summaryrefslogtreecommitdiff
blob: aa8012acce95d859dad1d56c6a488ba0841d17ae (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
/* Copyright (C) 2001-2019 Artifex Software, Inc.
   All Rights Reserved.

   This software is provided AS-IS with no warranty, either express or
   implied.

   This software is distributed under license and may not be copied,
   modified or distributed except as expressly authorized under the terms
   of the license contained in the file LICENSE in this distribution.

   Refer to licensing information at http://www.artifex.com or contact
   Artifex Software, Inc.,  1305 Grant Avenue - Suite 200, Novato,
   CA 94945, U.S.A., +1(415)492-9861, for further information.
*/


/* Fast monochrome image rendering */
#include "gx.h"
#include "memory_.h"
#include "gpcheck.h"
#include "gsbittab.h"
#include "gserrors.h"
#include "gxfixed.h"
#include "gxarith.h"
#include "gxmatrix.h"
#include "gsccolor.h"
#include "gspaint.h"
#include "gsutil.h"
#include "gxdevice.h"
#include "gxcmap.h"
#include "gxdcolor.h"
#include "gxgstate.h"
#include "gxdevmem.h"
#include "gdevmem.h"		/* for mem_mono_device */
#include "gxcpath.h"
#include "gximage.h"
#include "gzht.h"

#include "valgrind.h"

/* Conditionally include statistics code. */
#if defined(DEBUG) && !defined(GS_THREADSAFE)
#  define STATS
#endif

/* ------ Strategy procedure ------ */

/* Check the prototype. */
iclass_proc(gs_image_class_1_simple);

/* Use special fast logic for portrait or landscape black-and-white images. */
static irender_proc(image_render_skip);
static irender_proc(image_render_simple);
static irender_proc(image_render_landscape);

int
gs_image_class_1_simple(gx_image_enum * penum, irender_proc_t *render_fn)
{
    fixed ox = dda_current(penum->dda.pixel0.x);
    fixed oy = dda_current(penum->dda.pixel0.y);

    if (penum->use_rop || penum->spp != 1 || penum->bps != 1)
        return 0;
    switch (penum->posture) {
        case image_portrait:
            {			/* Use fast portrait algorithm. */
                long dev_width =
                    fixed2long_pixround(ox + penum->x_extent.x) -
                    fixed2long_pixround(ox);

                if (dev_width != penum->rect.w) {
                    /*
                     * Add an extra align_bitmap_mod of padding so that
                     * we can align scaled rows with the device.
                     */
                    long line_size =
                        bitmap_raster(any_abs(dev_width)) + align_bitmap_mod;

                    if (penum->adjust != 0 || line_size > max_uint)
                        return 0;
                    /* Must buffer a scan line. */
                    penum->line_width = any_abs(dev_width);
                    penum->line_size = (uint) line_size;
                    penum->line = gs_alloc_bytes(penum->memory,
                                            penum->line_size, "image line");
                    if (penum->line == 0) {
                        return gs_error_VMerror;
                    }
                }
                if_debug2m('b', penum->memory, "[b]render=simple, unpack=copy; rect.w=%d, dev_width=%ld\n",
                           penum->rect.w, dev_width);
                *render_fn = image_render_simple;
                break;
            }
        case image_landscape:
            {			/* Use fast landscape algorithm. */
                long dev_width =
                    fixed2long_pixround(oy + penum->x_extent.y) -
                    fixed2long_pixround(oy);
                long line_size =
                    (dev_width = any_abs(dev_width),
                     bitmap_raster(dev_width) * 8 +
                     ROUND_UP(dev_width, 8) * align_bitmap_mod);

                if ((dev_width != penum->rect.w && penum->adjust != 0) ||
                    line_size > max_uint
                    )
                    return 0;
                /* Must buffer a group of 8N scan lines. */
                penum->line_width = dev_width;
                penum->line_size = (uint) line_size;
                penum->line = gs_alloc_bytes(penum->memory,
                                             penum->line_size, "image line");
                if (penum->line == 0) {
                    return gs_error_VMerror;
                }
#ifdef PACIFY_VALGRIND
                memset(penum->line, 0, penum->line_size); /* For the number of scan lined < 8 */
#endif
                penum->xi_next = penum->line_xy = fixed2int_var_rounded(ox);
                if_debug3m('b', penum->memory,
                           "[b]render=landscape, unpack=copy; rect.w=%d, dev_width=%ld, line_size=%ld\n",
                           penum->rect.w, dev_width, line_size);
                *render_fn = image_render_landscape;
                /* Precompute values needed for rasterizing. */
                penum->dxy =
                    float2fixed(penum->matrix.xy +
                                fixed2float(fixed_epsilon) / 2);
                break;
            }
        default:
            return 0;
    }
    /* Precompute values needed for rasterizing. */
    penum->dxx =
        float2fixed(penum->matrix.xx + fixed2float(fixed_epsilon) / 2);
    /*
     * We don't want to spread the samples, but we have to reset unpack_bps
     * to prevent the buffer pointer from being incremented by 8 bytes per
     * input byte.
     */
    penum->unpack = sample_unpack_copy;
    penum->unpack_bps = 8;
    if (penum->use_mask_color) {
        /*
         * Set the masked color as 'no_color' to make it transparent
         *  according to the mask color range and the decoding.
         */
        penum->masked = true;
        if (penum->mask_color.values[0] == 1) {
            /* if v0 == 1, 1 is transparent since v1 must be == 1 to be a valid range */
            set_nonclient_dev_color(penum->map[0].inverted ? penum->icolor0 : penum->icolor1,
                        gx_no_color_index);
        } else if (penum->mask_color.values[1] == 0) {
            /* if v1 == 0, 0 is transparent since v0 must be == 0 to be a valid range */
            set_nonclient_dev_color(penum->map[0].inverted ? penum->icolor1 : penum->icolor0,
                        gx_no_color_index);
        } else {
            /*
             * The only other possible in-range value is v0 = 0, v1 = 1.
             * The image is completely transparent!
             */
            *render_fn = image_render_skip;
        }
        penum->map[0].decoding = sd_none;
    }
    return 0;
}

/* ------ Rendering procedures ------ */

#define DC_IS_NULL(pdc)\
  (gx_dc_is_pure(pdc) && (pdc)->colors.pure == gx_no_color_index)

/* Skip over a completely transparent image. */
static int
image_render_skip(gx_image_enum * penum, const byte * buffer, int data_x,
                  uint w, int h, gx_device * dev)
{
    return h;
}

/*
 * Scale (and possibly reverse) one scan line of a monobit image.
 * This is used for both portrait and landscape image processing.
 * We pass in an x offset (0 <= line_x < align_bitmap_mod * 8) so that
 * we can align the result with the eventual device X.
 *
 * To be precise, the input to this routine is the w bits starting at
 * bit data_x in buffer.  These w bits expand to abs(x_extent) bits,
 * either inverted (zero = 0xff) or not (zero = 0), starting at bit
 * line_x in line which corresponds to coordinate
 * fixed2int_pixround(xcur + min(x_extent, 0)).  Note that the entire
 * bytes containing the first and last output bits are affected: the
 * other bits in those bytes are set to zero (i.e., the value of the
 * 'zero' argument).
 */
#ifdef STATS
struct stats_image_fast_s {
    long
         calls, all0s, all1s, runs, lbit0, byte00, byte01, byte02, byte03,
         byte04, rbit0, lbit1, byte1, rbit1, thin, thin2, nwide, bwide,
         nfill, bfill;
} stats_image_fast;
#  define INCS(stat) ++stats_image_fast.stat
#  define ADDS(stat, n) stats_image_fast.stat += n
#else
#  define INCS(stat) DO_NOTHING
#  define ADDS(stat, n) DO_NOTHING
#endif
static inline void
fill_row(byte *line, int line_x, uint raster, int value)
{
    memset(line + (line_x >> 3), value, raster - (line_x >> 3));
}

static void
image_simple_expand(byte * line, int line_x, uint raster,
                    const byte * buffer, int data_x, uint w,
                    fixed xcur, fixed x_extent, byte zero /* 0 or 0xff */ )
{
    int dbitx = data_x & 7;
    byte sbit = 0x80 >> dbitx;
    byte sbitmask = 0xff >> dbitx;
    uint wx = dbitx + w;
    gx_dda_fixed xl;
    gx_dda_step_fixed dxx4, dxx8, dxx16, dxx24, dxx32;
    register const byte *psrc = buffer + (data_x >> 3);

    /*
     * The following 3 variables define the end of the input data row.
     * We would put them in a struct, except that no compiler that we
     * know of will optimize individual struct members as though they
     * were simple variables (e.g., by putting them in registers).
     *
     * endp points to the byte that contains the bit just beyond the
     * end of the row.  endx gives the bit number of this bit within
     * the byte, with 0 being the *least* significant bit.  endbit is
     * a mask for this bit.
     */
    const byte *endp = psrc + (wx >> 3);
    int endx = ~wx & 7;
    byte endbit = 1 << endx;

    /*
     * The following 3 variables do the same for start of the last run
     * of the input row (think of it as a pointer to just beyond the
     * end of the next-to-last run).
     */
    const byte *stop = endp;
    int stopx;
    byte stopbit = endbit;
    byte data;
    byte one = ~zero;
    fixed xl0;
#ifdef PACIFY_VALGRIND
    byte vbits;
#endif

    if (w == 0)
        return;
    INCS(calls);

    /* Scan backward for the last transition. */
    if (stopbit == 0x80)
        --stop, stopbit = 1;
    else
        stopbit <<= 1;
    /* Now (stop, stopbit) give the last bit of the row. */
#ifdef PACIFY_VALGRIND
    /* Here, we are dealing with a row of bits, rather than bytes.
     * If the width of the bits is not a multiple of 8, we don't
     * fill out the last byte, and valgrind (correctly) tracks the
     * bits in that last byte as being a mix of defined and undefined.
     * When we are scanning through the row bitwise, everything works
     * fine, but our "skip whole bytes" code can confuse valgrind.
     * We know that we won't match the "data == 0xff" for the final
     * byte (i.e. the undefinedness of some of the bits doesn't matter
     * to the correctness of the routine), but valgrind is not smart
     * enough to realise that we know this. Accordingly, we get a false
     * positive "undefined memory read".
     * How do we fix this? Well, one way would be to read in the
     * partial last byte, and explicitly set the undefined bits to
     * be 0.
     *   *stop &= ~(stopbit-1);
     * Unfortunately, stop is a const *, so we can't do that (we could
     * break const, but it is just conceivable that the caller might
     * pass the next string of bits out in a later call, and so we
     * might be corrupting valid data).
     * Instead, we make a call to a valgrind helper. */
    VALGRIND_GET_VBITS(stop,&vbits,1);
    if ((vbits & stopbit)==0) { /* At least our stop bit must be addressable already! */
      byte zero = 0;
      VALGRIND_SET_VBITS(stop,&zero,1);
    }
#endif
    {
        byte stopmask = -stopbit << 1;
        byte last = *stop;

        if (stop == psrc)	/* only 1 input byte */
            stopmask &= sbitmask;
        if (last & stopbit) {
            /* The last bit is a 1: look for a 0-to-1 transition. */
            if (~last & stopmask) {	/* Transition in last byte. */
                last |= stopbit - 1;
            } else {		/* No transition in the last byte. */
                while (stop > psrc && stop[-1] == 0xff)
                    --stop;
                if (stop == psrc ||
                    (stop == psrc + 1 && !(~*psrc & sbitmask))
                    ) {
                    /* The input is all 1s.  Clear the row and exit. */
                    INCS(all1s);
                    fill_row(line, line_x, raster, one);
                    goto end;
                }
                last = *--stop;
            }
            stopx = byte_bit_run_length_0[byte_reverse_bits[last]] - 1;
        } else {
            /* The last bit is a 0: look for a 1-to-0 transition. */
            if (last & stopmask) {	/* Transition in last byte. */
                last &= -stopbit;
            } else {		/* No transition in the last byte. */
                while (stop > psrc && stop[-1] == 0)
                    --stop;
                if (stop == psrc ||
                    (stop == psrc + 1 && !(*psrc & sbitmask))
                    ) {
                    /* The input is all 0s.  Clear the row and exit. */
                    INCS(all0s);
                    fill_row(line, line_x, raster, zero);
                    goto end;
                }
                last = *--stop;
            }
            stopx = byte_bit_run_length_0[byte_reverse_bits[last ^ 0xff]] - 1;
        }
        if (stopx < 0)
            stopx = 7, ++stop;
        stopbit = 1 << stopx;
    }

    /* Pre-clear the row. */
    fill_row(line, line_x, raster, zero);


    /* Extreme negative values of x_extent cause the xl0 calculation
     * to explode. Workaround this here. */
    if (x_extent < min_int + 0x100)
      x_extent += 0x100;

    /* Set up the DDAs. */
    xl0 =
        (x_extent >= 0 ?
         fixed_fraction(fixed_pre_pixround(xcur)) :
         fixed_fraction(fixed_pre_pixround(xcur + x_extent)) - x_extent);
    xl0 += int2fixed(line_x);
    /* We should never get a negative x10 here. If we do, all bets are off. */
    if (xl0 < 0)
        xl0 = 0, x_extent = 0;
    dda_init(xl, xl0, x_extent, w);
    dxx4 = xl.step;
    dda_step_add(dxx4, xl.step);
    /* egcc - 2.91.66 generates incorrect code for
     * dda_step_add(dxx4, dxx4);
     * Using the temp variable.
     */
    dxx8 = dxx4;
    dda_step_add(dxx4, dxx8);
    dxx8 = dxx4;
    dda_step_add(dxx8, dxx4);
    dxx16 = dxx8;
    dda_step_add(dxx16, dxx8);
    dxx24 = dxx16;
    dda_step_add(dxx24, dxx8);
    dxx32 = dxx24;
    dda_step_add(dxx32, dxx8);

    /*
     * Loop invariants:
     *      data = *psrc;
     *      sbit = 1 << n, 0<=n<=7.
     */
    for (data = *psrc;;) {
        int x0, n, bit;
        byte *bp;
        static const byte lmasks[9] = {
            0xff, 0x7f, 0x3f, 0x1f, 0xf, 7, 3, 1, 0
        };
        static const byte rmasks[9] = {
            0, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff
        };

        INCS(runs);

        /* Scan a run of zeros. */
        data ^= 0xff;		/* invert */
        while (data & sbit) {
            dda_next(xl);
            sbit >>= 1;
            INCS(lbit0);
        }
        if (!sbit) {		/* Scan a run of zero bytes. */
sw:	    if ((data = psrc[1]) != 0) {
                psrc++;
                INCS(byte00);
            } else if ((data = psrc[2]) != 0) {
                dda_state_next(xl.state, dxx8);
                psrc += 2;
                INCS(byte01);
            } else if ((data = psrc[3]) != 0) {
                dda_state_next(xl.state, dxx16);
                psrc += 3;
                INCS(byte02);
            } else if ((data = psrc[4]) != 0) {
                dda_state_next(xl.state, dxx24);
                psrc += 4;
                INCS(byte03);
            } else {
                dda_state_next(xl.state, dxx32);
                psrc += 4;
                INCS(byte04);
                goto sw;
            }
            if (data > 0xf)
                sbit = 0x80;
            else {
                sbit = 0x08;
                dda_state_next(xl.state, dxx4);
            }
            data ^= 0xff;	/* invert */
            while (data & sbit) {
                dda_next(xl);
                sbit >>= 1;
                INCS(rbit0);
            }
        }
        x0 = dda_current_fixed2int(xl);
        if (psrc >= stop && sbit == stopbit) {
            /*
             * We've scanned the last run of 0s.
             * Prepare to fill the final run of 1s.
             * Use int64_t to avoid overflow.
             */
            n = fixed2int((int64_t)xl0 + (int64_t)x_extent) - x0;
        } else {		/* Scan a run of ones. */
            /* We know the current bit is a one. */
            data ^= 0xff;	/* un-invert */
            do {
                dda_next(xl);
                sbit >>= 1;
                INCS(lbit1);
            }
            while (data & sbit);
            if (!sbit) {	/* Scan a run of 0xff bytes. */
                while ((data = *++psrc) == 0xff) {
                    dda_state_next(xl.state, dxx8);
                    INCS(byte1);
                }
                if (data < 0xf0)
                    sbit = 0x80;
                else {
                    sbit = 0x08;
                    dda_state_next(xl.state, dxx4);
                }
                while (data & sbit) {
                    dda_next(xl);
                    sbit >>= 1;
                    INCS(rbit1);
                }
            }
            n = dda_current_fixed2int(xl) - x0;
        }

        /* Fill the run in the scan line. */
        if (n < 0)
            x0 += n, n = -n;
        bp = line + (x0 >> 3);
        bit = x0 & 7;
        if ((n += bit) <= 8) {
            *bp ^= lmasks[bit] - lmasks[n];
            INCS(thin);
        } else if ((n -= 8) <= 8) {
            *bp ^= lmasks[bit];
            bp[1] ^= rmasks[n];
            INCS(thin2);
        } else {
            *bp++ ^= lmasks[bit];
            if (n >= 56) {
                int nb = n >> 3;

                memset(bp, one, nb);
                bp += nb;
                INCS(nwide);
                ADDS(bwide, nb);
            } else {
                ADDS(bfill, n >> 3);
                while ((n -= 8) >= 0)
                    *bp++ = one;
                INCS(nfill);
            }
            *bp ^= rmasks[n & 7];
        }
        if (psrc >= stop && sbit == stopbit)
            break;
    }
 end:
    VALGRIND_SET_VBITS(stop,&vbits,1);
}

/* Copy one rendered scan line to the device. */
static int
copy_portrait(gx_image_enum * penum, const byte * data, int dx, int raster,
              int x, int y, int w, int h, gx_device * dev)
{
    const gx_device_color *pdc0;
    const gx_device_color *pdc1;
    uint align = ALIGNMENT_MOD(data, align_bitmap_mod);

    /*
     * We know that the lookup table maps 1 bit to 1 bit,
     * so it can only have 2 states: straight-through or invert.
     */
    if (penum->map[0].table.lookup4x1to32[0])
        pdc0 = penum->icolor1, pdc1 = penum->icolor0;
    else
        pdc0 = penum->icolor0, pdc1 = penum->icolor1;
    data -= align;
    dx += align << 3;
    if (gx_dc_is_pure(pdc0) && gx_dc_is_pure(pdc1)) {
        /* Just use copy_mono. */
        dev_proc_copy_mono((*copy_mono)) =
            (h == 1 || (raster & (align_bitmap_mod - 1)) == 0 ?
             dev_proc(dev, copy_mono) : gx_copy_mono_unaligned);
        return (*copy_mono)
            (dev, data, dx, raster, gx_no_bitmap_id,
             x, y, w, h, pdc0->colors.pure, pdc1->colors.pure);
    }
    /*
     * At least one color isn't pure: if the other one is transparent, use
     * the opaque color's fill_masked procedure.  Note that we use a
     * slightly unusual representation for transparent here (per
     * gx_begin_image1): a pure color with pixel value gx_no_color_index.
     */
    {
        const gx_device_color *pdc;
        bool invert;

        if (DC_IS_NULL(pdc1)) {
            pdc = pdc0;
            invert = true;
        } else {
            if (!DC_IS_NULL(pdc0)) {
                int code = gx_device_color_fill_rectangle
                    (pdc0, x, y, w, h, dev, lop_default, NULL);

                if (code < 0)
                    return code;
            }
            pdc = pdc1;
            invert = false;
        }
        return (*pdc->type->fill_masked)
            (pdc, data, dx, raster, gx_no_bitmap_id, x, y, w, h,
             dev, lop_default, invert);

    }
}

/* Rendering procedure for a monobit image with no */
/* skew or rotation and pure colors. */
static int
image_render_simple(gx_image_enum * penum, const byte * buffer, int data_x,
                    uint w, int h, gx_device * dev)
{
    dev_proc_copy_mono((*copy_mono)) = dev_proc(dev, copy_mono);
    const fixed dxx = penum->dxx;
    const byte *line;
    uint line_width, line_size;
    int line_x;
    fixed xcur = dda_current(penum->dda.pixel0.x);
    int ix = fixed2int_pixround(xcur);
    int ixr;
    const int iy = penum->yci, ih = penum->hci;
    gx_device_color * const pdc0 = penum->icolor0;
    gx_device_color * const pdc1 = penum->icolor1;
    int dy;
    int code;

    if (h == 0)
        return 0;
    if ((!DC_IS_NULL(pdc0) &&
         (code = gx_color_load(pdc0, penum->pgs, dev)) < 0) ||
        (!DC_IS_NULL(pdc1) &&
         (code = gx_color_load(pdc1, penum->pgs, dev)) < 0)
        )
        return code;
    if (penum->line == 0) {	/* A direct BitBlt is possible. */
        line = buffer;
        line_size = (w + 7) >> 3;
        line_width = w;
        line_x = 0;
    } else if (copy_mono == dev_proc(&mem_mono_device, copy_mono) &&
               dxx > 0 && gx_dc_is_pure(pdc1) && gx_dc_is_pure(pdc0) &&
               /* We know the colors must be (0,1) or (1,0). */
               (pdc0->colors.pure ^ pdc1->colors.pure) == 1 &&
               !penum->clip_image &&
               /*
                * Even if clip_image is false, the clipping rectangle
                * might lie partly outside the device coordinate space
                * if the Margins values are non-zero.
                */
               ix >= 0 &&
               (ixr = fixed2int_pixround(xcur + penum->x_extent.x) - 1) <
                 dev->width &&
               iy >= 0 && iy + ih <= dev->height
        ) {
        /* Do the operation directly into the memory device bitmap. */
        int line_ix;
        int ib_left = ix >> 3, ib_right = ixr >> 3;
        byte *scan_line = scan_line_base((gx_device_memory *) dev, iy);
        byte save_left, save_right, mask;

        line_x = ix & (align_bitmap_mod * 8 - 1);
        line_ix = ix - line_x;
        line_size = (ixr >> 3) + 1 - (line_ix >> 3);
        line_width = ixr + 1 - ix;
        /* We must save and restore any unmodified bits in */
        /* the two edge bytes. */
        save_left = scan_line[ib_left];
        save_right = scan_line[ib_right];
        image_simple_expand(scan_line + (line_ix >> 3), line_x,
                            line_size, buffer, data_x, w, xcur,
                            penum->x_extent.x,
                            (byte)((pdc0->colors.pure == 0) !=
                             (penum->map[0].table.lookup4x1to32[0] == 0) ?
                             0xff : 0));
        if (ix & 7)
            mask = (byte) (0xff00 >> (ix & 7)),
                scan_line[ib_left] =
                (save_left & mask) + (scan_line[ib_left] & ~mask);
        if ((ixr + 1) & 7)
            mask = (byte) (0xff00 >> ((ixr + 1) & 7)),
                scan_line[ib_right] =
                (scan_line[ib_right] & mask) + (save_right & ~mask);
        if (ih <= 1)
            return 1;
        /****** MAY BE UNALIGNED ******/
        line = scan_line + (line_ix >> 3);
        for (dy = 1; dy < ih; dy++) {
            int code = (*copy_mono)
                (dev, line, line_x, line_size, gx_no_bitmap_id,
                 ix, iy + dy, line_width, 1,
                 (gx_color_index)0, (gx_color_index)1);

            if (code < 0)
                return code;
        }
        return 0;
    } else {
        line = penum->line;
        line_size = penum->line_size;
        line_width = penum->line_width;
        line_x = ix & (align_bitmap_mod * 8 - 1);
        image_simple_expand(penum->line, line_x, line_size,
                            buffer, data_x, w, xcur,
                            penum->x_extent.x, 0);
    }

    /* Finally, transfer the scan line to the device. */
    if (dxx < 0)
        ix -= line_width;
    for (dy = 0; dy < ih; dy++) {
        int code = copy_portrait(penum, line, line_x, line_size,
                                 ix, iy + dy, line_width, 1, dev);

        if (code < 0)
            return code;
    }

    return 1;
}

/* Rendering procedure for a 90 degree rotated monobit image */
/* with pure colors.  We buffer and then flip 8 scan lines at a time. */
static int copy_landscape(gx_image_enum *, int, int, bool, gx_device *);
static int
image_render_landscape(gx_image_enum * penum, const byte * buffer, int data_x,
                       uint w, int h, gx_device * dev)
{
    byte *line = penum->line;
    uint raster = bitmap_raster(penum->line_width);
    int ix = penum->xci, iw = penum->wci;
    int xinc, xmod;
    byte *row;
    const byte *orig_row = 0;
    bool y_neg = penum->dxy < 0;

    if (is_fneg(penum->matrix.yx))
        ix += iw, iw = -iw, xinc = -1;
    else
        xinc = 1;
    /*
     * Because of clipping, there may be discontinuous jumps in the values
     * of ix (xci).  If this happens, or if we are at the end of the data or
     * a client has requested flushing, flush the flipping buffer.
     */
    if (ix != penum->xi_next || h == 0) {
        int xi = penum->xi_next;
        int code =
            (xinc > 0 ?
             copy_landscape(penum, penum->line_xy, xi, y_neg, dev) :
             copy_landscape(penum, xi, penum->line_xy, y_neg, dev));

        if (code < 0)
            return code;
        penum->line_xy = penum->xi_next = ix;
        if (h == 0)
            return code;
    }
    for (; iw != 0; iw -= xinc) {
        if (xinc < 0)
            --ix;
        xmod = ix & 7;
        row = line + xmod * raster;
        if (orig_row == 0) {
            image_simple_expand(row, 0, raster,
                                buffer, data_x, w,
                                dda_current(penum->dda.pixel0.y),
                                penum->x_extent.y, 0);
            orig_row = row;
        } else
            memcpy(row, orig_row, raster);
        if (xinc > 0) {
            ++ix;
            if (xmod == 7) {
                int code =
                    copy_landscape(penum, penum->line_xy, ix, y_neg, dev);

                if (code < 0)
                    return code;
                orig_row = 0;
                penum->line_xy = ix;
            }
        } else {
            if (xmod == 0) {
                int code =
                    copy_landscape(penum, ix, penum->line_xy, y_neg, dev);

                if (code < 0)
                    return code;
                orig_row = 0;
                penum->line_xy = ix;
            }
        }
    }
    penum->xi_next = ix;
    return 0;
}

/* Flip and copy one group of scan lines. */
static int
copy_landscape(gx_image_enum * penum, int x0, int x1, bool y_neg,
               gx_device * dev)
{
    byte *line = penum->line;
    uint line_width = penum->line_width;
    uint raster = bitmap_raster(line_width);
    byte *flipped = line + raster * 8;
    int w = x1 - x0;
    int y = fixed2int_pixround(dda_current(penum->dda.pixel0.y));

    if (w == 0 || line_width == 0)
        return 0;
    /* Flip the buffered data from raster x 8 to align_bitmap_mod x */
    /* line_width. */
    if (line_width > 0) {
        int i = (line_width-1)>>3;

#ifdef PACIFY_VALGRIND
        if (line_width & 7) {
            memflip8x8_eol(line + i, raster,
                           flipped + (i << (log2_align_bitmap_mod + 3)),
                           align_bitmap_mod,
                           line_width & 7);
            i--;
        }
#endif

        for (; i >= 0; --i)
            memflip8x8(line + i, raster,
                       flipped + (i << (log2_align_bitmap_mod + 3)),
                       align_bitmap_mod);
    }
    /* Transfer the scan lines to the device. */
    if (w < 0)
        x0 = x1, w = -w;
    if (y_neg)
        y -= line_width;
    return copy_portrait(penum, flipped, x0 & 7, align_bitmap_mod,
                         x0, y, w, line_width, dev);
}