jidctred.c

M8手机图片查看器

  /* Add fudge factor here for final descale. */
  tmp4 += ONE << 2;

  tmp0 = tmp4 + tmp5;
  tmp2 = tmp4 - tmp5;

  /* Column 1 */
  tmp4 = DEQUANTIZE(coef_block[DCTSIZE*0+1], quantptr[DCTSIZE*0+1]);
  tmp5 = DEQUANTIZE(coef_block[DCTSIZE*1+1], quantptr[DCTSIZE*1+1]);

  tmp1 = tmp4 + tmp5;
  tmp3 = tmp4 - tmp5;

  /* Pass 2: process 2 rows, store into output array. */

  /* Row 0 */
  outptr = output_buf[0] + output_col;

  outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp0 + tmp1, 3) & RANGE_MASK];
  outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp0 - tmp1, 3) & RANGE_MASK];

  /* Row 1 */
  outptr = output_buf[1] + output_col;

  outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp2 + tmp3, 3) & RANGE_MASK];
  outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp2 - tmp3, 3) & RANGE_MASK];
}


/*
 * Perform dequantization and inverse DCT on one block of coefficients,
 * producing a reduced-size 1x1 output block.
 *
 * We hardly need an inverse DCT routine for this: just take the
 * average pixel value, which is one-eighth of the DC coefficient.
 */

GLOBAL(void)
jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	       JCOEFPTR coef_block,
	       JSAMPARRAY output_buf, JDIMENSION output_col)
{
  int dcval;
  ISLOW_MULT_TYPE * quantptr;
  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
  SHIFT_TEMPS

  /* 1x1 is trivial: just take the DC coefficient divided by 8. */
  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
  dcval = DEQUANTIZE(coef_block[0], quantptr[0]);
  dcval = (int) DESCALE((INT32) dcval, 3);

  output_buf[0][output_col] = range_limit[dcval & RANGE_MASK];
}


/*
 * Perform dequantization and inverse DCT on one block of coefficients,
 * producing a 9x9 output block.
 *
 * Optimized algorithm with 10 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/18).
 */

GLOBAL(void)
jpeg_idct_9x9 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
	       JCOEFPTR coef_block,
	       JSAMPARRAY output_buf, JDIMENSION output_col)
{
  INT32 tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13, tmp14;
  INT32 z1, z2, z3, z4;
  JCOEFPTR inptr;
  ISLOW_MULT_TYPE * quantptr;
  int * wsptr;
  JSAMPROW outptr;
  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
  int ctr;
  int workspace[8*9];	/* buffers data between passes */
  SHIFT_TEMPS

  /* Pass 1: process columns from input, store into work array. */

  inptr = coef_block;
  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
  wsptr = workspace;
  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
    /* Even part */

    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
    tmp0 <<= CONST_BITS;
    /* Add fudge factor here for final descale. */
    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);

    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);

    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
    tmp1 = tmp0 + tmp3;
    tmp2 = tmp0 - tmp3 - tmp3;

    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
    tmp11 = tmp2 + tmp0;
    tmp14 = tmp2 - tmp0 - tmp0;

    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */

    tmp10 = tmp1 + tmp0 - tmp3;
    tmp12 = tmp1 - tmp0 + tmp2;
    tmp13 = tmp1 - tmp2 + tmp3;

    /* Odd part */

    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);

    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */

    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
    tmp0 = tmp2 + tmp3 - z2;
    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
    tmp2 += z2 - tmp1;
    tmp3 += z2 + tmp1;
    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */

    /* Final output stage */

    wsptr[8*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
    wsptr[8*8] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
    wsptr[8*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
    wsptr[8*7] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
    wsptr[8*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
    wsptr[8*6] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
    wsptr[8*3] = (int) RIGHT_SHIFT(tmp13 + tmp3, CONST_BITS-PASS1_BITS);
    wsptr[8*5] = (int) RIGHT_SHIFT(tmp13 - tmp3, CONST_BITS-PASS1_BITS);
    wsptr[8*4] = (int) RIGHT_SHIFT(tmp14, CONST_BITS-PASS1_BITS);
  }

  /* Pass 2: process 9 rows from work array, store into output array. */

  wsptr = workspace;
  for (ctr = 0; ctr < 9; ctr++) {
    outptr = output_buf[ctr] + output_col;

    /* Even part */

    /* Add fudge factor here for final descale. */
    tmp0 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2));
    tmp0 <<= CONST_BITS;

    z1 = (INT32) wsptr[2];
    z2 = (INT32) wsptr[4];
    z3 = (INT32) wsptr[6];

    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
    tmp1 = tmp0 + tmp3;
    tmp2 = tmp0 - tmp3 - tmp3;

    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
    tmp11 = tmp2 + tmp0;
    tmp14 = tmp2 - tmp0 - tmp0;

    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */

    tmp10 = tmp1 + tmp0 - tmp3;
    tmp12 = tmp1 - tmp0 + tmp2;
    tmp13 = tmp1 - tmp2 + tmp3;

    /* Odd part */

    z1 = (INT32) wsptr[1];
    z2 = (INT32) wsptr[3];
    z3 = (INT32) wsptr[5];
    z4 = (INT32) wsptr[7];

    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */

    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
    tmp0 = tmp2 + tmp3 - z2;
    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
    tmp2 += z2 - tmp1;
    tmp3 += z2 + tmp1;
    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */

    /* Final output stage */

    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13 + tmp3,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp13 - tmp3,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];
    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp14,
					      CONST_BITS+PASS1_BITS+3)
			    & RANGE_MASK];

    wsptr += 8;		/* advance pointer to next row */
  }
}


/*
 * Perform dequantization and inverse DCT on one block of coefficients,
 * producing a 10x10 output block.
 *
 * Optimized algorithm with 12 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/20).
 */

GLOBAL(void)
jpeg_idct_10x10 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
		 JCOEFPTR coef_block,
		 JSAMPARRAY output_buf, JDIMENSION output_col)
{
  INT32 tmp10, tmp11, tmp12, tmp13, tmp14;
  INT32 tmp20, tmp21, tmp22, tmp23, tmp24;
  INT32 z1, z2, z3, z4, z5;
  JCOEFPTR inptr;
  ISLOW_MULT_TYPE * quantptr;
  int * wsptr;
  JSAMPROW outptr;
  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
  int ctr;
  int workspace[8*10];	/* buffers data between passes */
  SHIFT_TEMPS

  /* Pass 1: process columns from input, store into work array. */

  inptr = coef_block;
  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
  wsptr = workspace;
  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
    /* Even part */

    z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
    z3 <<= CONST_BITS;
    /* Add fudge factor here for final descale. */
    z3 += ONE << (CONST_BITS-PASS1_BITS-1);
    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
    tmp10 = z3 + z1;
    tmp11 = z3 - z2;

    tmp22 = RIGHT_SHIFT(z3 - ((z1 - z2) << 1),   /* c0 = (c4-c8)*2 */
			CONST_BITS-PASS1_BITS);

    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);

    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */

    tmp20 = tmp10 + tmp12;
    tmp24 = tmp10 - tmp12;
    tmp21 = tmp11 + tmp13;
    tmp23 = tmp11 - tmp13;

    /* Odd part */

    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);

    tmp11 = z2 + z4;
    tmp13 = z2 - z4;

    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
    z5 = z3 << CONST_BITS;

    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
    z4 = z5 + tmp12;

    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */

    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
    z4 = z5 - tmp12 - (tmp13 << (CONST_BITS - 1));

    tmp12 = (z1 - tmp13 - z3) << PASS1_BITS;

    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */

    /* Final output stage */

    wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
    wsptr[8*9] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
    wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
    wsptr[8*8] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
    wsptr[8*2] = (int) (tmp22 + tmp12);
    wsptr[8*7] = (int) (tmp22 - tmp12);
    wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
    wsptr[8*6] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
    wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
    wsptr[8*5] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
  }

  /* Pass 2: process 10 rows from work array, store into output array. */

  wsptr = workspace;
  for (ctr = 0; ctr < 10; ctr++) {
    outptr = output_buf[ctr] + output_col;

    /* Even part */

    /* Add fudge factor here for final descale. */
    z3 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2));
    z3 <<= CONST_BITS;
    z4 = (INT32) wsptr[4];
    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
    tmp10 = z3 + z1;
    tmp11 = z3 - z2;

    tmp22 = z3 - ((z1 - z2) << 1);               /* c0 = (c4-c8)*2 */

    z2 = (INT32) wsptr[2];
    z3 = (INT32) wsptr[6];

    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */

    tmp20 = tmp10 + tmp12;
    tmp24 = tmp10 - tmp12;
    tmp21 = tmp11 + tmp13;
    tmp23 = tmp11 - tmp13;

    /* Odd part */

    z1 = (INT32) wsptr[1];
    z2 = (INT32) wsptr[3];
    z3 = (INT32) wsptr[5];
    z3 <<= CONST_BITS;
    z4 = (INT32) wsptr[7];