summaryrefslogtreecommitdiffstats
path: root/apps/recorder/resize.c
blob: c16cec9dc9518d7b74dab5a4840860ba9bbd9d64 (plain)
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
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
/***************************************************************************
 *             __________               __   ___.
 *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
 *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
 *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
 *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
 *                     \/            \/     \/    \/            \/
 * $Id$
 *
 * Copyright (C) 2008 by Akio Idehara, Andrew Mahone
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 ****************************************************************************/

/*
 * Implementation of area average and linear row and vertical scalers, and
 * nearest-neighbor grey scaler (C) 2008 Andrew Mahone
 *
 * All files in this archive are subject to the GNU General Public License.
 * See the file COPYING in the source tree root for full license agreement.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 ****************************************************************************/

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdint.h>

#include "general.h"
#include "kernel.h"
#include "system.h"
#ifndef PLUGIN
#include "debug.h"
#endif
#include "lcd.h"
#include "file.h"
#ifdef HAVE_REMOTE_LCD
#include "lcd-remote.h"
#endif
#ifdef ROCKBOX_DEBUG_SCALERS
#define SDEBUGF DEBUGF
#else
#define SDEBUGF(...)
#endif
#ifndef __PCTOOL__
#include "config.h"
#include "system.h"
#include <bmp.h>
#include "resize.h"
#else
#undef DEBUGF
#define DEBUGF(...)
#endif
#include <jpeg_load.h>

#define MULUQ(a, b) ((a) * (b))
#define MULQ(a, b) ((a) * (b))

#ifdef HAVE_LCD_COLOR
#define CHANNEL_BYTES (sizeof(struct uint32_argb)/sizeof(uint32_t))
#else
#define CHANNEL_BYTES (sizeof(uint32_t)/sizeof(uint32_t)) /* packed */
#endif

/* calculate the maximum dimensions which will preserve the aspect ration of
   src while fitting in the constraints passed in dst, and store result in dst,
   returning 0 if rounding and 1 if not rounding.
*/
int recalc_dimension(struct dim *dst, struct dim *src)
{
    /* This only looks backwards. The input image size is being pre-scaled by
     * the inverse of the pixel aspect ratio, so that once the size it scaled
     * to meet the output constraints, the scaled image will have appropriate
     * proportions.
     */
    int sw = src->width * LCD_PIXEL_ASPECT_HEIGHT;
    int sh = src->height * LCD_PIXEL_ASPECT_WIDTH;
    int tmp;
    if (dst->width <= 0)
        dst->width = LCD_WIDTH;
    if (dst->height <= 0)
        dst->height = LCD_HEIGHT;
#ifndef HAVE_UPSCALER
    if (dst->width > sw || dst->height > sh)
    {
        dst->width = sw;
        dst->height = sh;
    }
    if (sw == dst->width && sh == dst->height)
        return 1;
#endif
    tmp = (sw * dst->height + (sh >> 1)) / sh;
    if (tmp > dst->width)
        dst->height = (sh * dst->width + (sw >> 1)) / sw;
    else
        dst->width = tmp;
    return src->width == dst->width && src->height == dst->height;
}

/* All of these scalers use variations of Bresenham's algorithm to convert from
   their input to output coordinates.  The error value is shifted from the
   "classic" version such that it is a useful input to the scaling calculation.
*/

#ifdef HAVE_LCD_COLOR
/* dither + pack on channel of RGB565, R an B share a packing macro */
#define PACKRB(v, delta)  ((31 * v + (v >> 3) + delta) >> 8)
#define PACKG(g, delta) ((63 * g + (g >> 2) + delta) >> 8)
#endif

/* read new img_part unconditionally, return false on failure */
#define FILL_BUF_INIT(img_part, store_part, args) { \
    img_part = store_part(args); \
    if (img_part == NULL) \
        return false; \
}

/* read new img_part if current one is empty, return false on failure */
#define FILL_BUF(img_part, store_part, args) { \
    if (img_part->len == 0) \
        img_part = store_part(args); \
    if (img_part == NULL) \
        return false; \
}

#if defined(CPU_COLDFIRE)
#define MAC(op1, op2, num) \
    asm volatile( \
        "mac.l %0, %1, %%acc" #num \
        : \
        : "%d" (op1), "d" (op2)\
    )
#define MAC_OUT(dest, num) \
    asm volatile( \
        "movclr.l %%acc" #num ", %0" \
        : "=d" (dest) \
    )
#endif

/* horizontal area average scaler */
static bool scale_h_area(void *out_line_ptr,
                         struct scaler_context *ctx, bool accum)
{
    SDEBUGF("scale_h_area\n");
    unsigned int ix, ox, oxe, mul;
    const uint32_t h_i_val = ctx->h_i_val,
                   h_o_val = ctx->h_o_val;
#ifdef HAVE_LCD_COLOR
    struct uint32_argb rgbvalacc = { 0, 0, 0, 0 },
                       rgbvaltmp = { 0, 0, 0, 0 },
                      *out_line = (struct uint32_argb *)out_line_ptr;
#else
    uint32_t acc = 0, tmp = 0, *out_line = (uint32_t*)out_line_ptr;
#endif
    struct img_part *part;
    FILL_BUF_INIT(part,ctx->store_part,ctx->args);
    ox = 0;
    oxe = 0;
    mul = 0;
    /* give other tasks a chance to run */
    yield();
    for (ix = 0; ix < (unsigned int)ctx->src->width; ix++)
    {
        oxe += h_o_val;
        /* end of current area has been reached */
        /* fill buffer if needed */
        FILL_BUF(part,ctx->store_part,ctx->args);
#ifdef HAVE_LCD_COLOR
        if (oxe >= h_i_val)
        {
            /* "reset" error, which now represents partial coverage of next
               pixel by the next area
            */
            oxe -= h_i_val;

#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
            /* add saved partial pixel from start of area */
            MAC(rgbvalacc.r, h_o_val, 0);
            MAC(rgbvalacc.g, h_o_val, 1);
            MAC(rgbvalacc.b, h_o_val, 2);
            MAC(rgbvalacc.a, h_o_val, 3);
            MAC(rgbvaltmp.r, mul, 0);
            MAC(rgbvaltmp.g, mul, 1);
            MAC(rgbvaltmp.b, mul, 2);
            MAC(rgbvaltmp.a, mul, 3);
            /* get new pixel , then add its partial coverage to this area */
            mul = h_o_val - oxe;
            rgbvaltmp.r = part->buf->red;
            rgbvaltmp.g = part->buf->green;
            rgbvaltmp.b = part->buf->blue;
            rgbvaltmp.a = part->buf->alpha;
            MAC(rgbvaltmp.r, mul, 0);
            MAC(rgbvaltmp.g, mul, 1);
            MAC(rgbvaltmp.b, mul, 2);
            MAC(rgbvaltmp.a, mul, 3);
            MAC_OUT(rgbvalacc.r, 0);
            MAC_OUT(rgbvalacc.g, 1);
            MAC_OUT(rgbvalacc.b, 2);
            MAC_OUT(rgbvalacc.a, 3);
#else
/* generic C math */
            /* add saved partial pixel from start of area */
            rgbvalacc.r = rgbvalacc.r * h_o_val + rgbvaltmp.r * mul;
            rgbvalacc.g = rgbvalacc.g * h_o_val + rgbvaltmp.g * mul;
            rgbvalacc.b = rgbvalacc.b * h_o_val + rgbvaltmp.b * mul;
            rgbvalacc.a = rgbvalacc.a * h_o_val + rgbvaltmp.a * mul;

            /* get new pixel , then add its partial coverage to this area */
            rgbvaltmp.r = part->buf->red;
            rgbvaltmp.g = part->buf->green;
            rgbvaltmp.b = part->buf->blue;
            rgbvaltmp.a = part->buf->alpha;
            mul = h_o_val - oxe;
            rgbvalacc.r += rgbvaltmp.r * mul;
            rgbvalacc.g += rgbvaltmp.g * mul;
            rgbvalacc.b += rgbvaltmp.b * mul;
            rgbvalacc.a += rgbvaltmp.a * mul;
#endif /* CPU */
            rgbvalacc.r = (rgbvalacc.r + (1 << 21)) >> 22;
            rgbvalacc.g = (rgbvalacc.g + (1 << 21)) >> 22;
            rgbvalacc.b = (rgbvalacc.b + (1 << 21)) >> 22;
            rgbvalacc.a = (rgbvalacc.a + (1 << 21)) >> 22;
            /* store or accumulate to output row */
            if (accum)
            {
                rgbvalacc.r += out_line[ox].r;
                rgbvalacc.g += out_line[ox].g;
                rgbvalacc.b += out_line[ox].b;
                rgbvalacc.a += out_line[ox].a;
            }
            out_line[ox].r = rgbvalacc.r;
            out_line[ox].g = rgbvalacc.g;
            out_line[ox].b = rgbvalacc.b;
            out_line[ox].a = rgbvalacc.a;
            /* reset accumulator */
            rgbvalacc.r = 0;
            rgbvalacc.g = 0;
            rgbvalacc.b = 0;
            rgbvalacc.a = 0;
            mul = oxe;
            ox += 1;
        /* inside an area */
        } else {
            /* add pixel value to accumulator */
            rgbvalacc.r += part->buf->red;
            rgbvalacc.g += part->buf->green;
            rgbvalacc.b += part->buf->blue;
            rgbvalacc.a += part->buf->alpha;
        }
#else
        if (oxe >= h_i_val)
        {
            /* "reset" error, which now represents partial coverage of next
               pixel by the next area
            */
            oxe -= h_i_val;
#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
            /* add saved partial pixel from start of area */
            MAC(acc, h_o_val, 0);
            MAC(tmp, mul, 0);
            /* get new pixel , then add its partial coverage to this area */
            tmp = *(part->buf);
            mul = h_o_val - oxe;
            MAC(tmp, mul, 0);
            MAC_OUT(acc, 0);
#else
/* generic C math */
            /* add saved partial pixel from start of area */
            acc = (acc * h_o_val) + (tmp * mul);

            /* get new pixel , then add its partial coverage to this area */
            tmp = *(part->buf);
            mul = h_o_val - oxe;
            acc += tmp * mul;
#endif /* CPU */
            /* round, divide, and either store or accumulate to output row */
            acc = (acc + (1 << 21)) >> 22;
            if (accum)
            {
                acc += out_line[ox];
            }
            out_line[ox] = acc;
            /* reset accumulator */
            acc = 0;
            mul = oxe;
            ox += 1;
        /* inside an area */
        } else {
            /* add pixel value to accumulator */
            acc += *(part->buf);
        }
#endif
        part->buf++;
        part->len--;
    }
    return true;
}

/* vertical area average scaler */
static inline bool scale_v_area(struct rowset *rset, struct scaler_context *ctx)
{
    uint32_t mul, oy, iy, oye;
    const uint32_t v_i_val = ctx->v_i_val,
                   v_o_val = ctx->v_o_val;

    /* Set up rounding and scale factors */
    mul = 0;
    oy = rset->rowstart;
    oye = 0;
    uint32_t *rowacc = (uint32_t *) ctx->buf,
             *rowtmp = rowacc + ctx->bm->width * CHANNEL_BYTES,
             *rowacc_px, *rowtmp_px;
    memset((void *)ctx->buf, 0, ctx->bm->width * 2 * sizeof(uint32_t)*CHANNEL_BYTES);
    SDEBUGF("scale_v_area\n");
    /* zero the accumulator and temp rows */
    for (iy = 0; iy < (unsigned int)ctx->src->height; iy++)
    {
        oye += v_o_val;
        /* end of current area has been reached */
        if (oye >= v_i_val)
        {
            /* "reset" error, which now represents partial coverage of the next
               row by the next area
            */
            oye -= v_i_val;
            /* add stored partial row to accumulator */
            for(rowacc_px = rowacc, rowtmp_px = rowtmp; rowacc_px != rowtmp;
                rowacc_px++, rowtmp_px++)
                *rowacc_px = *rowacc_px * v_o_val + *rowtmp_px * mul;
            /* store new scaled row in temp row */
            if(!ctx->h_scaler(rowtmp, ctx, false))
                return false;
            /* add partial coverage by new row to this area, then round and
               scale to final value
            */
            mul = v_o_val - oye;
            for(rowacc_px = rowacc, rowtmp_px = rowtmp; rowacc_px != rowtmp;
                rowacc_px++, rowtmp_px++)
                *rowacc_px += mul * *rowtmp_px;
            ctx->output_row(oy, (void*)rowacc, ctx);
            /* clear accumulator row, store partial coverage for next row */
            memset((void *)rowacc, 0, ctx->bm->width * sizeof(uint32_t) * CHANNEL_BYTES);
            mul = oye;
            oy += rset->rowstep;
        /* inside an area */
        } else {
            /* accumulate new scaled row to rowacc */
            if (!ctx->h_scaler(rowacc, ctx, true))
                return false;
        }
    }
    return true;
}

#ifdef HAVE_UPSCALER
/* horizontal linear scaler */
static bool scale_h_linear(void *out_line_ptr, struct scaler_context *ctx,
                           bool accum)
{
    unsigned int ix, ox, ixe;
    const uint32_t h_i_val = ctx->h_i_val,
                   h_o_val = ctx->h_o_val;
    /* type x = x is an ugly hack for hiding an unitialized data warning. The
       values are conditionally initialized before use, but other values are
       set such that this will occur before these are used.
    */
#ifdef HAVE_LCD_COLOR
    struct uint32_argb rgbval=rgbval, rgbinc=rgbinc,
                      *out_line = (struct uint32_argb*)out_line_ptr;
#else
    uint32_t val=val, inc=inc, *out_line = (uint32_t*)out_line_ptr;
#endif
    struct img_part *part;
    SDEBUGF("scale_h_linear\n");
    FILL_BUF_INIT(part,ctx->store_part,ctx->args);
    ix = 0;
    /* The error is set so that values are initialized on the first pass. */
    ixe = h_o_val;
    /* give other tasks a chance to run */
    yield();
    for (ox = 0; ox < (uint32_t)ctx->bm->width; ox++)
    {
#ifdef HAVE_LCD_COLOR
        if (ixe >= h_o_val)
        {
            /* Store the new "current" pixel value in rgbval, and the color
               step value in rgbinc.
            */
            ixe -= h_o_val;
            rgbinc.r = -(part->buf->red);
            rgbinc.g = -(part->buf->green);
            rgbinc.b = -(part->buf->blue);
            rgbinc.a = -(part->buf->alpha);
#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
            MAC(part->buf->red, h_o_val, 0);
            MAC(part->buf->green, h_o_val, 1);
            MAC(part->buf->blue, h_o_val, 2);
            MAC(part->buf->alpha, h_o_val, 3);
#else
/* generic C math */
            rgbval.r = (part->buf->red) * h_o_val;
            rgbval.g = (part->buf->green) * h_o_val;
            rgbval.b = (part->buf->blue) * h_o_val;
            rgbval.a = (part->buf->alpha) * h_o_val;
#endif /* CPU */
            ix += 1;
            /* If this wasn't the last pixel, add the next one to rgbinc. */
            if (LIKELY(ix < (uint32_t)ctx->src->width)) {
                part->buf++;
                part->len--;
                /* Fetch new pixels if needed */
                FILL_BUF(part,ctx->store_part,ctx->args);
                rgbinc.r += part->buf->red;
                rgbinc.g += part->buf->green;
                rgbinc.b += part->buf->blue;
                rgbinc.a += part->buf->alpha;
                /* Add a partial step to rgbval, in this pixel isn't precisely
                   aligned with the new source pixel
                */
#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
                MAC(rgbinc.r, ixe, 0);
                MAC(rgbinc.g, ixe, 1);
                MAC(rgbinc.b, ixe, 2);
                MAC(rgbinc.a, ixe, 3);
#else
/* generic C math */
                rgbval.r += rgbinc.r * ixe;
                rgbval.g += rgbinc.g * ixe;
                rgbval.b += rgbinc.b * ixe;
                rgbval.a += rgbinc.a * ixe;
#endif
            }
#if defined(CPU_COLDFIRE)
/* get final EMAC result out of ACC registers */
            MAC_OUT(rgbval.r, 0);
            MAC_OUT(rgbval.g, 1);
            MAC_OUT(rgbval.b, 2);
            MAC_OUT(rgbval.a, 3);
#endif
            /* Now multiply the color increment to its proper value */
            rgbinc.r *= h_i_val;
            rgbinc.g *= h_i_val;
            rgbinc.b *= h_i_val;
            rgbinc.a *= h_i_val;
        } else {
            rgbval.r += rgbinc.r;
            rgbval.g += rgbinc.g;
            rgbval.b += rgbinc.b;
            rgbval.a += rgbinc.a;
        }
        /* round and scale values, and accumulate or store to output */
        if (accum)
        {
            out_line[ox].r += (rgbval.r + (1 << 21)) >> 22;
            out_line[ox].g += (rgbval.g + (1 << 21)) >> 22;
            out_line[ox].b += (rgbval.b + (1 << 21)) >> 22;
            out_line[ox].a += (rgbval.a + (1 << 21)) >> 22;
        } else {
            out_line[ox].r = (rgbval.r + (1 << 21)) >> 22;
            out_line[ox].g = (rgbval.g + (1 << 21)) >> 22;
            out_line[ox].b = (rgbval.b + (1 << 21)) >> 22;
            out_line[ox].a = (rgbval.a + (1 << 21)) >> 22;
        }
#else
        if (ixe >= h_o_val)
        {
            /* Store the new "current" pixel value in rgbval, and the color
               step value in rgbinc.
            */
            ixe -= h_o_val;
            val = *(part->buf);
            inc = -val;
#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
            MAC(val, h_o_val, 0);
#else
/* generic C math */
            val = val * h_o_val;
#endif
            ix += 1;
            /* If this wasn't the last pixel, add the next one to rgbinc. */
            if (LIKELY(ix < (uint32_t)ctx->src->width)) {
                part->buf++;
                part->len--;
                /* Fetch new pixels if needed */
                FILL_BUF(part,ctx->store_part,ctx->args);
                inc += *(part->buf);
                /* Add a partial step to rgbval, in this pixel isn't precisely
                   aligned with the new source pixel
                */
#if defined(CPU_COLDFIRE)
/* Coldfire EMAC math */
                MAC(inc, ixe, 0);
#else
/* generic C math */
                val += inc * ixe;
#endif
            }
#if defined(CPU_COLDFIRE)
/* get final EMAC result out of ACC register */
            MAC_OUT(val, 0);
#endif
            /* Now multiply the color increment to its proper value */
/* generic C math */
            inc *= h_i_val;
        } else
            val += inc;
        /* round and scale values, and accumulate or store to output */
        if (accum)
        {
            out_line[ox] += (val + (1 << 21)) >> 22;
        } else {
            out_line[ox] = (val + (1 << 21)) >> 22;
        }
#endif
        ixe += h_i_val;
    }
    return true;
}

/* vertical linear scaler */
static inline bool scale_v_linear(struct rowset *rset,
                                  struct scaler_context *ctx)
{
    uint32_t iy, iye;
    int32_t oy;
    const uint32_t v_i_val = ctx->v_i_val,
                   v_o_val = ctx->v_o_val;
    /* Set up our buffers, to store the increment and current value for each
       column, and one temp buffer used to read in new rows.
    */
    uint32_t *rowinc = (uint32_t *)(ctx->buf),
             *rowval = rowinc + ctx->bm->width * CHANNEL_BYTES,
             *rowtmp = rowval + ctx->bm->width * CHANNEL_BYTES,
             *rowinc_px, *rowval_px, *rowtmp_px;

    SDEBUGF("scale_v_linear\n");
    iy = 0;
    iye = v_o_val;
    /* get first scaled row in rowtmp */
    if(!ctx->h_scaler((void*)rowtmp, ctx, false))
        return false;
    for (oy = rset->rowstart; oy != rset->rowstop; oy += rset->rowstep)
    {
        if (iye >= v_o_val)
        {
            iye -= v_o_val;
            iy += 1;
            for(rowinc_px = rowinc, rowtmp_px = rowtmp, rowval_px = rowval;
                rowinc_px < rowval; rowinc_px++, rowtmp_px++, rowval_px++)
            {
                *rowinc_px = -*rowtmp_px;
                *rowval_px = *rowtmp_px * v_o_val;
            }
            if (iy < (uint32_t)ctx->src->height)
            {
                if (!ctx->h_scaler((void*)rowtmp, ctx, false))
                    return false;
                for(rowinc_px = rowinc, rowtmp_px = rowtmp, rowval_px = rowval;
                    rowinc_px < rowval; rowinc_px++, rowtmp_px++, rowval_px++)
                {
                    *rowinc_px += *rowtmp_px;
                    *rowval_px += *rowinc_px * iye;
                    *rowinc_px *= v_i_val;
                }
            }
        } else
            for(rowinc_px = rowinc, rowval_px = rowval; rowinc_px < rowval;
                rowinc_px++, rowval_px++)
                *rowval_px += *rowinc_px;
        ctx->output_row(oy, (void*)rowval, ctx);
        iye += v_i_val;
    }
    return true;
}
#endif /* HAVE_UPSCALER */

#if defined(HAVE_LCD_COLOR) && (defined(HAVE_JPEG) || defined(PLUGIN))
static void output_row_32_native_fromyuv(uint32_t row, void * row_in,
                               struct scaler_context *ctx)
{
#if LCD_STRIDEFORMAT == VERTICAL_STRIDE
#define DEST_STEP   (ctx->bm->height)
#define Y_STEP      (1)
#else
#define DEST_STEP   (1)
#define Y_STEP      (BM_WIDTH(ctx->bm->width,FORMAT_NATIVE,0))
#endif

    int col;
    uint8_t dy = DITHERY(row);
    struct uint32_argb *qp = (struct uint32_argb *)row_in;
    SDEBUGF("output_row: y: %lu in: %p\n",row, row_in);
    fb_data *dest = (fb_data *)ctx->bm->data + Y_STEP * row;
    int delta = 127;
    unsigned r, g, b, y, u, v;
    
    for (col = 0; col < ctx->bm->width; col++) {
        (void) delta;
        if (ctx->dither)
            delta = DITHERXDY(col,dy);
        y = SC_OUT(qp->b, ctx);
        u = SC_OUT(qp->g, ctx);
        v = SC_OUT(qp->r, ctx);
        qp++;
        yuv_to_rgb(y, u, v, &r, &g, &b);
#if LCD_DEPTH < 24
        r = (31 * r + (r >> 3) + delta) >> 8;
        g = (63 * g + (g >> 2) + delta) >> 8;
        b = (31 * b + (b >> 3) + delta) >> 8;
#endif
        *dest = FB_RGBPACK_LCD(r, g, b);
        dest += DEST_STEP;
    }
}
#endif

#if !defined(PLUGIN) || LCD_DEPTH > 1
static void output_row_32_native(uint32_t row, void * row_in,
                              struct scaler_context *ctx)
{
    int col;
    int fb_width = BM_WIDTH(ctx->bm->width,FORMAT_NATIVE,0);
    uint8_t dy = DITHERY(row);
#ifdef HAVE_LCD_COLOR
    struct uint32_argb *qp = (struct uint32_argb*)row_in;
#else
    uint32_t *qp = (uint32_t*)row_in;
#endif
    SDEBUGF("output_row: y: %lu in: %p\n",row, row_in);
#if LCD_DEPTH == 2
#if LCD_PIXELFORMAT == HORIZONTAL_PACKING
                /* greyscale iPods */
                fb_data *dest = (fb_data *)ctx->bm->data + fb_width * row;
                int shift = 6;
                int delta = 127;
                unsigned bright;
                unsigned data = 0;

                for (col = 0; col < ctx->bm->width; col++) {
                    if (ctx->dither)
                        delta = DITHERXDY(col,dy);
                    bright = SC_OUT(*qp++, ctx);
                    bright = (3 * bright + (bright >> 6) + delta) >> 8;
                    data |= (~bright & 3) << shift;
                    shift -= 2;
                    if (shift < 0) {
                        *dest++ = data;
                        data = 0;
                        shift = 6;
                    }
                }
                if (shift < 6)
                    *dest++ = data;
#elif LCD_PIXELFORMAT == VERTICAL_PACKING
                /* iriver H1x0 */
                fb_data *dest = (fb_data *)ctx->bm->data + fb_width *
                                (row >> 2);
                int shift = 2 * (row & 3);
                int delta = 127;
                unsigned bright;

                for (col = 0; col < ctx->bm->width; col++) {
                    if (ctx->dither)
                        delta = DITHERXDY(col,dy);
                    bright = SC_OUT(*qp++, ctx);
                    bright = (3 * bright + (bright >> 6) + delta) >> 8;
                    *dest++ |= (~bright & 3) << shift;
                }
#elif LCD_PIXELFORMAT == VERTICAL_INTERLEAVED
                /* iAudio M3 */
                fb_data *dest = (fb_data *)ctx->bm->data + fb_width *
                                (row >> 3);
                int shift = row & 7;
                int delta = 127;
                unsigned bright;

                for (col = 0; col < ctx->bm->width; col++) {
                    if (ctx->dither)
                        delta = DITHERXDY(col,dy);
                    bright = SC_OUT(*qp++, ctx);
                    bright = (3 * bright + (bright >> 6) + delta) >> 8;
                    *dest++ |= vi_pattern[bright] << shift;
                }
#endif /* LCD_PIXELFORMAT */
#elif LCD_DEPTH >= 16
                /* iriver h300, colour iPods, X5 */
                (void)fb_width;
                fb_data *dest = STRIDE_MAIN((fb_data *)ctx->bm->data + fb_width * row,
                                            (fb_data *)ctx->bm->data + row);
                int delta = 127;
                unsigned r, g, b;
                struct uint32_argb q0;
                /* setup alpha channel buffer */
                unsigned char *bm_alpha = NULL;
                if (ctx->bm->alpha_offset > 0)
                    bm_alpha = ctx->bm->data + ctx->bm->alpha_offset;
                if (bm_alpha)
                    bm_alpha += ALIGN_UP(ctx->bm->width, 2)*row/2;

                for (col = 0; col < ctx->bm->width; col++) {
                    (void) delta;
                    if (ctx->dither)
                        delta = DITHERXDY(col,dy);
                    q0 = *qp++;
                    r = SC_OUT(q0.r, ctx);
                    g = SC_OUT(q0.g, ctx);
                    b = SC_OUT(q0.b, ctx);
#if LCD_DEPTH < 24
                    r = (31 * r + (r >> 3) + delta) >> 8;
                    g = (63 * g + (g >> 2) + delta) >> 8;
                    b = (31 * b + (b >> 3) + delta) >> 8;
#endif
                    *dest = FB_RGBPACK_LCD(r, g, b);
                    dest += STRIDE_MAIN(1, ctx->bm->height);
                    if (bm_alpha) {
                        /* pack alpha channel for 2 pixels into 1 byte */
                        unsigned alpha = SC_OUT(q0.a, ctx);
                        if (col%2)
                            *bm_alpha++ |= alpha&0xf0;
                        else
                            *bm_alpha = alpha>>4;
                    }
                }
#endif /* LCD_DEPTH */
}
#endif

#if defined(PLUGIN) && LCD_DEPTH > 1
unsigned int get_size_native(struct bitmap *bm)
{
    return BM_SIZE(bm->width,bm->height,FORMAT_NATIVE,0);
}

const struct custom_format format_native = {
    .output_row_8 = output_row_8_native,
#if defined(HAVE_LCD_COLOR) && (defined(HAVE_JPEG) || defined(PLUGIN))
    .output_row_32 = {
        output_row_32_native,
        output_row_32_native_fromyuv
    },
#else
    .output_row_32 = output_row_32_native,
#endif
    .get_size = get_size_native
};
#endif

int resize_on_load(struct bitmap *bm, bool dither, struct dim *src,
                   struct rowset *rset, unsigned char *buf, unsigned int len,
                   const struct custom_format *format,
                   IF_PIX_FMT(int format_index,)
                   struct img_part* (*store_part)(void *args),
                   void *args)
{
    const int sw = src->width;
    const int sh = src->height;
    const int dw = bm->width;
    const int dh = bm->height;
    int ret;
    /* buffer for 1 line + 2 spare lines */
#ifdef HAVE_LCD_COLOR
    unsigned int needed = sizeof(struct uint32_argb) * 3 * bm->width;
#else
    unsigned int needed = sizeof(uint32_t) * 3 * bm->width;
#endif
#if MAX_SC_STACK_ALLOC
    uint8_t sc_buf[(needed <= len  || needed > MAX_SC_STACK_ALLOC) ?
                   0 : needed];
#endif
    ALIGN_BUFFER(buf, len, sizeof(uint32_t));
    if (needed > len)
    {
#if MAX_SC_STACK_ALLOC
        if (needed > MAX_SC_STACK_ALLOC)
        {
            DEBUGF("unable to allocate required buffer: %d needed, "
                   "%d available, %d permitted from stack\n",
                   needed, len, MAX_SC_STACK_ALLOC);
            return 0;
        }
        if (sizeof(sc_buf) < needed)
        {
            DEBUGF("failed to allocate large enough buffer on stack: "
                   "%d needed, only got %d",
                   needed, MAX_SC_STACK_ALLOC);
            return 0;
        }
#else
        DEBUGF("unable to allocate required buffer: %d needed, "
               "%d available\n", needed, len);
        return 0;
#endif
    }

    struct scaler_context ctx;
#ifdef HAVE_ADJUSTABLE_CPU_FREQ
    cpu_boost(true);
#endif
    ctx.store_part = store_part;
    ctx.args = args;
#if MAX_SC_STACK_ALLOC
    ctx.buf = needed > len ? sc_buf : buf;
#else
    ctx.buf = buf;
#endif
    ctx.len = len;
    ctx.bm = bm;
    ctx.src = src;
    ctx.dither = dither;
#if !defined(PLUGIN)
#if defined(HAVE_LCD_COLOR) && defined(HAVE_JPEG)
    ctx.output_row = format_index ? output_row_32_native_fromyuv
                                  : output_row_32_native;
#else
    ctx.output_row = output_row_32_native;
#endif
    if (format)
#endif
#ifdef HAVE_LCD_COLOR
        ctx.output_row = format->output_row_32[format_index];
#else
        ctx.output_row = format->output_row_32;
#endif
#ifdef HAVE_UPSCALER
    if (sw > dw)
    {
#endif
        ctx.h_scaler = scale_h_area;
        uint32_t h_div = (1U << 24) / sw;
        ctx.h_i_val = sw * h_div;
        ctx.h_o_val = dw * h_div;
#ifdef HAVE_UPSCALER
    } else {
        ctx.h_scaler = scale_h_linear;
        uint32_t h_div = (1U << 24) / (dw - 1);
        ctx.h_i_val = (sw - 1) * h_div;
        ctx.h_o_val = (dw - 1) * h_div;
    }
#endif
#ifdef CPU_COLDFIRE
    unsigned old_macsr = coldfire_get_macsr();
    coldfire_set_macsr(EMAC_UNSIGNED);
#endif
#ifdef HAVE_UPSCALER
    if (sh > dh)
#endif
    {
        uint32_t v_div = (1U << 22) / sh;
        ctx.v_i_val = sh * v_div;
        ctx.v_o_val = dh * v_div;
        ret = scale_v_area(rset, &ctx);
    }
#ifdef HAVE_UPSCALER
    else
    {
        uint32_t v_div = (1U << 22) / dh;
        ctx.v_i_val = (sh - 1) * v_div;
        ctx.v_o_val = (dh - 1) * v_div;
        ret = scale_v_linear(rset, &ctx);
    }
#endif
#ifdef CPU_COLDFIRE
    /* Restore emac status; other modules like tone control filter
     * calculation may rely on it. */
    coldfire_set_macsr(old_macsr);
#endif
#ifdef HAVE_ADJUSTABLE_CPU_FREQ
    cpu_boost(false);
#endif
    if (!ret)
        return 0;
    return 1;
}