3rdparty/libjpeg/jdmainct.c

/*
 * jdmainct.c
 *
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
 * This file contains the main buffer controller for decompression.
 * The main buffer lies between the JPEG decompressor proper and the
 * post-processor; it holds downsampled data in the JPEG colorspace.
 *
 * Note that this code is bypassed in raw-data mode, since the application
 * supplies the equivalent of the main buffer in that case.
 */

#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"


/*
 * In the current system design, the main buffer need never be a full-image
 * buffer; any full-height buffers will be found inside the coefficient or
 * postprocessing controllers.  Nonetheless, the main controller is not
 * trivial.  Its responsibility is to provide context rows for upsampling/
 * rescaling, and doing this in an efficient fashion is a bit tricky.
 *
 * Postprocessor input data is counted in "row groups".  A row group
 * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
 * sample rows of each component.  (We require DCT_scaled_size values to be
 * chosen such that these numbers are integers.  In practice DCT_scaled_size
 * values will likely be powers of two, so we actually have the stronger
 * condition that DCT_scaled_size / min_DCT_scaled_size is an integer.)
 * Upsampling will typically produce max_v_samp_factor pixel rows from each
 * row group (times any additional scale factor that the upsampler is
 * applying).
 *
 * The coefficient controller will deliver data to us one iMCU row at a time;
 * each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or
 * exactly min_DCT_scaled_size row groups.  (This amount of data corresponds
 * to one row of MCUs when the image is fully interleaved.)  Note that the
 * number of sample rows varies across components, but the number of row
 * groups does not.  Some garbage sample rows may be included in the last iMCU
 * row at the bottom of the image.
 *
 * Depending on the vertical scaling algorithm used, the upsampler may need
 * access to the sample row(s) above and below its current input row group.
 * The upsampler is required to set need_context_rows TRUE at global selection
 * time if so.  When need_context_rows is FALSE, this controller can simply
 * obtain one iMCU row at a time from the coefficient controller and dole it
 * out as row groups to the postprocessor.
 *
 * When need_context_rows is TRUE, this controller guarantees that the buffer
 * passed to postprocessing contains at least one row group's worth of samples
 * above and below the row group(s) being processed.  Note that the context
 * rows "above" the first passed row group appear at negative row offsets in
 * the passed buffer.  At the top and bottom of the image, the required
 * context rows are manufactured by duplicating the first or last real sample
 * row; this avoids having special cases in the upsampling inner loops.
 *
 * The amount of context is fixed at one row group just because that's a
 * convenient number for this controller to work with.  The existing
 * upsamplers really only need one sample row of context.  An upsampler
 * supporting arbitrary output rescaling might wish for more than one row
 * group of context when shrinking the image; tough, we don't handle that.
 * (This is justified by the assumption that downsizing will be handled mostly
 * by adjusting the DCT_scaled_size values, so that the actual scale factor at
 * the upsample step needn't be much less than one.)
 *
 * To provide the desired context, we have to retain the last two row groups
 * of one iMCU row while reading in the next iMCU row.  (The last row group
 * can't be processed until we have another row group for its below-context,
 * and so we have to save the next-to-last group too for its above-context.)
 * We could do this most simply by copying data around in our buffer, but
 * that'd be very slow.  We can avoid copying any data by creating a rather
 * strange pointer structure.  Here's how it works.  We allocate a workspace
 * consisting of M+2 row groups (where M = min_DCT_scaled_size is the number
 * of row groups per iMCU row).  We create two sets of redundant pointers to
 * the workspace.  Labeling the physical row groups 0 to M+1, the synthesized
 * pointer lists look like this:
 *                   M+1                          M-1
 * master pointer --> 0         master pointer --> 0
 *                    1                            1
 *                   ...                          ...
 *                   M-3                          M-3
 *                   M-2                           M
 *                   M-1                          M+1
 *                    M                           M-2
 *                   M+1                          M-1
 *                    0                            0
 * We read alternate iMCU rows using each master pointer; thus the last two
 * row groups of the previous iMCU row remain un-overwritten in the workspace.
 * The pointer lists are set up so that the required context rows appear to
 * be adjacent to the proper places when we pass the pointer lists to the
 * upsampler.
 *
 * The above pictures describe the normal state of the pointer lists.
 * At top and bottom of the image, we diddle the pointer lists to duplicate
 * the first or last sample row as necessary (this is cheaper than copying
 * sample rows around).
 *
 * This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1.  In that
 * situation each iMCU row provides only one row group so the buffering logic
 * must be different (eg, we must read two iMCU rows before we can emit the
 * first row group).  For now, we simply do not support providing context
 * rows when min_DCT_scaled_size is 1.  That combination seems unlikely to
 * be worth providing --- if someone wants a 1/8th-size preview, they probably
 * want it quick and dirty, so a context-free upsampler is sufficient.
 */


/* Private buffer controller object */

typedef struct {
  struct jpeg_d_main_controller pub; /* public fields */

  /* Pointer to allocated workspace (M or M+2 row groups). */
  JSAMPARRAY buffer[MAX_COMPONENTS];

  boolean buffer_full;          /* Have we gotten an iMCU row from decoder? */
  JDIMENSION rowgroup_ctr;      /* counts row groups output to postprocessor */

  /* Remaining fields are only used in the context case. */

  /* These are the master pointers to the funny-order pointer lists. */
  JSAMPIMAGE xbuffer[2];        /* pointers to weird pointer lists */

  int whichptr;                 /* indicates which pointer set is now in use */
  int context_state;            /* process_data state machine status */
  JDIMENSION rowgroups_avail;   /* row groups available to postprocessor */
  JDIMENSION iMCU_row_ctr;      /* counts iMCU rows to detect image top/bot */
} my_main_controller;

typedef my_main_controller * my_main_ptr;

/* context_state values: */
#define CTX_PREPARE_FOR_IMCU    0       /* need to prepare for MCU row */
#define CTX_PROCESS_IMCU        1       /* feeding iMCU to postprocessor */
#define CTX_POSTPONED_ROW       2       /* feeding postponed row group */


/* Forward declarations */
METHODDEF(void) process_data_simple_main
        JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
             JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
METHODDEF(void) process_data_context_main
        JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
             JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
#ifdef QUANT_2PASS_SUPPORTED
METHODDEF(void) process_data_crank_post
        JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
             JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
#endif


LOCAL(void)
alloc_funny_pointers (j_decompress_ptr cinfo)
/* Allocate space for the funny pointer lists.
 * This is done only once, not once per pass.
 */
{
  my_main_ptr main = (my_main_ptr) cinfo->main;
  int ci, rgroup;
  int M = cinfo->min_DCT_v_scaled_size;
  jpeg_component_info *compptr;
  JSAMPARRAY xbuf;

  /* Get top-level space for component array pointers.
   * We alloc both arrays with one call to save a few cycles.
   */
  main->xbuffer[0] = (JSAMPIMAGE)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
                                cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
  main->xbuffer[1] = main->xbuffer[0] + cinfo->num_components;

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
      cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
    /* Get space for pointer lists --- M+4 row groups in each list.
     * We alloc both pointer lists with one call to save a few cycles.
     */
    xbuf = (JSAMPARRAY)
      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
                                  2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
    xbuf += rgroup;             /* want one row group at negative offsets */
    main->xbuffer[0][ci] = xbuf;
    xbuf += rgroup * (M + 4);
    main->xbuffer[1][ci] = xbuf;
  }
}


LOCAL(void)
make_funny_pointers (j_decompress_ptr cinfo)
/* Create the funny pointer lists discussed in the comments above.
 * The actual workspace is already allocated (in main->buffer),
 * and the space for the pointer lists is allocated too.
 * This routine just fills in the curiously ordered lists.
 * This will be repeated at the beginning of each pass.
 */
{
  my_main_ptr main = (my_main_ptr) cinfo->main;
  int ci, i, rgroup;
  int M = cinfo->min_DCT_v_scaled_size;
  jpeg_component_info *compptr;
  JSAMPARRAY buf, xbuf0, xbuf1;

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
      cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
    xbuf0 = main->xbuffer[0][ci];
    xbuf1 = main->xbuffer[1][ci];
    /* First copy the workspace pointers as-is */
    buf = main->buffer[ci];
    for (i = 0; i < rgroup * (M + 2); i++) {
      xbuf0[i] = xbuf1[i] = buf[i];
    }
    /* In the second list, put the last four row groups in swapped order */
    for (i = 0; i < rgroup * 2; i++) {
      xbuf1[rgroup*(M-2) + i] = buf[rgroup*M + i];
      xbuf1[rgroup*M + i] = buf[rgroup*(M-2) + i];
    }
    /* The wraparound pointers at top and bottom will be filled later
     * (see set_wraparound_pointers, below).  Initially we want the "above"
     * pointers to duplicate the first actual data line.  This only needs
     * to happen in xbuffer[0].
     */
    for (i = 0; i < rgroup; i++) {
      xbuf0[i - rgroup] = xbuf0[0];
    }
  }
}


LOCAL(void)
set_wraparound_pointers (j_decompress_ptr cinfo)
/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
 * This changes the pointer list state from top-of-image to the normal state.
 */
{
  my_main_ptr main = (my_main_ptr) cinfo->main;
  int ci, i, rgroup;
  int M = cinfo->min_DCT_v_scaled_size;
  jpeg_component_info *compptr;
  JSAMPARRAY xbuf0, xbuf1;

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
      cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
    xbuf0 = main->xbuffer[0][ci];
    xbuf1 = main->xbuffer[1][ci];
    for (i = 0; i < rgroup; i++) {
      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
    }
  }
}


LOCAL(void)
set_bottom_pointers (j_decompress_ptr cinfo)
/* Change the pointer lists to duplicate the last sample row at the bottom
 * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
 * Also sets rowgroups_avail to indicate number of nondummy row groups in row.
 */
{
  my_main_ptr main = (my_main_ptr) cinfo->main;
  int ci, i, rgroup, iMCUheight, rows_left;
  jpeg_component_info *compptr;
  JSAMPARRAY xbuf;

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    /* Count sample rows in one iMCU row and in one row group */
    iMCUheight = compptr->v_samp_factor * compptr->DCT_v_scaled_size;
    rgroup = iMCUheight / cinfo->min_DCT_v_scaled_size;
    /* Count nondummy sample rows remaining for this component */
    rows_left = (int) (compptr->downsampled_height % (JDIMENSION) iMCUheight);
    if (rows_left == 0) rows_left = iMCUheight;
    /* Count nondummy row groups.  Should get same answer for each component,
     * so we need only do it once.
     */
    if (ci == 0) {
      main->rowgroups_avail = (JDIMENSION) ((rows_left-1) / rgroup + 1);
    }
    /* Duplicate the last real sample row rgroup*2 times; this pads out the
     * last partial rowgroup and ensures at least one full rowgroup of context.
     */
    xbuf = main->xbuffer[main->whichptr][ci];
    for (i = 0; i < rgroup * 2; i++) {
      xbuf[rows_left + i] = xbuf[rows_left-1];
    }
  }
}


/*
 * Initialize for a processing pass.
 */

METHODDEF(void)
start_pass_main (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
{
  my_main_ptr main = (my_main_ptr) cinfo->main;

  switch (pass_mode) {
  case JBUF_PASS_THRU:
    if (cinfo->upsample->need_context_rows) {
      main->pub.process_data = process_data_context_main;
      make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
      main->whichptr = 0;       /* Read first iMCU row into xbuffer[0] */
      main->context_state = CTX_PREPARE_FOR_IMCU;
      main->iMCU_row_ctr = 0;
    } else {
      /* Simple case with no context needed */
      main->pub.process_data = process_data_simple_main;
    }
    main->buffer_full = FALSE;  /* Mark buffer empty */
    main->rowgroup_ctr = 0;
    break;
#ifdef QUANT_2PASS_SUPPORTED
  case JBUF_CRANK_DEST:
    /* For last pass of 2-pass quantization, just crank the postprocessor */
    main->pub.process_data = process_data_crank_post;
    break;
#endif
  default:
    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
    break;
  }
}


/*
 * Process some data.
 * This handles the simple case where no context is required.
 */

METHODDEF(void)
process_data_simple_main (j_decompress_ptr cinfo,
                          JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
                          JDIMENSION out_rows_avail)
{
  my_main_ptr main = (my_main_ptr) cinfo->main;
  JDIMENSION rowgroups_avail;

  /* Read input data if we haven't filled the main buffer yet */
  if (! main->buffer_full) {
    if (! (*cinfo->coef->decompress_data) (cinfo, main->buffer))
      return;                   /* suspension forced, can do nothing more */
    main->buffer_full = TRUE;   /* OK, we have an iMCU row to work with */
  }

  /* There are always min_DCT_scaled_size row groups in an iMCU row. */
  rowgroups_avail = (JDIMENSION) cinfo->min_DCT_v_scaled_size;
  /* Note: at the bottom of the image, we may pass extra garbage row groups
   * to the postprocessor.  The postprocessor has to check for bottom
   * of image anyway (at row resolution), so no point in us doing it too.
   */

  /* Feed the postprocessor */
  (*cinfo->post->post_process_data) (cinfo, main->buffer,
                                     &main->rowgroup_ctr, rowgroups_avail,
                                     output_buf, out_row_ctr, out_rows_avail);

  /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
  if (main->rowgroup_ctr >= rowgroups_avail) {
    main->buffer_full = FALSE;
    main->rowgroup_ctr = 0;
  }
}


/*
 * Process some data.
 * This handles the case where context rows must be provided.
 */

METHODDEF(void)
process_data_context_main (j_decompress_ptr cinfo,
                           JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
                           JDIMENSION out_rows_avail)
{
  my_main_ptr main = (my_main_ptr) cinfo->main;

  /* Read input data if we haven't filled the main buffer yet */
  if (! main->buffer_full) {
    if (! (*cinfo->coef->decompress_data) (cinfo,
                                           main->xbuffer[main->whichptr]))
      return;                   /* suspension forced, can do nothing more */
    main->buffer_full = TRUE;   /* OK, we have an iMCU row to work with */
    main->iMCU_row_ctr++;       /* count rows received */
  }

  /* Postprocessor typically will not swallow all the input data it is handed
   * in one call (due to filling the output buffer first).  Must be prepared
   * to exit and restart.  This switch lets us keep track of how far we got.
   * Note that each case falls through to the next on successful completion.
   */
  switch (main->context_state) {
  case CTX_POSTPONED_ROW:
    /* Call postprocessor using previously set pointers for postponed row */
    (*cinfo->post->post_process_data) (cinfo, main->xbuffer[main->whichptr],
                        &main->rowgroup_ctr, main->rowgroups_avail,
                        output_buf, out_row_ctr, out_rows_avail);
    if (main->rowgroup_ctr < main->rowgroups_avail)
      return;                   /* Need to suspend */
    main->context_state = CTX_PREPARE_FOR_IMCU;
    if (*out_row_ctr >= out_rows_avail)
      return;                   /* Postprocessor exactly filled output buf */
    /*FALLTHROUGH*/
  case CTX_PREPARE_FOR_IMCU:
    /* Prepare to process first M-1 row groups of this iMCU row */
    main->rowgroup_ctr = 0;
    main->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_v_scaled_size - 1);
    /* Check for bottom of image: if so, tweak pointers to "duplicate"
     * the last sample row, and adjust rowgroups_avail to ignore padding rows.
     */
    if (main->iMCU_row_ctr == cinfo->total_iMCU_rows)
      set_bottom_pointers(cinfo);
    main->context_state = CTX_PROCESS_IMCU;
    /*FALLTHROUGH*/
  case CTX_PROCESS_IMCU:
    /* Call postprocessor using previously set pointers */
    (*cinfo->post->post_process_data) (cinfo, main->xbuffer[main->whichptr],
                        &main->rowgroup_ctr, main->rowgroups_avail,
                        output_buf, out_row_ctr, out_rows_avail);
    if (main->rowgroup_ctr < main->rowgroups_avail)
      return;                   /* Need to suspend */
    /* After the first iMCU, change wraparound pointers to normal state */
    if (main->iMCU_row_ctr == 1)
      set_wraparound_pointers(cinfo);
    /* Prepare to load new iMCU row using other xbuffer list */
    main->whichptr ^= 1;        /* 0=>1 or 1=>0 */
    main->buffer_full = FALSE;
    /* Still need to process last row group of this iMCU row, */
    /* which is saved at index M+1 of the other xbuffer */
    main->rowgroup_ctr = (JDIMENSION) (cinfo->min_DCT_v_scaled_size + 1);
    main->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_v_scaled_size + 2);
    main->context_state = CTX_POSTPONED_ROW;
  }
}


/*
 * Process some data.
 * Final pass of two-pass quantization: just call the postprocessor.
 * Source data will be the postprocessor controller's internal buffer.
 */

#ifdef QUANT_2PASS_SUPPORTED

METHODDEF(void)
process_data_crank_post (j_decompress_ptr cinfo,
                         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
                         JDIMENSION out_rows_avail)
{
  (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
                                     (JDIMENSION *) NULL, (JDIMENSION) 0,
                                     output_buf, out_row_ctr, out_rows_avail);
}

#endif /* QUANT_2PASS_SUPPORTED */


/*
 * Initialize main buffer controller.
 */

GLOBAL(void)
jinit_d_main_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
{
  my_main_ptr main;
  int ci, rgroup, ngroups;
  jpeg_component_info *compptr;

  main = (my_main_ptr)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
                                SIZEOF(my_main_controller));
  cinfo->main = (struct jpeg_d_main_controller *) main;
  main->pub.start_pass = start_pass_main;

  if (need_full_buffer)         /* shouldn't happen */
    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);

  /* Allocate the workspace.
   * ngroups is the number of row groups we need.
   */
  if (cinfo->upsample->need_context_rows) {
    if (cinfo->min_DCT_v_scaled_size < 2) /* unsupported, see comments above */
      ERREXIT(cinfo, JERR_NOTIMPL);
    alloc_funny_pointers(cinfo); /* Alloc space for xbuffer[] lists */
    ngroups = cinfo->min_DCT_v_scaled_size + 2;
  } else {
    ngroups = cinfo->min_DCT_v_scaled_size;
  }

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
      cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
    main->buffer[ci] = (*cinfo->mem->alloc_sarray)
                        ((j_common_ptr) cinfo, JPOOL_IMAGE,
                         compptr->width_in_blocks * compptr->DCT_h_scaled_size,
                         (JDIMENSION) (rgroup * ngroups));
  }
}
1	/*
2	* jdmainct.c
3	*
4	* Copyright (C) 1994-1996, Thomas G. Lane.
5	* This file is part of the Independent JPEG Group's software.
6	* For conditions of distribution and use, see the accompanying README file.
7	*
8	* This file contains the main buffer controller for decompression.
9	* The main buffer lies between the JPEG decompressor proper and the
10	* post-processor; it holds downsampled data in the JPEG colorspace.
11	*
12	* Note that this code is bypassed in raw-data mode, since the application
13	* supplies the equivalent of the main buffer in that case.
14	*/
15
16	#define JPEG_INTERNALS
17	#include "jinclude.h"
18	#include "jpeglib.h"
19
20
21	/*
22	* In the current system design, the main buffer need never be a full-image
23	* buffer; any full-height buffers will be found inside the coefficient or
24	* postprocessing controllers. Nonetheless, the main controller is not
25	* trivial. Its responsibility is to provide context rows for upsampling/
26	* rescaling, and doing this in an efficient fashion is a bit tricky.
27	*
28	* Postprocessor input data is counted in "row groups". A row group
29	* is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
30	* sample rows of each component. (We require DCT_scaled_size values to be
31	* chosen such that these numbers are integers. In practice DCT_scaled_size
32	* values will likely be powers of two, so we actually have the stronger
33	* condition that DCT_scaled_size / min_DCT_scaled_size is an integer.)
34	* Upsampling will typically produce max_v_samp_factor pixel rows from each
35	* row group (times any additional scale factor that the upsampler is
36	* applying).
37	*
38	* The coefficient controller will deliver data to us one iMCU row at a time;
39	* each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or
40	* exactly min_DCT_scaled_size row groups. (This amount of data corresponds
41	* to one row of MCUs when the image is fully interleaved.) Note that the
42	* number of sample rows varies across components, but the number of row
43	* groups does not. Some garbage sample rows may be included in the last iMCU
44	* row at the bottom of the image.
45	*
46	* Depending on the vertical scaling algorithm used, the upsampler may need
47	* access to the sample row(s) above and below its current input row group.
48	* The upsampler is required to set need_context_rows TRUE at global selection
49	* time if so. When need_context_rows is FALSE, this controller can simply
50	* obtain one iMCU row at a time from the coefficient controller and dole it
51	* out as row groups to the postprocessor.
52	*
53	* When need_context_rows is TRUE, this controller guarantees that the buffer
54	* passed to postprocessing contains at least one row group's worth of samples
55	* above and below the row group(s) being processed. Note that the context
56	* rows "above" the first passed row group appear at negative row offsets in
57	* the passed buffer. At the top and bottom of the image, the required
58	* context rows are manufactured by duplicating the first or last real sample
59	* row; this avoids having special cases in the upsampling inner loops.
60	*
61	* The amount of context is fixed at one row group just because that's a
62	* convenient number for this controller to work with. The existing
63	* upsamplers really only need one sample row of context. An upsampler
64	* supporting arbitrary output rescaling might wish for more than one row
65	* group of context when shrinking the image; tough, we don't handle that.
66	* (This is justified by the assumption that downsizing will be handled mostly
67	* by adjusting the DCT_scaled_size values, so that the actual scale factor at
68	* the upsample step needn't be much less than one.)
69	*
70	* To provide the desired context, we have to retain the last two row groups
71	* of one iMCU row while reading in the next iMCU row. (The last row group
72	* can't be processed until we have another row group for its below-context,
73	* and so we have to save the next-to-last group too for its above-context.)
74	* We could do this most simply by copying data around in our buffer, but
75	* that'd be very slow. We can avoid copying any data by creating a rather
76	* strange pointer structure. Here's how it works. We allocate a workspace
77	* consisting of M+2 row groups (where M = min_DCT_scaled_size is the number
78	* of row groups per iMCU row). We create two sets of redundant pointers to
79	* the workspace. Labeling the physical row groups 0 to M+1, the synthesized
80	* pointer lists look like this:
81	* M+1 M-1
82	* master pointer --> 0 master pointer --> 0
83	* 1 1
84	* ... ...
85	* M-3 M-3
86	* M-2 M
87	* M-1 M+1
88	* M M-2
89	* M+1 M-1
90	* 0 0
91	* We read alternate iMCU rows using each master pointer; thus the last two
92	* row groups of the previous iMCU row remain un-overwritten in the workspace.
93	* The pointer lists are set up so that the required context rows appear to
94	* be adjacent to the proper places when we pass the pointer lists to the
95	* upsampler.
96	*
97	* The above pictures describe the normal state of the pointer lists.
98	* At top and bottom of the image, we diddle the pointer lists to duplicate
99	* the first or last sample row as necessary (this is cheaper than copying
100	* sample rows around).
101	*
102	* This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1. In that
103	* situation each iMCU row provides only one row group so the buffering logic
104	* must be different (eg, we must read two iMCU rows before we can emit the
105	* first row group). For now, we simply do not support providing context
106	* rows when min_DCT_scaled_size is 1. That combination seems unlikely to
107	* be worth providing --- if someone wants a 1/8th-size preview, they probably
108	* want it quick and dirty, so a context-free upsampler is sufficient.
109	*/
110
111
112	/* Private buffer controller object */
113
114	typedef struct {
115	struct jpeg_d_main_controller pub; /* public fields */
116
117	/* Pointer to allocated workspace (M or M+2 row groups). */
118	JSAMPARRAY buffer[MAX_COMPONENTS];
119
120	boolean buffer_full; /* Have we gotten an iMCU row from decoder? */
121	JDIMENSION rowgroup_ctr; /* counts row groups output to postprocessor */
122
123	/* Remaining fields are only used in the context case. */
124
125	/* These are the master pointers to the funny-order pointer lists. */
126	JSAMPIMAGE xbuffer[2]; /* pointers to weird pointer lists */
127
128	int whichptr; /* indicates which pointer set is now in use */
129	int context_state; /* process_data state machine status */
130	JDIMENSION rowgroups_avail; /* row groups available to postprocessor */
131	JDIMENSION iMCU_row_ctr; /* counts iMCU rows to detect image top/bot */
132	} my_main_controller;
133
134	typedef my_main_controller * my_main_ptr;
135
136	/* context_state values: */
137	#define CTX_PREPARE_FOR_IMCU 0 /* need to prepare for MCU row */
138	#define CTX_PROCESS_IMCU 1 /* feeding iMCU to postprocessor */
139	#define CTX_POSTPONED_ROW 2 /* feeding postponed row group */
140
141
142	/* Forward declarations */
143	METHODDEF(void) process_data_simple_main
144	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
145	JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
146	METHODDEF(void) process_data_context_main
147	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
148	JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
149	#ifdef QUANT_2PASS_SUPPORTED
150	METHODDEF(void) process_data_crank_post
151	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
152	JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
153	#endif
154
155
156	LOCAL(void)
157	alloc_funny_pointers (j_decompress_ptr cinfo)
158	/* Allocate space for the funny pointer lists.
159	* This is done only once, not once per pass.
160	*/
161	{
162	my_main_ptr main = (my_main_ptr) cinfo->main;
163	int ci, rgroup;
164	int M = cinfo->min_DCT_v_scaled_size;
165	jpeg_component_info *compptr;
166	JSAMPARRAY xbuf;
167
168	/* Get top-level space for component array pointers.
169	* We alloc both arrays with one call to save a few cycles.
170	*/
171	main->xbuffer[0] = (JSAMPIMAGE)
172	(*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
173	cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
174	main->xbuffer[1] = main->xbuffer[0] + cinfo->num_components;
175
176	for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
177	ci++, compptr++) {
178	rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
179	cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
180	/* Get space for pointer lists --- M+4 row groups in each list.
181	* We alloc both pointer lists with one call to save a few cycles.
182	*/
183	xbuf = (JSAMPARRAY)
184	(*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
185	2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
186	xbuf += rgroup; /* want one row group at negative offsets */
187	main->xbuffer[0][ci] = xbuf;
188	xbuf += rgroup * (M + 4);
189	main->xbuffer[1][ci] = xbuf;
190	}
191	}
192
193
194	LOCAL(void)
195	make_funny_pointers (j_decompress_ptr cinfo)
196	/* Create the funny pointer lists discussed in the comments above.
197	* The actual workspace is already allocated (in main->buffer),
198	* and the space for the pointer lists is allocated too.
199	* This routine just fills in the curiously ordered lists.
200	* This will be repeated at the beginning of each pass.
201	*/
202	{
203	my_main_ptr main = (my_main_ptr) cinfo->main;
204	int ci, i, rgroup;
205	int M = cinfo->min_DCT_v_scaled_size;
206	jpeg_component_info *compptr;
207	JSAMPARRAY buf, xbuf0, xbuf1;
208
209	for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
210	ci++, compptr++) {
211	rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
212	cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
213	xbuf0 = main->xbuffer[0][ci];
214	xbuf1 = main->xbuffer[1][ci];
215	/* First copy the workspace pointers as-is */
216	buf = main->buffer[ci];
217	for (i = 0; i < rgroup * (M + 2); i++) {
218	xbuf0[i] = xbuf1[i] = buf[i];
219	}
220	/* In the second list, put the last four row groups in swapped order */
221	for (i = 0; i < rgroup * 2; i++) {
222	xbuf1[rgroup(M-2) + i] = buf[rgroupM + i];
223	xbuf1[rgroupM + i] = buf[rgroup(M-2) + i];
224	}
225	/* The wraparound pointers at top and bottom will be filled later
226	* (see set_wraparound_pointers, below). Initially we want the "above"
227	* pointers to duplicate the first actual data line. This only needs
228	* to happen in xbuffer[0].
229	*/
230	for (i = 0; i < rgroup; i++) {
231	xbuf0[i - rgroup] = xbuf0[0];
232	}
233	}
234	}
235
236
237	LOCAL(void)
238	set_wraparound_pointers (j_decompress_ptr cinfo)
239	/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
240	* This changes the pointer list state from top-of-image to the normal state.
241	*/
242	{
243	my_main_ptr main = (my_main_ptr) cinfo->main;
244	int ci, i, rgroup;
245	int M = cinfo->min_DCT_v_scaled_size;
246	jpeg_component_info *compptr;
247	JSAMPARRAY xbuf0, xbuf1;
248
249	for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
250	ci++, compptr++) {
251	rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
252	cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
253	xbuf0 = main->xbuffer[0][ci];
254	xbuf1 = main->xbuffer[1][ci];
255	for (i = 0; i < rgroup; i++) {
256	xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
257	xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
258	xbuf0[rgroup*(M+2) + i] = xbuf0[i];
259	xbuf1[rgroup*(M+2) + i] = xbuf1[i];
260	}
261	}
262	}
263
264
265	LOCAL(void)
266	set_bottom_pointers (j_decompress_ptr cinfo)
267	/* Change the pointer lists to duplicate the last sample row at the bottom
268	* of the image. whichptr indicates which xbuffer holds the final iMCU row.
269	* Also sets rowgroups_avail to indicate number of nondummy row groups in row.
270	*/
271	{
272	my_main_ptr main = (my_main_ptr) cinfo->main;
273	int ci, i, rgroup, iMCUheight, rows_left;
274	jpeg_component_info *compptr;
275	JSAMPARRAY xbuf;
276
277	for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
278	ci++, compptr++) {
279	/* Count sample rows in one iMCU row and in one row group */
280	iMCUheight = compptr->v_samp_factor * compptr->DCT_v_scaled_size;
281	rgroup = iMCUheight / cinfo->min_DCT_v_scaled_size;
282	/* Count nondummy sample rows remaining for this component */
283	rows_left = (int) (compptr->downsampled_height % (JDIMENSION) iMCUheight);
284	if (rows_left == 0) rows_left = iMCUheight;
285	/* Count nondummy row groups. Should get same answer for each component,
286	* so we need only do it once.
287	*/
288	if (ci == 0) {
289	main->rowgroups_avail = (JDIMENSION) ((rows_left-1) / rgroup + 1);
290	}
291	/* Duplicate the last real sample row rgroup*2 times; this pads out the
292	* last partial rowgroup and ensures at least one full rowgroup of context.
293	*/
294	xbuf = main->xbuffer[main->whichptr][ci];
295	for (i = 0; i < rgroup * 2; i++) {
296	xbuf[rows_left + i] = xbuf[rows_left-1];
297	}
298	}
299	}
300
301
302	/*
303	* Initialize for a processing pass.
304	*/
305
306	METHODDEF(void)
307	start_pass_main (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
308	{
309	my_main_ptr main = (my_main_ptr) cinfo->main;
310
311	switch (pass_mode) {
312	case JBUF_PASS_THRU:
313	if (cinfo->upsample->need_context_rows) {
314	main->pub.process_data = process_data_context_main;
315	make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
316	main->whichptr = 0; /* Read first iMCU row into xbuffer[0] */
317	main->context_state = CTX_PREPARE_FOR_IMCU;
318	main->iMCU_row_ctr = 0;
319	} else {
320	/* Simple case with no context needed */
321	main->pub.process_data = process_data_simple_main;
322	}
323	main->buffer_full = FALSE; /* Mark buffer empty */
324	main->rowgroup_ctr = 0;
325	break;
326	#ifdef QUANT_2PASS_SUPPORTED
327	case JBUF_CRANK_DEST:
328	/* For last pass of 2-pass quantization, just crank the postprocessor */
329	main->pub.process_data = process_data_crank_post;
330	break;
331	#endif
332	default:
333	ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
334	break;
335	}
336	}
337
338
339	/*
340	* Process some data.
341	* This handles the simple case where no context is required.
342	*/
343
344	METHODDEF(void)
345	process_data_simple_main (j_decompress_ptr cinfo,
346	JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
347	JDIMENSION out_rows_avail)
348	{
349	my_main_ptr main = (my_main_ptr) cinfo->main;
350	JDIMENSION rowgroups_avail;
351
352	/* Read input data if we haven't filled the main buffer yet */
353	if (! main->buffer_full) {
354	if (! (*cinfo->coef->decompress_data) (cinfo, main->buffer))
355	return; /* suspension forced, can do nothing more */
356	main->buffer_full = TRUE; /* OK, we have an iMCU row to work with */
357	}
358
359	/* There are always min_DCT_scaled_size row groups in an iMCU row. */
360	rowgroups_avail = (JDIMENSION) cinfo->min_DCT_v_scaled_size;
361	/* Note: at the bottom of the image, we may pass extra garbage row groups
362	* to the postprocessor. The postprocessor has to check for bottom
363	* of image anyway (at row resolution), so no point in us doing it too.
364	*/
365
366	/* Feed the postprocessor */
367	(*cinfo->post->post_process_data) (cinfo, main->buffer,
368	&main->rowgroup_ctr, rowgroups_avail,
369	output_buf, out_row_ctr, out_rows_avail);
370
371	/* Has postprocessor consumed all the data yet? If so, mark buffer empty */
372	if (main->rowgroup_ctr >= rowgroups_avail) {
373	main->buffer_full = FALSE;
374	main->rowgroup_ctr = 0;
375	}
376	}
377
378
379	/*
380	* Process some data.
381	* This handles the case where context rows must be provided.
382	*/
383
384	METHODDEF(void)
385	process_data_context_main (j_decompress_ptr cinfo,
386	JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
387	JDIMENSION out_rows_avail)
388	{
389	my_main_ptr main = (my_main_ptr) cinfo->main;
390
391	/* Read input data if we haven't filled the main buffer yet */
392	if (! main->buffer_full) {
393	if (! (*cinfo->coef->decompress_data) (cinfo,
394	main->xbuffer[main->whichptr]))
395	return; /* suspension forced, can do nothing more */
396	main->buffer_full = TRUE; /* OK, we have an iMCU row to work with */
397	main->iMCU_row_ctr++; /* count rows received */
398	}
399
400	/* Postprocessor typically will not swallow all the input data it is handed
401	* in one call (due to filling the output buffer first). Must be prepared
402	* to exit and restart. This switch lets us keep track of how far we got.
403	* Note that each case falls through to the next on successful completion.
404	*/
405	switch (main->context_state) {
406	case CTX_POSTPONED_ROW:
407	/* Call postprocessor using previously set pointers for postponed row */
408	(*cinfo->post->post_process_data) (cinfo, main->xbuffer[main->whichptr],
409	&main->rowgroup_ctr, main->rowgroups_avail,
410	output_buf, out_row_ctr, out_rows_avail);
411	if (main->rowgroup_ctr < main->rowgroups_avail)
412	return; /* Need to suspend */
413	main->context_state = CTX_PREPARE_FOR_IMCU;
414	if (*out_row_ctr >= out_rows_avail)
415	return; /* Postprocessor exactly filled output buf */
416	/FALLTHROUGH/
417	case CTX_PREPARE_FOR_IMCU:
418	/* Prepare to process first M-1 row groups of this iMCU row */
419	main->rowgroup_ctr = 0;
420	main->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_v_scaled_size - 1);
421	/* Check for bottom of image: if so, tweak pointers to "duplicate"
422	* the last sample row, and adjust rowgroups_avail to ignore padding rows.
423	*/
424	if (main->iMCU_row_ctr == cinfo->total_iMCU_rows)
425	set_bottom_pointers(cinfo);
426	main->context_state = CTX_PROCESS_IMCU;
427	/FALLTHROUGH/
428	case CTX_PROCESS_IMCU:
429	/* Call postprocessor using previously set pointers */
430	(*cinfo->post->post_process_data) (cinfo, main->xbuffer[main->whichptr],
431	&main->rowgroup_ctr, main->rowgroups_avail,
432	output_buf, out_row_ctr, out_rows_avail);
433	if (main->rowgroup_ctr < main->rowgroups_avail)
434	return; /* Need to suspend */
435	/* After the first iMCU, change wraparound pointers to normal state */
436	if (main->iMCU_row_ctr == 1)
437	set_wraparound_pointers(cinfo);
438	/* Prepare to load new iMCU row using other xbuffer list */
439	main->whichptr ^= 1; /* 0=>1 or 1=>0 */
440	main->buffer_full = FALSE;
441	/* Still need to process last row group of this iMCU row, */
442	/* which is saved at index M+1 of the other xbuffer */
443	main->rowgroup_ctr = (JDIMENSION) (cinfo->min_DCT_v_scaled_size + 1);
444	main->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_v_scaled_size + 2);
445	main->context_state = CTX_POSTPONED_ROW;
446	}
447	}
448
449
450	/*
451	* Process some data.
452	* Final pass of two-pass quantization: just call the postprocessor.
453	* Source data will be the postprocessor controller's internal buffer.
454	*/
455
456	#ifdef QUANT_2PASS_SUPPORTED
457
458	METHODDEF(void)
459	process_data_crank_post (j_decompress_ptr cinfo,
460	JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
461	JDIMENSION out_rows_avail)
462	{
463	(*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
464	(JDIMENSION *) NULL, (JDIMENSION) 0,
465	output_buf, out_row_ctr, out_rows_avail);
466	}
467
468	#endif /* QUANT_2PASS_SUPPORTED */
469
470
471	/*
472	* Initialize main buffer controller.
473	*/
474
475	GLOBAL(void)
476	jinit_d_main_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
477	{
478	my_main_ptr main;
479	int ci, rgroup, ngroups;
480	jpeg_component_info *compptr;
481
482	main = (my_main_ptr)
483	(*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
484	SIZEOF(my_main_controller));
485	cinfo->main = (struct jpeg_d_main_controller *) main;
486	main->pub.start_pass = start_pass_main;
487
488	if (need_full_buffer) /* shouldn't happen */
489	ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
490
491	/* Allocate the workspace.
492	* ngroups is the number of row groups we need.
493	*/
494	if (cinfo->upsample->need_context_rows) {
495	if (cinfo->min_DCT_v_scaled_size < 2) /* unsupported, see comments above */
496	ERREXIT(cinfo, JERR_NOTIMPL);
497	alloc_funny_pointers(cinfo); /* Alloc space for xbuffer[] lists */
498	ngroups = cinfo->min_DCT_v_scaled_size + 2;
499	} else {
500	ngroups = cinfo->min_DCT_v_scaled_size;
501	}
502
503	for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
504	ci++, compptr++) {
505	rgroup = (compptr->v_samp_factor * compptr->DCT_v_scaled_size) /
506	cinfo->min_DCT_v_scaled_size; /* height of a row group of component */
507	main->buffer[ci] = (*cinfo->mem->alloc_sarray)
508	((j_common_ptr) cinfo, JPOOL_IMAGE,
509	compptr->width_in_blocks * compptr->DCT_h_scaled_size,
510	(JDIMENSION) (rgroup * ngroups));
511	}
512	}