2 * Copyright (c) 2002-2007, Communications and Remote Sensing Laboratory, Universite catholique de Louvain (UCL), Belgium
3 * Copyright (c) 2002-2007, Professor Benoit Macq
4 * Copyright (c) 2001-2003, David Janssens
5 * Copyright (c) 2002-2003, Yannick Verschueren
6 * Copyright (c) 2003-2007, Francois-Olivier Devaux and Antonin Descampe
7 * Copyright (c) 2005, Herve Drolon, FreeImage Team
8 * Copyright (c) 2007, Jonathan Ballard <dzonatas@dzonux.net>
9 * Copyright (c) 2007, Callum Lerwick <seg@haxxed.com>
10 * Copyright (c) 2008, Jerome Fimes, Communications & Systemes <jerome.fimes@c-s.fr>
11 * All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS `AS IS'
23 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 * POSSIBILITY OF SUCH DAMAGE.
36 #include <xmmintrin.h>
43 #include "opj_malloc.h"
46 /** @defgroup DWT DWT - Implementation of a discrete wavelet transform */
49 #define WS(i) v->mem[(i)*2]
50 #define WD(i) v->mem[(1+(i)*2)]
52 /** @name Local data structures */
55 typedef struct dwt_local {
66 typedef struct v4dwt_local {
73 static const OPJ_FLOAT32 dwt_alpha = 1.586134342f; // 12994
74 static const OPJ_FLOAT32 dwt_beta = 0.052980118f; // 434
75 static const OPJ_FLOAT32 dwt_gamma = -0.882911075f; // -7233
76 static const OPJ_FLOAT32 delta = -0.443506852f; // -3633
78 static const OPJ_FLOAT32 K = 1.230174105f; // 10078
79 /* FIXME: What is this constant? */
80 static const OPJ_FLOAT32 c13318 = 1.625732422f;
85 Virtual function type for wavelet transform in 1-D
87 typedef void (*DWT1DFN)(dwt_t* v);
89 /** @name Local static functions */
93 Forward lazy transform (horizontal)
95 static void dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
97 Forward lazy transform (vertical)
99 static void dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas);
101 Inverse lazy transform (horizontal)
103 static void dwt_interleave_h(dwt_t* h, OPJ_INT32 *a);
105 Inverse lazy transform (vertical)
107 static void dwt_interleave_v(dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x);
109 Forward 5-3 wavelet transform in 1-D
111 static void dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
113 Inverse 5-3 wavelet transform in 1-D
115 static void dwt_decode_1(dwt_t *v);
117 Forward 9-7 wavelet transform in 1-D
119 static void dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas);
121 Explicit calculation of the Quantization Stepsizes
123 static void dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize);
125 Inverse wavelet transform in 2-D.
127 static bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 i, DWT1DFN fn);
129 static OPJ_UINT32 dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i);
131 static INLINE bool dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) );
136 #define S(i) a[(i)*2]
137 #define D(i) a[(1+(i)*2)]
138 #define S_(i) ((i)<0?S(0):((i)>=sn?S(sn-1):S(i)))
139 #define D_(i) ((i)<0?D(0):((i)>=dn?D(dn-1):D(i)))
141 #define SS_(i) ((i)<0?S(0):((i)>=dn?S(dn-1):S(i)))
142 #define DD_(i) ((i)<0?D(0):((i)>=sn?D(sn-1):D(i)))
145 /* This table contains the norms of the 5-3 wavelets for different bands. */
147 static const OPJ_FLOAT64 dwt_norms[4][10] = {
148 {1.000, 1.500, 2.750, 5.375, 10.68, 21.34, 42.67, 85.33, 170.7, 341.3},
149 {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9},
150 {1.038, 1.592, 2.919, 5.703, 11.33, 22.64, 45.25, 90.48, 180.9},
151 {.7186, .9218, 1.586, 3.043, 6.019, 12.01, 24.00, 47.97, 95.93}
155 /* This table contains the norms of the 9-7 wavelets for different bands. */
157 static const OPJ_FLOAT64 dwt_norms_real[4][10] = {
158 {1.000, 1.965, 4.177, 8.403, 16.90, 33.84, 67.69, 135.3, 270.6, 540.9},
159 {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0},
160 {2.022, 3.989, 8.355, 17.04, 34.27, 68.63, 137.3, 274.6, 549.0},
161 {2.080, 3.865, 8.307, 17.18, 34.71, 69.59, 139.3, 278.6, 557.2}
165 ==========================================================
167 ==========================================================
171 /* Forward lazy transform (horizontal). */
173 static void dwt_deinterleave_h(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) {
176 OPJ_INT32 * l_dest = b;
177 OPJ_INT32 * l_src = a+cas;
195 /* Forward lazy transform (vertical). */
197 static void dwt_deinterleave_v(OPJ_INT32 *a, OPJ_INT32 *b, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 x, OPJ_INT32 cas) {
199 OPJ_INT32 * l_dest = b;
200 OPJ_INT32 * l_src = a+cas;
208 /* b[i*x]=a[2*i+cas]; */
220 /*b[(sn+i)*x]=a[(2*i+1-cas)];*/
225 /* Inverse lazy transform (horizontal). */
227 static void dwt_interleave_h(dwt_t* h, OPJ_INT32 *a) {
229 OPJ_INT32 *bi = h->mem + h->cas;
238 bi = h->mem + 1 - h->cas;
249 /* Inverse lazy transform (vertical). */
251 static void dwt_interleave_v(dwt_t* v, OPJ_INT32 *a, OPJ_INT32 x) {
253 OPJ_INT32 *bi = v->mem + v->cas;
260 ai = a + (v->sn * x);
261 bi = v->mem + 1 - v->cas;
272 /* Forward 5-3 wavelet transform in 1-D. */
274 static void dwt_encode_1(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) {
278 if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */
279 for (i = 0; i < dn; i++) D(i) -= (S_(i) + S_(i + 1)) >> 1;
280 for (i = 0; i < sn; i++) S(i) += (D_(i - 1) + D_(i) + 2) >> 2;
283 if (!sn && dn == 1) /* NEW : CASE ONE ELEMENT */
286 for (i = 0; i < dn; i++) S(i) -= (DD_(i) + DD_(i - 1)) >> 1;
287 for (i = 0; i < sn; i++) D(i) += (SS_(i) + SS_(i + 1) + 2) >> 2;
293 /* Inverse 5-3 wavelet transform in 1-D. */
295 static void dwt_decode_1_(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) {
299 if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */
300 for (i = 0; i < sn; i++) S(i) -= (D_(i - 1) + D_(i) + 2) >> 2;
301 for (i = 0; i < dn; i++) D(i) += (S_(i) + S_(i + 1)) >> 1;
304 if (!sn && dn == 1) /* NEW : CASE ONE ELEMENT */
307 for (i = 0; i < sn; i++) D(i) -= (SS_(i) + SS_(i + 1) + 2) >> 2;
308 for (i = 0; i < dn; i++) S(i) += (DD_(i) + DD_(i - 1)) >> 1;
314 /* Inverse 5-3 wavelet transform in 1-D. */
316 static void dwt_decode_1(dwt_t *v) {
317 dwt_decode_1_(v->mem, v->dn, v->sn, v->cas);
321 /* Forward 9-7 wavelet transform in 1-D. */
323 static void dwt_encode_1_real(OPJ_INT32 *a, OPJ_INT32 dn, OPJ_INT32 sn, OPJ_INT32 cas) {
326 if ((dn > 0) || (sn > 1)) { /* NEW : CASE ONE ELEMENT */
327 for (i = 0; i < dn; i++)
328 D(i) -= fix_mul(S_(i) + S_(i + 1), 12993);
329 for (i = 0; i < sn; i++)
330 S(i) -= fix_mul(D_(i - 1) + D_(i), 434);
331 for (i = 0; i < dn; i++)
332 D(i) += fix_mul(S_(i) + S_(i + 1), 7233);
333 for (i = 0; i < sn; i++)
334 S(i) += fix_mul(D_(i - 1) + D_(i), 3633);
335 for (i = 0; i < dn; i++)
336 D(i) = fix_mul(D(i), 5038); /*5038 */
337 for (i = 0; i < sn; i++)
338 S(i) = fix_mul(S(i), 6659); /*6660 */
341 if ((sn > 0) || (dn > 1)) { /* NEW : CASE ONE ELEMENT */
342 for (i = 0; i < dn; i++)
343 S(i) -= fix_mul(DD_(i) + DD_(i - 1), 12993);
344 for (i = 0; i < sn; i++)
345 D(i) -= fix_mul(SS_(i) + SS_(i + 1), 434);
346 for (i = 0; i < dn; i++)
347 S(i) += fix_mul(DD_(i) + DD_(i - 1), 7233);
348 for (i = 0; i < sn; i++)
349 D(i) += fix_mul(SS_(i) + SS_(i + 1), 3633);
350 for (i = 0; i < dn; i++)
351 S(i) = fix_mul(S(i), 5038); /*5038 */
352 for (i = 0; i < sn; i++)
353 D(i) = fix_mul(D(i), 6659); /*6660 */
358 static void dwt_encode_stepsize(OPJ_INT32 stepsize, OPJ_INT32 numbps, opj_stepsize_t *bandno_stepsize) {
360 p = int_floorlog2(stepsize) - 13;
361 n = 11 - int_floorlog2(stepsize);
362 bandno_stepsize->mant = (n < 0 ? stepsize >> -n : stepsize << n) & 0x7ff;
363 bandno_stepsize->expn = numbps - p;
367 ==========================================================
369 ==========================================================
373 /* Forward 5-3 wavelet transform in 2-D. */
375 INLINE bool dwt_encode_procedure(opj_tcd_tilecomp_t * tilec,void (*p_function)(OPJ_INT32 *, OPJ_INT32,OPJ_INT32,OPJ_INT32) )
383 OPJ_INT32 rw; /* width of the resolution level computed */
384 OPJ_INT32 rh; /* height of the resolution level computed */
385 OPJ_INT32 l_data_size;
387 opj_tcd_resolution_t * l_cur_res = 0;
388 opj_tcd_resolution_t * l_last_res = 0;
390 w = tilec->x1-tilec->x0;
391 l = tilec->numresolutions-1;
394 l_cur_res = tilec->resolutions + l;
395 l_last_res = l_cur_res - 1;
397 rw = l_cur_res->x1 - l_cur_res->x0;
398 rh = l_cur_res->y1 - l_cur_res->y0;
400 l_data_size = dwt_max_resolution( tilec->resolutions,tilec->numresolutions) * sizeof(OPJ_INT32);
401 bj = opj_malloc(l_data_size);
412 OPJ_INT32 rw1; /* width of the resolution level once lower than computed one */
413 OPJ_INT32 rh1; /* height of the resolution level once lower than computed one */
414 OPJ_INT32 cas_col; /* 0 = non inversion on horizontal filtering 1 = inversion between low-pass and high-pass filtering */
415 OPJ_INT32 cas_row; /* 0 = non inversion on vertical filtering 1 = inversion between low-pass and high-pass filtering */
418 rw = l_cur_res->x1 - l_cur_res->x0;
419 rh = l_cur_res->y1 - l_cur_res->y0;
420 rw1 = l_last_res->x1 - l_last_res->x0;
421 rh1 = l_last_res->y1 - l_last_res->y0;
423 cas_row = l_cur_res->x0 & 1;
424 cas_col = l_cur_res->y0 & 1;
437 (*p_function) (bj, dn, sn, cas_col);
438 dwt_deinterleave_v(bj, aj, dn, sn, w, cas_col);
442 for (j = 0; j < rh; j++)
445 for (k = 0; k < rw; k++) bj[k] = aj[k];
446 (*p_function) (bj, dn, sn, cas_row);
447 dwt_deinterleave_h(bj, aj, dn, sn, cas_row);
449 l_cur_res = l_last_res;
455 /* Forward 5-3 wavelet transform in 2-D. */
457 bool dwt_encode(opj_tcd_tilecomp_t * tilec)
459 return dwt_encode_procedure(tilec,dwt_encode_1);
463 /* Inverse 5-3 wavelet transform in 2-D. */
465 bool dwt_decode(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres) {
466 return dwt_decode_tile(tilec, numres, &dwt_decode_1);
471 /* Get gain of 5-3 wavelet transform. */
473 OPJ_UINT32 dwt_getgain(OPJ_UINT32 orient) {
476 if (orient == 1 || orient == 2)
482 /* Get norm of 5-3 wavelet. */
484 OPJ_FLOAT64 dwt_getnorm(OPJ_UINT32 level, OPJ_UINT32 orient) {
485 return dwt_norms[orient][level];
489 /* Forward 9-7 wavelet transform in 2-D. */
491 bool dwt_encode_real(opj_tcd_tilecomp_t * tilec)
493 return dwt_encode_procedure(tilec,dwt_encode_1_real);
499 /* Get gain of 9-7 wavelet transform. */
501 OPJ_UINT32 dwt_getgain_real(OPJ_UINT32 orient) {
507 /* Get norm of 9-7 wavelet. */
509 OPJ_FLOAT64 dwt_getnorm_real(OPJ_UINT32 level, OPJ_UINT32 orient) {
510 return dwt_norms_real[orient][level];
513 void dwt_calc_explicit_stepsizes(opj_tccp_t * tccp, OPJ_UINT32 prec) {
514 OPJ_UINT32 numbands, bandno;
515 numbands = 3 * tccp->numresolutions - 2;
516 for (bandno = 0; bandno < numbands; bandno++) {
517 OPJ_FLOAT64 stepsize;
518 OPJ_UINT32 resno, level, orient, gain;
520 resno = (bandno == 0) ? 0 : ((bandno - 1) / 3 + 1);
521 orient = (bandno == 0) ? 0 : ((bandno - 1) % 3 + 1);
522 level = tccp->numresolutions - 1 - resno;
523 gain = (tccp->qmfbid == 0) ? 0 : ((orient == 0) ? 0 : (((orient == 1) || (orient == 2)) ? 1 : 2));
524 if (tccp->qntsty == J2K_CCP_QNTSTY_NOQNT) {
527 OPJ_FLOAT64 norm = dwt_norms_real[orient][level];
528 stepsize = (1 << (gain)) / norm;
530 dwt_encode_stepsize((OPJ_INT32) floor(stepsize * 8192.0), prec + gain, &tccp->stepsizes[bandno]);
536 /* Determine maximum computed resolution level for inverse wavelet transform */
538 static OPJ_UINT32 dwt_max_resolution(opj_tcd_resolution_t* restrict r, OPJ_UINT32 i) {
543 if( mr < ( w = r->x1 - r->x0 ) )
545 if( mr < ( w = r->y1 - r->y0 ) )
553 /* Inverse wavelet transform in 2-D. */
555 static bool dwt_decode_tile(opj_tcd_tilecomp_t* tilec, OPJ_UINT32 numres, DWT1DFN dwt_1D) {
559 opj_tcd_resolution_t* tr = tilec->resolutions;
561 OPJ_UINT32 rw = tr->x1 - tr->x0; /* width of the resolution level computed */
562 OPJ_UINT32 rh = tr->y1 - tr->y0; /* height of the resolution level computed */
564 OPJ_UINT32 w = tilec->x1 - tilec->x0;
566 h.mem = opj_aligned_malloc(dwt_max_resolution(tr, numres) * sizeof(OPJ_INT32));
576 OPJ_INT32 * restrict tiledp = tilec->data;
583 rw = tr->x1 - tr->x0;
584 rh = tr->y1 - tr->y0;
589 for(j = 0; j < rh; ++j) {
590 dwt_interleave_h(&h, &tiledp[j*w]);
592 memcpy(&tiledp[j*w], h.mem, rw * sizeof(OPJ_INT32));
598 for(j = 0; j < rw; ++j){
600 dwt_interleave_v(&v, &tiledp[j], w);
602 for(k = 0; k < rh; ++k) {
603 tiledp[k * w + j] = v.mem[k];
607 opj_aligned_free(h.mem);
611 static void v4dwt_interleave_h(v4dwt_t* restrict w, OPJ_FLOAT32* restrict a, OPJ_INT32 x, OPJ_INT32 size){
612 OPJ_FLOAT32* restrict bi = (OPJ_FLOAT32*) (w->wavelet + w->cas);
613 OPJ_INT32 count = w->sn;
615 for(k = 0; k < 2; ++k){
616 for(i = 0; i < count; ++i){
620 if(j >= size) continue;
623 if(j >= size) continue;
626 if(j >= size) continue;
629 bi = (OPJ_FLOAT32*) (w->wavelet + 1 - w->cas);
636 static void v4dwt_interleave_v(v4dwt_t* restrict v , OPJ_FLOAT32* restrict a , OPJ_INT32 x){
637 v4* restrict bi = v->wavelet + v->cas;
639 for(i = 0; i < v->sn; ++i){
640 memcpy(&bi[i*2], &a[i*x], 4 * sizeof(OPJ_FLOAT32));
643 bi = v->wavelet + 1 - v->cas;
644 for(i = 0; i < v->dn; ++i){
645 memcpy(&bi[i*2], &a[i*x], 4 * sizeof(OPJ_FLOAT32));
651 static void v4dwt_decode_step1_sse(v4* w, OPJ_INT32 count, const __m128 c){
652 __m128* restrict vw = (__m128*) w;
654 for(i = 0; i < count; ++i){
655 __m128 tmp = vw[i*2];
660 static void v4dwt_decode_step2_sse(v4* l, v4* w, OPJ_INT32 k, OPJ_INT32 m, __m128 c){
661 __m128* restrict vl = (__m128*) l;
662 __m128* restrict vw = (__m128*) w;
664 for(i = 0; i < m; ++i){
665 __m128 tmp1 = vl[ 0];
666 __m128 tmp2 = vw[-1];
667 __m128 tmp3 = vw[ 0];
668 vw[-1] = tmp2 + ((tmp1 + tmp3) * c);
686 static void v4dwt_decode_step1(v4* w, OPJ_INT32 count, const OPJ_FLOAT32 c){
687 OPJ_FLOAT32* restrict fw = (OPJ_FLOAT32*) w;
689 for(i = 0; i < count; ++i){
690 OPJ_FLOAT32 tmp1 = fw[i*8 ];
691 OPJ_FLOAT32 tmp2 = fw[i*8 + 1];
692 OPJ_FLOAT32 tmp3 = fw[i*8 + 2];
693 OPJ_FLOAT32 tmp4 = fw[i*8 + 3];
695 fw[i*8 + 1] = tmp2 * c;
696 fw[i*8 + 2] = tmp3 * c;
697 fw[i*8 + 3] = tmp4 * c;
701 static void v4dwt_decode_step2(v4* l, v4* w, OPJ_INT32 k, OPJ_INT32 m, OPJ_FLOAT32 c){
702 OPJ_FLOAT32* restrict fl = (OPJ_FLOAT32*) l;
703 OPJ_FLOAT32* restrict fw = (OPJ_FLOAT32*) w;
705 for(i = 0; i < m; ++i){
706 OPJ_FLOAT32 tmp1_1 = fl[0];
707 OPJ_FLOAT32 tmp1_2 = fl[1];
708 OPJ_FLOAT32 tmp1_3 = fl[2];
709 OPJ_FLOAT32 tmp1_4 = fl[3];
710 OPJ_FLOAT32 tmp2_1 = fw[-4];
711 OPJ_FLOAT32 tmp2_2 = fw[-3];
712 OPJ_FLOAT32 tmp2_3 = fw[-2];
713 OPJ_FLOAT32 tmp2_4 = fw[-1];
714 OPJ_FLOAT32 tmp3_1 = fw[0];
715 OPJ_FLOAT32 tmp3_2 = fw[1];
716 OPJ_FLOAT32 tmp3_3 = fw[2];
717 OPJ_FLOAT32 tmp3_4 = fw[3];
718 fw[-4] = tmp2_1 + ((tmp1_1 + tmp3_1) * c);
719 fw[-3] = tmp2_2 + ((tmp1_2 + tmp3_2) * c);
720 fw[-2] = tmp2_3 + ((tmp1_3 + tmp3_3) * c);
721 fw[-1] = tmp2_4 + ((tmp1_4 + tmp3_4) * c);
736 OPJ_FLOAT32 tmp1 = fw[-4];
737 OPJ_FLOAT32 tmp2 = fw[-3];
738 OPJ_FLOAT32 tmp3 = fw[-2];
739 OPJ_FLOAT32 tmp4 = fw[-1];
752 /* Inverse 9-7 wavelet transform in 1-D. */
754 static void v4dwt_decode(v4dwt_t* restrict dwt){
757 if(!((dwt->dn > 0) || (dwt->sn > 1))){
763 if(!((dwt->sn > 0) || (dwt->dn > 1))) {
770 v4dwt_decode_step1_sse(dwt->wavelet+a, dwt->sn, _mm_set1_ps(K));
771 v4dwt_decode_step1_sse(dwt->wavelet+b, dwt->dn, _mm_set1_ps(c13318));
772 v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(delta));
773 v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_gamma));
774 v4dwt_decode_step2_sse(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), _mm_set1_ps(dwt_beta));
775 v4dwt_decode_step2_sse(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), _mm_set1_ps(dwt_alpha));
777 v4dwt_decode_step1(dwt->wavelet+a, dwt->sn, K);
778 v4dwt_decode_step1(dwt->wavelet+b, dwt->dn, c13318);
779 v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), delta);
780 v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_gamma);
781 v4dwt_decode_step2(dwt->wavelet+b, dwt->wavelet+a+1, dwt->sn, int_min(dwt->sn, dwt->dn-a), dwt_beta);
782 v4dwt_decode_step2(dwt->wavelet+a, dwt->wavelet+b+1, dwt->dn, int_min(dwt->dn, dwt->sn-b), dwt_alpha);
787 /* Inverse 9-7 wavelet transform in 2-D. */
789 bool dwt_decode_real(opj_tcd_tilecomp_t* restrict tilec, OPJ_UINT32 numres){
793 opj_tcd_resolution_t* res = tilec->resolutions;
795 OPJ_UINT32 rw = res->x1 - res->x0; /* width of the resolution level computed */
796 OPJ_UINT32 rh = res->y1 - res->y0; /* height of the resolution level computed */
798 OPJ_UINT32 w = tilec->x1 - tilec->x0;
800 h.wavelet = (v4*) opj_aligned_malloc((dwt_max_resolution(res, numres)+5) * sizeof(v4));
801 v.wavelet = h.wavelet;
804 OPJ_FLOAT32 * restrict aj = (OPJ_FLOAT32*) tilec->data;
805 OPJ_UINT32 bufsize = (tilec->x1 - tilec->x0) * (tilec->y1 - tilec->y0);
813 rw = res->x1 - res->x0; /* width of the resolution level computed */
814 rh = res->y1 - res->y0; /* height of the resolution level computed */
819 for(j = rh; j > 0; j -= 4){
820 v4dwt_interleave_h(&h, aj, w, bufsize);
827 aj[k ] = h.wavelet[k].f[0];
828 aj[k+w ] = h.wavelet[k].f[1];
829 aj[k+w*2] = h.wavelet[k].f[2];
830 aj[k+w*3] = h.wavelet[k].f[3];
838 case 3: aj[k+w*2] = h.wavelet[k].f[2];
839 case 2: aj[k+w ] = h.wavelet[k].f[1];
840 case 1: aj[k ] = h.wavelet[k].f[0];
851 aj = (OPJ_FLOAT32*) tilec->data;
852 for(j = rw; j > 0; j -= 4){
853 v4dwt_interleave_v(&v, aj, w);
857 for(k = 0; k < rh; ++k){
858 memcpy(&aj[k*w], &v.wavelet[k], 4 * sizeof(OPJ_FLOAT32));
862 for(k = 0; k < rh; ++k){
863 memcpy(&aj[k*w], &v.wavelet[k], j * sizeof(OPJ_FLOAT32));
870 opj_aligned_free(h.wavelet);