mpegaudiodec.c

Trolltech公司发布的图形界面操作系统。可在qt-embedded-2.3.10平台上编译为嵌入式图形界面操作系统。

        for(i=0;i<16;i++) {
            double f;
            int e, k;

            for(j=0;j<2;j++) {
                e = -(j + 1) * ((i + 1) >> 1);
                f = pow(2.0, e / 4.0);
                k = i & 1;
                is_table_lsf[j][k ^ 1][i] = FIXR(f);
                is_table_lsf[j][k][i] = FIXR(1.0);
                dprintf("is_table_lsf %d %d: %x %x\n", 
                        i, j, is_table_lsf[j][0][i], is_table_lsf[j][1][i]);
            }
        }

        for(i=0;i<8;i++) {
            float ci, cs, ca;
            ci = ci_table[i];
            cs = 1.0 / sqrt(1.0 + ci * ci);
            ca = cs * ci;
            csa_table[i][0] = FIX(cs);
            csa_table[i][1] = FIX(ca);
        }

        /* compute mdct windows */
        for(i=0;i<36;i++) {
            int v;
            v = FIXR(sin(M_PI * (i + 0.5) / 36.0));
            mdct_win[0][i] = v;
            mdct_win[1][i] = v;
            mdct_win[3][i] = v;
        }
        for(i=0;i<6;i++) {
            mdct_win[1][18 + i] = FIXR(1.0);
            mdct_win[1][24 + i] = FIXR(sin(M_PI * ((i + 6) + 0.5) / 12.0));
            mdct_win[1][30 + i] = FIXR(0.0);

            mdct_win[3][i] = FIXR(0.0);
            mdct_win[3][6 + i] = FIXR(sin(M_PI * (i + 0.5) / 12.0));
            mdct_win[3][12 + i] = FIXR(1.0);
        }

        for(i=0;i<12;i++)
            mdct_win[2][i] = FIXR(sin(M_PI * (i + 0.5) / 12.0));
        
        /* NOTE: we do frequency inversion adter the MDCT by changing
           the sign of the right window coefs */
        for(j=0;j<4;j++) {
            for(i=0;i<36;i+=2) {
                mdct_win[j + 4][i] = mdct_win[j][i];
                mdct_win[j + 4][i + 1] = -mdct_win[j][i + 1];
            }
        }

#if defined(DEBUG)
        for(j=0;j<8;j++) {
            printf("win%d=\n", j);
            for(i=0;i<36;i++)
                printf("%f, ", (double)mdct_win[j][i] / FRAC_ONE);
            printf("\n");
        }
#endif
        init = 1;
    }

    s->inbuf_index = 0;
    s->inbuf = &s->inbuf1[s->inbuf_index][BACKSTEP_SIZE];
    s->inbuf_ptr = s->inbuf;
#ifdef DEBUG
    s->frame_count = 0;
#endif
    return 0;
}

/* tab[i][j] = 1.0 / (2.0 * cos(pi*(2*k+1) / 2^(6 - j))) */

/* cos(i*pi/64) */

#define COS0_0  FIXR(0.50060299823519630134)
#define COS0_1  FIXR(0.50547095989754365998)
#define COS0_2  FIXR(0.51544730992262454697)
#define COS0_3  FIXR(0.53104259108978417447)
#define COS0_4  FIXR(0.55310389603444452782)
#define COS0_5  FIXR(0.58293496820613387367)
#define COS0_6  FIXR(0.62250412303566481615)
#define COS0_7  FIXR(0.67480834145500574602)
#define COS0_8  FIXR(0.74453627100229844977)
#define COS0_9  FIXR(0.83934964541552703873)
#define COS0_10 FIXR(0.97256823786196069369)
#define COS0_11 FIXR(1.16943993343288495515)
#define COS0_12 FIXR(1.48416461631416627724)
#define COS0_13 FIXR(2.05778100995341155085)
#define COS0_14 FIXR(3.40760841846871878570)
#define COS0_15 FIXR(10.19000812354805681150)

#define COS1_0 FIXR(0.50241928618815570551)
#define COS1_1 FIXR(0.52249861493968888062)
#define COS1_2 FIXR(0.56694403481635770368)
#define COS1_3 FIXR(0.64682178335999012954)
#define COS1_4 FIXR(0.78815462345125022473)
#define COS1_5 FIXR(1.06067768599034747134)
#define COS1_6 FIXR(1.72244709823833392782)
#define COS1_7 FIXR(5.10114861868916385802)

#define COS2_0 FIXR(0.50979557910415916894)
#define COS2_1 FIXR(0.60134488693504528054)
#define COS2_2 FIXR(0.89997622313641570463)
#define COS2_3 FIXR(2.56291544774150617881)

#define COS3_0 FIXR(0.54119610014619698439)
#define COS3_1 FIXR(1.30656296487637652785)

#define COS4_0 FIXR(0.70710678118654752439)

/* butterfly operator */
#define BF(a, b, c)\
{\
    tmp0 = tab[a] + tab[b];\
    tmp1 = tab[a] - tab[b];\
    tab[a] = tmp0;\
    tab[b] = MULL(tmp1, c);\
}

#define BF1(a, b, c, d)\
{\
    BF(a, b, COS4_0);\
    BF(c, d, -COS4_0);\
    tab[c] += tab[d];\
}

#define BF2(a, b, c, d)\
{\
    BF(a, b, COS4_0);\
    BF(c, d, -COS4_0);\
    tab[c] += tab[d];\
    tab[a] += tab[c];\
    tab[c] += tab[b];\
    tab[b] += tab[d];\
}

#define ADD(a, b) tab[a] += tab[b]

/* DCT32 without 1/sqrt(2) coef zero scaling. */
static void dct32(int32_t *out, int32_t *tab)
{
    int tmp0, tmp1;

    /* pass 1 */
    BF(0, 31, COS0_0);
    BF(1, 30, COS0_1);
    BF(2, 29, COS0_2);
    BF(3, 28, COS0_3);
    BF(4, 27, COS0_4);
    BF(5, 26, COS0_5);
    BF(6, 25, COS0_6);
    BF(7, 24, COS0_7);
    BF(8, 23, COS0_8);
    BF(9, 22, COS0_9);
    BF(10, 21, COS0_10);
    BF(11, 20, COS0_11);
    BF(12, 19, COS0_12);
    BF(13, 18, COS0_13);
    BF(14, 17, COS0_14);
    BF(15, 16, COS0_15);

    /* pass 2 */
    BF(0, 15, COS1_0);
    BF(1, 14, COS1_1);
    BF(2, 13, COS1_2);
    BF(3, 12, COS1_3);
    BF(4, 11, COS1_4);
    BF(5, 10, COS1_5);
    BF(6,  9, COS1_6);
    BF(7,  8, COS1_7);
    
    BF(16, 31, -COS1_0);
    BF(17, 30, -COS1_1);
    BF(18, 29, -COS1_2);
    BF(19, 28, -COS1_3);
    BF(20, 27, -COS1_4);
    BF(21, 26, -COS1_5);
    BF(22, 25, -COS1_6);
    BF(23, 24, -COS1_7);
    
    /* pass 3 */
    BF(0, 7, COS2_0);
    BF(1, 6, COS2_1);
    BF(2, 5, COS2_2);
    BF(3, 4, COS2_3);
    
    BF(8, 15, -COS2_0);
    BF(9, 14, -COS2_1);
    BF(10, 13, -COS2_2);
    BF(11, 12, -COS2_3);
    
    BF(16, 23, COS2_0);
    BF(17, 22, COS2_1);
    BF(18, 21, COS2_2);
    BF(19, 20, COS2_3);
    
    BF(24, 31, -COS2_0);
    BF(25, 30, -COS2_1);
    BF(26, 29, -COS2_2);
    BF(27, 28, -COS2_3);

    /* pass 4 */
    BF(0, 3, COS3_0);
    BF(1, 2, COS3_1);
    
    BF(4, 7, -COS3_0);
    BF(5, 6, -COS3_1);
    
    BF(8, 11, COS3_0);
    BF(9, 10, COS3_1);
    
    BF(12, 15, -COS3_0);
    BF(13, 14, -COS3_1);
    
    BF(16, 19, COS3_0);
    BF(17, 18, COS3_1);
    
    BF(20, 23, -COS3_0);
    BF(21, 22, -COS3_1);
    
    BF(24, 27, COS3_0);
    BF(25, 26, COS3_1);
    
    BF(28, 31, -COS3_0);
    BF(29, 30, -COS3_1);
    
    /* pass 5 */
    BF1(0, 1, 2, 3);
    BF2(4, 5, 6, 7);
    BF1(8, 9, 10, 11);
    BF2(12, 13, 14, 15);
    BF1(16, 17, 18, 19);
    BF2(20, 21, 22, 23);
    BF1(24, 25, 26, 27);
    BF2(28, 29, 30, 31);
    
    /* pass 6 */
    
    ADD( 8, 12);
    ADD(12, 10);
    ADD(10, 14);
    ADD(14,  9);
    ADD( 9, 13);
    ADD(13, 11);
    ADD(11, 15);

    out[ 0] = tab[0];
    out[16] = tab[1];
    out[ 8] = tab[2];
    out[24] = tab[3];
    out[ 4] = tab[4];
    out[20] = tab[5];
    out[12] = tab[6];
    out[28] = tab[7];
    out[ 2] = tab[8];
    out[18] = tab[9];
    out[10] = tab[10];
    out[26] = tab[11];
    out[ 6] = tab[12];
    out[22] = tab[13];
    out[14] = tab[14];
    out[30] = tab[15];
    
    ADD(24, 28);
    ADD(28, 26);
    ADD(26, 30);
    ADD(30, 25);
    ADD(25, 29);
    ADD(29, 27);
    ADD(27, 31);

    out[ 1] = tab[16] + tab[24];
    out[17] = tab[17] + tab[25];
    out[ 9] = tab[18] + tab[26];
    out[25] = tab[19] + tab[27];
    out[ 5] = tab[20] + tab[28];
    out[21] = tab[21] + tab[29];
    out[13] = tab[22] + tab[30];
    out[29] = tab[23] + tab[31];
    out[ 3] = tab[24] + tab[20];
    out[19] = tab[25] + tab[21];
    out[11] = tab[26] + tab[22];
    out[27] = tab[27] + tab[23];
    out[ 7] = tab[28] + tab[18];
    out[23] = tab[29] + tab[19];
    out[15] = tab[30] + tab[17];
    out[31] = tab[31];
}

#define OUT_SHIFT (WFRAC_BITS + FRAC_BITS - 15)

#if FRAC_BITS <= 15

static inline int round_sample(int sum)
{
    int sum1;
    sum1 = (sum + (1 << (OUT_SHIFT - 1))) >> OUT_SHIFT;
    if (sum1 < -32768)
        sum1 = -32768;
    else if (sum1 > 32767)
        sum1 = 32767;
    return sum1;
}

#if defined(ARCH_POWERPC_405)

/* signed 16x16 -> 32 multiply add accumulate */
#define MACS(rt, ra, rb) \
    asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb));

/* signed 16x16 -> 32 multiply */
#define MULS(ra, rb) \
    ({ int __rt; asm ("mullhw %0, %1, %2" : "=r" (__rt) : "r" (ra), "r" (rb)); __rt; })

#else

/* signed 16x16 -> 32 multiply add accumulate */
#define MACS(rt, ra, rb) rt += (ra) * (rb)

/* signed 16x16 -> 32 multiply */
#define MULS(ra, rb) ((ra) * (rb))

#endif

#else

static inline int round_sample(int64_t sum) 
{
    int sum1;
    sum1 = (int)((sum + (int64_t_C(1) << (OUT_SHIFT - 1))) >> OUT_SHIFT);
    if (sum1 < -32768)
        sum1 = -32768;
    else if (sum1 > 32767)
        sum1 = 32767;
    return sum1;
}

#define MULS(ra, rb) MUL64(ra, rb)

#endif

#define SUM8(sum, op, w, p) \
{                                               \
    sum op MULS((w)[0 * 64], p[0 * 64]);\
    sum op MULS((w)[1 * 64], p[1 * 64]);\
    sum op MULS((w)[2 * 64], p[2 * 64]);\
    sum op MULS((w)[3 * 64], p[3 * 64]);\
    sum op MULS((w)[4 * 64], p[4 * 64]);\
    sum op MULS((w)[5 * 64], p[5 * 64]);\
    sum op MULS((w)[6 * 64], p[6 * 64]);\
    sum op MULS((w)[7 * 64], p[7 * 64]);\
}

#define SUM8P2(sum1, op1, sum2, op2, w1, w2, p) \
{                                               \
    int tmp;\
    tmp = p[0 * 64];\
    sum1 op1 MULS((w1)[0 * 64], tmp);\
    sum2 op2 MULS((w2)[0 * 64], tmp);\
    tmp = p[1 * 64];\
    sum1 op1 MULS((w1)[1 * 64], tmp);\
    sum2 op2 MULS((w2)[1 * 64], tmp);\
    tmp = p[2 * 64];\
    sum1 op1 MULS((w1)[2 * 64], tmp);\
    sum2 op2 MULS((w2)[2 * 64], tmp);\
    tmp = p[3 * 64];\
    sum1 op1 MULS((w1)[3 * 64], tmp);\
    sum2 op2 MULS((w2)[3 * 64], tmp);\
    tmp = p[4 * 64];\
    sum1 op1 MULS((w1)[4 * 64], tmp);\
    sum2 op2 MULS((w2)[4 * 64], tmp);\
    tmp = p[5 * 64];\
    sum1 op1 MULS((w1)[5 * 64], tmp);\
    sum2 op2 MULS((w2)[5 * 64], tmp);\
    tmp = p[6 * 64];\
    sum1 op1 MULS((w1)[6 * 64], tmp);\
    sum2 op2 MULS((w2)[6 * 64], tmp);\
    tmp = p[7 * 64];\
    sum1 op1 MULS((w1)[7 * 64], tmp);\
    sum2 op2 MULS((w2)[7 * 64], tmp);\
}


/* 32 sub band synthesis filter. Input: 32 sub band samples, Output:
   32 samples. */
/* XXX: optimize by avoiding ring buffer usage */
static void synth_filter(MPADecodeContext *s1,
                         int ch, int16_t *samples, int incr, 
                         int32_t sb_samples[SBLIMIT])
{
    int32_t tmp[32];
    register MPA_INT *synth_buf;
    const register MPA_INT *w, *w2, *p;
    int j, offset, v;
    int16_t *samples2;
#if FRAC_BITS <= 15
    int sum, sum2;
#else
    int64_t sum, sum2;
#endif
    
    dct32(tmp, sb_samples);
    
    offset = s1->synth_buf_offset[ch];
    synth_buf = s1->synth_buf[ch] + offset;

    for(j=0;j<32;j++) {
        v = tmp[j];
#if FRAC_BITS <= 15
        /* NOTE: can cause a loss in precision if very high amplitude
           sound */
        if (v > 32767)
            v = 32767;
        else if (v < -32768)
            v = -32768;
#endif
        synth_buf[j] = v;
    }
    /* copy to avoid wrap */
    memcpy(synth_buf + 512, synth_buf, 32 * sizeof(MPA_INT));

    samples2 = samples + 31 * incr;
    w = window;
    w2 = window + 31;

    sum = 0;
    p = synth_buf + 16;
    SUM8(sum, +=, w, p);
    p = synth_buf + 48;
    SUM8(sum, -=, w + 32, p);
    *samples = round_sample(sum);
    samples += incr;
    w++;

    /* we calculate two samples at the same time to avoid one memory
       access per two sample */
    for(j=1;j<16;j++) {
        sum = 0;