annotate jfdctint.c @ 3990:746a60ba3177 libavcodec

enable CMOV_IS_FAST as its faster or equal speed on every cpu (duron, athlon, PM, P3) from which ive seen benchmarks, it might be slower on P4 but noone has posted benchmarks ...

/*
 * jfdctint.c
 *
 * This file is part of the Independent JPEG Group's software.
 *
 * The authors make NO WARRANTY or representation, either express or implied,
 * with respect to this software, its quality, accuracy, merchantability, or
 * fitness for a particular purpose.  This software is provided "AS IS", and
 * you, its user, assume the entire risk as to its quality and accuracy.
 *
 * This software is copyright (C) 1991-1996, Thomas G. Lane.
 * All Rights Reserved except as specified below.
 *
 * Permission is hereby granted to use, copy, modify, and distribute this
 * software (or portions thereof) for any purpose, without fee, subject to
 * these conditions:
 * (1) If any part of the source code for this software is distributed, then
 * this README file must be included, with this copyright and no-warranty
 * notice unaltered; and any additions, deletions, or changes to the original
 * files must be clearly indicated in accompanying documentation.
 * (2) If only executable code is distributed, then the accompanying
 * documentation must state that "this software is based in part on the work
 * of the Independent JPEG Group".
 * (3) Permission for use of this software is granted only if the user accepts
 * full responsibility for any undesirable consequences; the authors accept
 * NO LIABILITY for damages of any kind.
 *
 * These conditions apply to any software derived from or based on the IJG
 * code, not just to the unmodified library.  If you use our work, you ought
 * to acknowledge us.
 *
 * Permission is NOT granted for the use of any IJG author's name or company
 * name in advertising or publicity relating to this software or products
 * derived from it.  This software may be referred to only as "the Independent
 * JPEG Group's software".
 *
 * We specifically permit and encourage the use of this software as the basis
 * of commercial products, provided that all warranty or liability claims are
 * assumed by the product vendor.
 *
 * This file contains a slow-but-accurate integer implementation of the
 * forward DCT (Discrete Cosine Transform).
 *
 * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
 * on each column.  Direct algorithms are also available, but they are
 * much more complex and seem not to be any faster when reduced to code.
 *
 * This implementation is based on an algorithm described in
 *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
 *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
 *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
 * The primary algorithm described there uses 11 multiplies and 29 adds.
 * We use their alternate method with 12 multiplies and 32 adds.
 * The advantage of this method is that no data path contains more than one
 * multiplication; this allows a very simple and accurate implementation in
 * scaled fixed-point arithmetic, with a minimal number of shifts.
 */

/**
 * @file jfdctint.c
 * Independent JPEG Group's slow & accurate dct.
 */

#include <stdlib.h>
#include <stdio.h>
#include "common.h"
#include "dsputil.h"

#define SHIFT_TEMPS
#define DCTSIZE 8
#define BITS_IN_JSAMPLE 8
#define GLOBAL(x) x
#define RIGHT_SHIFT(x, n) ((x) >> (n))
#define MULTIPLY16C16(var,const) ((var)*(const))

#if 1 //def USE_ACCURATE_ROUNDING
#define DESCALE(x,n)  RIGHT_SHIFT((x) + (1 << ((n) - 1)), n)
#else
#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
#endif


/*
 * This module is specialized to the case DCTSIZE = 8.
 */

#if DCTSIZE != 8
  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
#endif


/*
 * The poop on this scaling stuff is as follows:
 *
 * Each 1-D DCT step produces outputs which are a factor of sqrt(N)
 * larger than the true DCT outputs.  The final outputs are therefore
 * a factor of N larger than desired; since N=8 this can be cured by
 * a simple right shift at the end of the algorithm.  The advantage of
 * this arrangement is that we save two multiplications per 1-D DCT,
 * because the y0 and y4 outputs need not be divided by sqrt(N).
 * In the IJG code, this factor of 8 is removed by the quantization step
 * (in jcdctmgr.c), NOT in this module.
 *
 * We have to do addition and subtraction of the integer inputs, which
 * is no problem, and multiplication by fractional constants, which is
 * a problem to do in integer arithmetic.  We multiply all the constants
 * by CONST_SCALE and convert them to integer constants (thus retaining
 * CONST_BITS bits of precision in the constants).  After doing a
 * multiplication we have to divide the product by CONST_SCALE, with proper
 * rounding, to produce the correct output.  This division can be done
 * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
 * as long as possible so that partial sums can be added together with
 * full fractional precision.
 *
 * The outputs of the first pass are scaled up by PASS1_BITS bits so that
 * they are represented to better-than-integral precision.  These outputs
 * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
 * with the recommended scaling.  (For 12-bit sample data, the intermediate
 * array is int32_t anyway.)
 *
 * To avoid overflow of the 32-bit intermediate results in pass 2, we must
 * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
 * shows that the values given below are the most effective.
 */

#if BITS_IN_JSAMPLE == 8
#define CONST_BITS  13
#define PASS1_BITS  4   /* set this to 2 if 16x16 multiplies are faster */
#else
#define CONST_BITS  13
#define PASS1_BITS  1   /* lose a little precision to avoid overflow */
#endif

/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
 * causing a lot of useless floating-point operations at run time.
 * To get around this we use the following pre-calculated constants.
 * If you change CONST_BITS you may want to add appropriate values.
 * (With a reasonable C compiler, you can just rely on the FIX() macro...)
 */

#if CONST_BITS == 13
#define FIX_0_298631336  ((int32_t)  2446)      /* FIX(0.298631336) */
#define FIX_0_390180644  ((int32_t)  3196)      /* FIX(0.390180644) */
#define FIX_0_541196100  ((int32_t)  4433)      /* FIX(0.541196100) */
#define FIX_0_765366865  ((int32_t)  6270)      /* FIX(0.765366865) */
#define FIX_0_899976223  ((int32_t)  7373)      /* FIX(0.899976223) */
#define FIX_1_175875602  ((int32_t)  9633)      /* FIX(1.175875602) */
#define FIX_1_501321110  ((int32_t)  12299)     /* FIX(1.501321110) */
#define FIX_1_847759065  ((int32_t)  15137)     /* FIX(1.847759065) */
#define FIX_1_961570560  ((int32_t)  16069)     /* FIX(1.961570560) */
#define FIX_2_053119869  ((int32_t)  16819)     /* FIX(2.053119869) */
#define FIX_2_562915447  ((int32_t)  20995)     /* FIX(2.562915447) */
#define FIX_3_072711026  ((int32_t)  25172)     /* FIX(3.072711026) */
#else
#define FIX_0_298631336  FIX(0.298631336)
#define FIX_0_390180644  FIX(0.390180644)
#define FIX_0_541196100  FIX(0.541196100)
#define FIX_0_765366865  FIX(0.765366865)
#define FIX_0_899976223  FIX(0.899976223)
#define FIX_1_175875602  FIX(1.175875602)
#define FIX_1_501321110  FIX(1.501321110)
#define FIX_1_847759065  FIX(1.847759065)
#define FIX_1_961570560  FIX(1.961570560)
#define FIX_2_053119869  FIX(2.053119869)
#define FIX_2_562915447  FIX(2.562915447)
#define FIX_3_072711026  FIX(3.072711026)
#endif


/* Multiply an int32_t variable by an int32_t constant to yield an int32_t result.
 * For 8-bit samples with the recommended scaling, all the variable
 * and constant values involved are no more than 16 bits wide, so a
 * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
 * For 12-bit samples, a full 32-bit multiplication will be needed.
 */

#if BITS_IN_JSAMPLE == 8 && CONST_BITS<=13 && PASS1_BITS<=2
#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
#else
#define MULTIPLY(var,const)  ((var) * (const))
#endif


static always_inline void row_fdct(DCTELEM * data){
  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  int_fast32_t tmp10, tmp11, tmp12, tmp13;
  int_fast32_t z1, z2, z3, z4, z5;
  DCTELEM *dataptr;
  int ctr;
  SHIFT_TEMPS

  /* Pass 1: process rows. */
  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
  /* furthermore, we scale the results by 2**PASS1_BITS. */

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
    tmp0 = dataptr[0] + dataptr[7];
    tmp7 = dataptr[0] - dataptr[7];
    tmp1 = dataptr[1] + dataptr[6];
    tmp6 = dataptr[1] - dataptr[6];
    tmp2 = dataptr[2] + dataptr[5];
    tmp5 = dataptr[2] - dataptr[5];
    tmp3 = dataptr[3] + dataptr[4];
    tmp4 = dataptr[3] - dataptr[4];

    /* Even part per LL&M figure 1 --- note that published figure is faulty;
     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
     */

    tmp10 = tmp0 + tmp3;
    tmp13 = tmp0 - tmp3;
    tmp11 = tmp1 + tmp2;
    tmp12 = tmp1 - tmp2;

    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);

    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
    dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
                                   CONST_BITS-PASS1_BITS);
    dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
                                   CONST_BITS-PASS1_BITS);

    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
     * cK represents cos(K*pi/16).
     * i0..i3 in the paper are tmp4..tmp7 here.
     */

    z1 = tmp4 + tmp7;
    z2 = tmp5 + tmp6;
    z3 = tmp4 + tmp6;
    z4 = tmp5 + tmp7;
    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */

    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */

    z3 += z5;
    z4 += z5;

    dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
    dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
    dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
    dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);

    dataptr += DCTSIZE;         /* advance pointer to next row */
  }
}

/*
 * Perform the forward DCT on one block of samples.
 */

GLOBAL(void)
ff_jpeg_fdct_islow (DCTELEM * data)
{
  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  int_fast32_t tmp10, tmp11, tmp12, tmp13;
  int_fast32_t z1, z2, z3, z4, z5;
  DCTELEM *dataptr;
  int ctr;
  SHIFT_TEMPS

  row_fdct(data);

  /* Pass 2: process columns.
   * We remove the PASS1_BITS scaling, but leave the results scaled up
   * by an overall factor of 8.
   */

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];

    /* Even part per LL&M figure 1 --- note that published figure is faulty;
     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
     */

    tmp10 = tmp0 + tmp3;
    tmp13 = tmp0 - tmp3;
    tmp11 = tmp1 + tmp2;
    tmp12 = tmp1 - tmp2;

    dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
    dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);

    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
    dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
                                           CONST_BITS+PASS1_BITS);
    dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
                                           CONST_BITS+PASS1_BITS);

    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
     * cK represents cos(K*pi/16).
     * i0..i3 in the paper are tmp4..tmp7 here.
     */

    z1 = tmp4 + tmp7;
    z2 = tmp5 + tmp6;
    z3 = tmp4 + tmp6;
    z4 = tmp5 + tmp7;
    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */

    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */

    z3 += z5;
    z4 += z5;

    dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
                                           CONST_BITS+PASS1_BITS);
    dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
                                           CONST_BITS+PASS1_BITS);
    dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
                                           CONST_BITS+PASS1_BITS);
    dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
                                           CONST_BITS+PASS1_BITS);

    dataptr++;                  /* advance pointer to next column */
  }
}

/*
 * The secret of DCT2-4-8 is really simple -- you do the usual 1-DCT
 * on the rows and then, instead of doing even and odd, part on the colums
 * you do even part two times.
 */
GLOBAL(void)
ff_fdct248_islow (DCTELEM * data)
{
  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  int_fast32_t tmp10, tmp11, tmp12, tmp13;
  int_fast32_t z1;
  DCTELEM *dataptr;
  int ctr;
  SHIFT_TEMPS

  row_fdct(data);

  /* Pass 2: process columns.
   * We remove the PASS1_BITS scaling, but leave the results scaled up
   * by an overall factor of 8.
   */

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
     tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*1];
     tmp1 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3];
     tmp2 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*6] + dataptr[DCTSIZE*7];
     tmp4 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*1];
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3];
     tmp6 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*5];
     tmp7 = dataptr[DCTSIZE*6] - dataptr[DCTSIZE*7];

     tmp10 = tmp0 + tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
     tmp13 = tmp0 - tmp3;

     dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
     dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);

     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
                                            CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
                                            CONST_BITS+PASS1_BITS);

     tmp10 = tmp4 + tmp7;
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp5 - tmp6;
     tmp13 = tmp4 - tmp7;

     dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
     dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);

     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
                                            CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
                                            CONST_BITS+PASS1_BITS);

     dataptr++;                 /* advance pointer to next column */
  }
}