JSAMPARRAY output_buf, int num_rows)
/* Fast path for out_color_components==3, with ordered dithering */
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  register int pixcode;
  register JSAMPROW input_ptr;
  register JSAMPROW output_ptr;
  JSAMPROW colorindex0 = cquantize->colorindex[0];
  JSAMPROW colorindex1 = cquantize->colorindex[1];
  JSAMPROW colorindex2 = cquantize->colorindex[2];
  int * dither0;		/* points to active row of dither matrix */
  int * dither1;
  int * dither2;
  int row_index, col_index;	/* current indexes into dither matrix */
  int row;
  JDIMENSION col;
  JDIMENSION width = cinfo->output_width;

  for (row = 0; row < num_rows; row++) {
    row_index = cquantize->row_index;
    input_ptr = input_buf[row];
    output_ptr = output_buf[row];
    dither0 = cquantize->odither[0][row_index];
    dither1 = cquantize->odither[1][row_index];
    dither2 = cquantize->odither[2][row_index];
    col_index = 0;

    for (col = width; col > 0; col--) {
      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) +
					dither0[col_index]]);
      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) +
					dither1[col_index]]);
      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) +
					dither2[col_index]]);
      *output_ptr++ = (JSAMPLE) pixcode;
      col_index = (col_index + 1) & ODITHER_MASK;
    }
    row_index = (row_index + 1) & ODITHER_MASK;
    cquantize->row_index = row_index;
  }
}


METHODDEF(void)
quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
		    JSAMPARRAY output_buf, int num_rows)
/* General case, with Floyd-Steinberg dithering */
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  register LOCFSERROR cur;	/* current error or pixel value */
  LOCFSERROR belowerr;		/* error for pixel below cur */
  LOCFSERROR bpreverr;		/* error for below/prev col */
  LOCFSERROR bnexterr;		/* error for below/next col */
  LOCFSERROR delta;
  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
  register JSAMPROW input_ptr;
  register JSAMPROW output_ptr;
  JSAMPROW colorindex_ci;
  JSAMPROW colormap_ci;
  int pixcode;
  int nc = cinfo->out_color_components;
  int dir;			/* 1 for left-to-right, -1 for right-to-left */
  int dirnc;			/* dir * nc */
  int ci;
  int row;
  JDIMENSION col;
  JDIMENSION width = cinfo->output_width;
  JSAMPLE *range_limit = cinfo->sample_range_limit;
  SHIFT_TEMPS

  for (row = 0; row < num_rows; row++) {
    /* Initialize output values to 0 so can process components separately */
    jzero_far((void FAR *) output_buf[row],
	      (size_t) (width * SIZEOF(JSAMPLE)));
    for (ci = 0; ci < nc; ci++) {
      input_ptr = input_buf[row] + ci;
      output_ptr = output_buf[row];
      if (cquantize->on_odd_row) {
	/* work right to left in this row */
	input_ptr += (width-1) * nc; /* so point to rightmost pixel */
	output_ptr += width-1;
	dir = -1;
	dirnc = -nc;
	errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
      } else {
	/* work left to right in this row */
	dir = 1;
	dirnc = nc;
	errorptr = cquantize->fserrors[ci]; /* => entry before first column */
      }
      colorindex_ci = cquantize->colorindex[ci];
      colormap_ci = cquantize->sv_colormap[ci];
      /* Preset error values: no error propagated to first pixel from left */
      cur = 0;
      /* and no error propagated to row below yet */
      belowerr = bpreverr = 0;

      for (col = width; col > 0; col--) {
	/* cur holds the error propagated from the previous pixel on the
	 * current line.  Add the error propagated from the previous line
	 * to form the complete error correction term for this pixel, and
	 * round the error term (which is expressed * 16) to an integer.
	 * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
	 * for either sign of the error value.
	 * Note: errorptr points to *previous* column's array entry.
	 */
	cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
	/* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
	 * The maximum error is +- MAXJSAMPLE; this sets the required size
	 * of the range_limit array.
	 */
	cur += GETJSAMPLE(*input_ptr);
	cur = GETJSAMPLE(range_limit[cur]);
	/* Select output value, accumulate into output code for this pixel */
	pixcode = GETJSAMPLE(colorindex_ci[cur]);
	*output_ptr += (JSAMPLE) pixcode;
	/* Compute actual representation error at this pixel */
	/* Note: we can do this even though we don't have the final */
	/* pixel code, because the colormap is orthogonal. */
	cur -= GETJSAMPLE(colormap_ci[pixcode]);
	/* Compute error fractions to be propagated to adjacent pixels.
	 * Add these into the running sums, and simultaneously shift the
	 * next-line error sums left by 1 column.
	 */
	bnexterr = cur;
	delta = cur * 2;
	cur += delta;		/* form error * 3 */
	errorptr[0] = (FSERROR) (bpreverr + cur);
	cur += delta;		/* form error * 5 */
	bpreverr = belowerr + cur;
	belowerr = bnexterr;
	cur += delta;		/* form error * 7 */
	/* At this point cur contains the 7/16 error value to be propagated
	 * to the next pixel on the current line, and all the errors for the
	 * next line have been shifted over. We are therefore ready to move on.
	 */
	input_ptr += dirnc;	/* advance input ptr to next column */
	output_ptr += dir;	/* advance output ptr to next column */
	errorptr += dir;	/* advance errorptr to current column */
      }
      /* Post-loop cleanup: we must unload the final error value into the
       * final fserrors[] entry.  Note we need not unload belowerr because
       * it is for the dummy column before or after the actual array.
       */
      errorptr[0] = (FSERROR) bpreverr; /* unload prev err into array */
    }
    cquantize->on_odd_row = (cquantize->on_odd_row ? FALSE : TRUE);
  }
}


/*
 * Allocate workspace for Floyd-Steinberg errors.
 */

LOCAL(void)
alloc_fs_workspace (j_decompress_ptr cinfo)
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  size_t arraysize;
  int i;

  arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
  for (i = 0; i < cinfo->out_color_components; i++) {
    cquantize->fserrors[i] = (FSERRPTR)
      (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
  }
}


/*
 * Initialize for one-pass color quantization.
 */

METHODDEF(void)
start_pass_1_quant (j_decompress_ptr cinfo, boolean is_pre_scan)
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  size_t arraysize;
  int i;

  /* Install my colormap. */
  cinfo->colormap = cquantize->sv_colormap;
  cinfo->actual_number_of_colors = cquantize->sv_actual;

  /* Initialize for desired dithering mode. */
  switch (cinfo->dither_mode) {
  case JDITHER_NONE:
    if (cinfo->out_color_components == 3)
      cquantize->pub.color_quantize = color_quantize3;
    else
      cquantize->pub.color_quantize = color_quantize;
    break;
  case JDITHER_ORDERED:
    if (cinfo->out_color_components == 3)
      cquantize->pub.color_quantize = quantize3_ord_dither;
    else
      cquantize->pub.color_quantize = quantize_ord_dither;
    cquantize->row_index = 0;	/* initialize state for ordered dither */
    /* If user changed to ordered dither from another mode,
     * we must recreate the color index table with padding.
     * This will cost extra space, but probably isn't very likely.
     */
    if (! cquantize->is_padded)
      create_colorindex(cinfo);
    /* Create ordered-dither tables if we didn't already. */
    if (cquantize->odither[0] == NULL)
      create_odither_tables(cinfo);
    break;
  case JDITHER_FS:
    cquantize->pub.color_quantize = quantize_fs_dither;
    cquantize->on_odd_row = FALSE; /* initialize state for F-S dither */
    /* Allocate Floyd-Steinberg workspace if didn't already. */
    if (cquantize->fserrors[0] == NULL)
      alloc_fs_workspace(cinfo);
    /* Initialize the propagated errors to zero. */
    arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
    for (i = 0; i < cinfo->out_color_components; i++)
      jzero_far((void FAR *) cquantize->fserrors[i], arraysize);
    break;
  default:
    ERREXIT(cinfo, JERR_NOT_COMPILED);
    break;
  }
}


/*
 * Finish up at the end of the pass.
 */

METHODDEF(void)
finish_pass_1_quant (j_decompress_ptr cinfo)
{
  /* no work in 1-pass case */
}


/*
 * Switch to a new external colormap between output passes.
 * Shouldn't get to this module!
 */

METHODDEF(void)
new_color_map_1_quant (j_decompress_ptr cinfo)
{
  ERREXIT(cinfo, JERR_MODE_CHANGE);
}


/*
 * Module initialization routine for 1-pass color quantization.
 */

GLOBAL(void)
jinit_1pass_quantizer (j_decompress_ptr cinfo)
{
  my_cquantize_ptr cquantize;

  cquantize = (my_cquantize_ptr)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				SIZEOF(my_cquantizer));
  cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
  cquantize->pub.start_pass = start_pass_1_quant;
  cquantize->pub.finish_pass = finish_pass_1_quant;
  cquantize->pub.new_color_map = new_color_map_1_quant;
  cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */
  cquantize->odither[0] = NULL;	/* Also flag odither arrays not allocated */

  /* Make sure my internal arrays won't overflow */
  if (cinfo->out_color_components > MAX_Q_COMPS)
    ERREXIT1(cinfo, JERR_QUANT_COMPONENTS, MAX_Q_COMPS);
  /* Make sure colormap indexes can be represented by JSAMPLEs */
  if (cinfo->desired_number_of_colors > (MAXJSAMPLE+1))
    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXJSAMPLE+1);

  /* Create the colormap and color index table. */
  create_colormap(cinfo);
  create_colorindex(cinfo);

  /* Allocate Floyd-Steinberg workspace now if requested.
   * We do this now since it is FAR storage and may affect the memory
   * manager's space calculations.  If the user changes to FS dither
   * mode in a later pass, we will allocate the space then, and will
   * possibly overrun the max_memory_to_use setting.
   */
  if (cinfo->dither_mode == JDITHER_FS)
    alloc_fs_workspace(cinfo);
}

#endif /* QUANT_1PASS_SUPPORTED */