b); // a00 0 a01 0 a02 0 a03 0 -- 16 bit long
src_r1 = _mm_unpackhi_epi16(src_r0_r1, zero_8x16b); // a10 0 a11 0 a12 0 a13 0 -- 16 bit long
src_r2 = _mm_unpacklo_epi16(src_r2_r3, zero_8x16b); // a20 0 a21 0 a22 0 a23 0 -- 16 bit long
src_r3 = _mm_unpackhi_epi16(src_r2_r3, zero_8x16b); // a30 0 a31 0 a32 0 a33 0 -- 16 bit long
temp4 = _mm_madd_epi16(src_r0, temp4); //a00*b00*q00 a10*b10*q10 a20*b20*q20 a30*b30 q30 -- 32 bits long
temp5 = _mm_madd_epi16(src_r1, temp5);
temp6 = _mm_madd_epi16(src_r2, temp6);
temp7 = _mm_madd_epi16(src_r3, temp7);
if (u4_qp_div_6 >= 4) {
resq_r0 = _mm_slli_epi32(temp4, u4_qp_div_6 - 4);
resq_r1 = _mm_slli_epi32(temp5, u4_qp_div_6 - 4);
resq_r2 = _mm_slli_epi32(temp6, u4_qp_div_6 - 4);
resq_r3 = _mm_slli_epi32(temp7, u4_qp_div_6 - 4);
} else {
temp4 = _mm_add_epi32(temp4, add_rshift);
temp5 = _mm_add_epi32(temp5, add_rshift);
temp6 = _mm_add_epi32(temp6, add_rshift);
temp7 = _mm_add_epi32(temp7, add_rshift);
resq_r0 = _mm_srai_epi32(temp4, 4 - u4_qp_div_6);
resq_r1 = _mm_srai_epi32(temp5, 4 - u4_qp_div_6);
resq_r2 = _mm_srai_epi32(temp6, 4 - u4_qp_div_6);
resq_r3 = _mm_srai_epi32(temp7, 4 - u4_qp_div_6);
}
resq_r0 = _mm_insert_epi32(resq_r0,(WORD32)pi2_dc_ld_addr[0],0);
/* Perform Inverse transform */
/*-------------------------------------------------------------*/
/* IDCT [ Horizontal transformation ] */
/*-------------------------------------------------------------*/
// Matrix transpose
/*
* a0 a1 a2 a3
* b0 b1 b2 b3
* c0 c1 c2 c3
* d0 d1 d2 d3
*/
temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 b0 a1 b1
temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //c0 d0 c1 d1
temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //a2 b2 a3 b3
temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 d2 c3 d3
resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 b0 c0 d0
resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //a1 b1 c1 d1
resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //a2 b2 c2 d2
resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //a3 b3 c3 d3
//Transform starts -- horizontal transform
/*------------------------------------------------------------------*/
/* z0 = w0 + w2 */
temp0 = _mm_add_epi32(resq_r0, resq_r2);
/* z1 = w0 - w2 */
temp1 = _mm_sub_epi32(resq_r0, resq_r2);
/* z2 = (w1 >> 1) - w3 */
temp2 = _mm_srai_epi32(resq_r1, 1); //(w1>>1)
temp2 = _mm_sub_epi32(temp2, resq_r3); //(w1>>1) - w3
/* z3 = w1 + (w3 >> 1) */
temp3 = _mm_srai_epi32(resq_r3, 1); //(w3>>1) + w1
temp3 = _mm_add_epi32(temp3, resq_r1);
/*----------------------------------------------------------*/
/* x0 = z0 + z3 */
resq_r0 = _mm_add_epi32(temp0, temp3);
/* x1 = z1 + z2 */
resq_r1 = _mm_add_epi32(temp1, temp2);
/* x2 = z1 - z2 */
resq_r2 = _mm_sub_epi32(temp1, temp2);
/* x3 = z0 - z3 */
resq_r3 = _mm_sub_epi32(temp0, temp3);
// Matrix transpose
/*
* a0 b0 c0 d0
* a1 b1 c1 d1
* a2 b2 c2 d2
* a3 b3 c3 d3
*/
temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); //a0 a1 b0 b1
temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); //a2 a3 b2 b3
temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); //c0 c1 d0 d1
temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); //c2 c3 d2 d3
resq_r0 = _mm_unpacklo_epi64(temp1, temp3); //a0 a1 a2 a3
resq_r1 = _mm_unpackhi_epi64(temp1, temp3); //b0 b1 b2 b3
resq_r2 = _mm_unpacklo_epi64(temp2, temp4); //c0 c1 c2 c3
resq_r3 = _mm_unpackhi_epi64(temp2, temp4); //d0 d1 d2 d3
//Transform ends -- horizontal transform
//Load pred buffer
pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); //p00 p01 p02 p03 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[pred_strd])); //p10 p11 p12 p13 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * pred_strd])); //p20 p21 p22 p23 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * pred_strd])); //p30 p31 p32 p33 0 0 0 0 0 0 0 0 -- all 8 bits
pred_r0 = _mm_and_si128(pred_r0, chroma_mask);
pred_r1 = _mm_and_si128(pred_r1, chroma_mask);
pred_r2 = _mm_and_si128(pred_r2, chroma_mask);
pred_r3 = _mm_and_si128(pred_r3, chroma_mask);
pred_r0 = _mm_cvtepu16_epi32(pred_r0); //p00 p01 p02 p03 -- all 32 bits
pred_r1 = _mm_cvtepu16_epi32(pred_r1); //p10 p11 p12 p13 -- all 32 bits
pred_r2 = _mm_cvtepu16_epi32(pred_r2); //p20 p21 p22 p23 -- all 32 bits
pred_r3 = _mm_cvtepu16_epi32(pred_r3); //p30 p31 p32 p33 -- all 32 bits
/*--------------------------------------------------------------*/
/* IDCT [ Vertical transformation] and Xij = (xij + 32)>>6 */
/* */
/* Add the prediction and store it back to same buffer */
/*--------------------------------------------------------------*/
/* z0j = y0j + y2j */
temp0 = _mm_add_epi32(resq_r0, resq_r2);
/* z1j = y0j - y2j */
temp1 = _mm_sub_epi32(resq_r0, resq_r2);
/* z2j = (y1j>>1) - y3j */
temp2 = _mm_srai_epi32(resq_r1, 1); //(y1j>>1)
temp2 = _mm_sub_epi32(temp2, resq_r3);
/* z3j = y1j + (y3j>>1) */
temp3 = _mm_srai_epi32(resq_r3, 1); //(y3j>>1)
temp3 = _mm_add_epi32(temp3, resq_r1);
/* x0j = z0j + z3j */
temp4 = _mm_add_epi32(temp0, temp3);
temp4 = _mm_add_epi32(temp4, value_32);
temp4 = _mm_srai_epi32(temp4, 6);
temp4 = _mm_add_epi32(temp4, pred_r0);
/* x1j = z1j + z2j */
temp5 = _mm_add_epi32(temp1, temp2);
temp5 = _mm_add_epi32(temp5, value_32);
temp5 = _mm_srai_epi32(temp5, 6);
temp5 = _mm_add_epi32(temp5, pred_r1);
/* x2j = z1j - z2j */
temp6 = _mm_sub_epi32(temp1, temp2);
temp6 = _mm_add_epi32(temp6, value_32);
temp6 = _mm_srai_epi32(temp6, 6);
temp6 = _mm_add_epi32(temp6, pred_r2);
/* x3j = z0j - z3j */
temp7 = _mm_sub_epi32(temp0, temp3);
temp7 = _mm_add_epi32(temp7, value_32);
temp7 = _mm_srai_epi32(temp7, 6);
temp7 = _mm_add_epi32(temp7, pred_r3);
// 32-bit to 16-bit conversion
temp0 = _mm_packs_epi32(temp4, temp5);
temp1 = _mm_packs_epi32(temp6, temp7);
/*------------------------------------------------------------------*/
//Clipping the results to 8 bits
sign_reg = _mm_cmpgt_epi16(temp0, zero_8x16b); // sign check
temp0 = _mm_and_si128(temp0, sign_reg);
sign_reg = _mm_cmpgt_epi16(temp1, zero_8x16b);
temp1 = _mm_and_si128(temp1, sign_reg);
resq_r0 = _mm_packus_epi16(temp0, temp1);
resq_r1 = _mm_srli_si128(resq_r0, 4);
resq_r2 = _mm_srli_si128(resq_r1, 4);
resq_r3 = _mm_srli_si128(resq_r2, 4);
resq_r0 = _mm_cvtepu8_epi16(resq_r0); //p00 p01 p02 p03 -- all 16 bits
resq_r1 = _mm_cvtepu8_epi16(resq_r1); //p10 p11 p12 p13 -- all 16 bits
resq_r2 = _mm_cvtepu8_epi16(resq_r2); //p20 p21 p22 p23 -- all 16 bits
resq_r3 = _mm_cvtepu8_epi16(resq_r3); //p30 p31 p32 p33 -- all 16 bits
chroma_mask = _mm_set1_epi16 (0xFF00);
out_r0 = _mm_loadl_epi64((__m128i *) (&pu1_out[0]));
out_r1 = _mm_loadl_epi64((__m128i *) (&pu1_out[out_strd]));
out_r2 = _mm_loadl_epi64((__m128i *) (&pu1_out[2 * out_strd]));
out_r3 = _mm_loadl_epi64((__m128i *) (&pu1_out[3 * out_strd]));
out_r0 = _mm_and_si128(out_r0, chroma_mask);
out_r1 = _mm_and_si128(out_r1, chroma_mask);
out_r2 = _mm_and_si128(out_r2, chroma_mask);
out_r3 = _mm_and_si128(out_r3, chroma_mask);
out_r0 = _mm_add_epi8(out_r0, resq_r0);
out_r1 = _mm_add_epi8(out_r1, resq_r1);
out_r2 = _mm_add_epi8(out_r2, resq_r2);
out_r3 = _mm_add_epi8(out_r3, resq_r3);
_mm_storel_epi64((__m128i *)(&pu1_out[0]), out_r0);
_mm_storel_epi64((__m128i *)(&pu1_out[out_strd]), out_r1);
_mm_storel_epi64((__m128i *)(&pu1_out[2 * out_strd]), out_r2);
_mm_storel_epi64((__m128i *)(&pu1_out[3 * out_strd]), out_r3);
}
|
