dsputil.h

播放H264文件的播放器

     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
    tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?

    qpel_mc_func put_qpel_pixels_tab[2][16];
    qpel_mc_func avg_qpel_pixels_tab[2][16];
    qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
    qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
    qpel_mc_func put_mspel_pixels_tab[8];

    /**
     * h264 Chroma MC
     */
    h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
    /* This is really one func used in VC-1 decoding */
    h264_chroma_mc_func put_no_rnd_h264_chroma_pixels_tab[3];
    h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];

    qpel_mc_func put_h264_qpel_pixels_tab[4][16];
    qpel_mc_func avg_h264_qpel_pixels_tab[4][16];

    qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
    qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];

    h264_weight_func weight_h264_pixels_tab[10];
    h264_biweight_func biweight_h264_pixels_tab[10];

    /* AVS specific */
    qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
    qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
    void (*cavs_filter_lv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
    void (*cavs_filter_lh)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
    void (*cavs_filter_cv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
    void (*cavs_filter_ch)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
    void (*cavs_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);

    me_cmp_func pix_abs[2][4];

    /* huffyuv specific */
    void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
    void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
    void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
    /**
     * subtract huffyuv's variant of median prediction
     * note, this might read from src1[-1], src2[-1]
     */
    void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
    void (*add_hfyu_median_prediction)(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top);
    /* this might write to dst[w] */
    void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
    void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);

    void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
    void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
    /* v/h_loop_filter_luma_intra: align 16 */
    void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
    void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
    void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
    void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
    void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
    void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
    // h264_loop_filter_strength: simd only. the C version is inlined in h264.c
    void (*h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
                                      int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);

    void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
    void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);

    void (*h261_loop_filter)(uint8_t *src, int stride);

    void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
    void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);

    void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
    void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);

    void (*vp6_filter_diag4)(uint8_t *dst, uint8_t *src, int stride,
                             const int16_t *h_weights,const int16_t *v_weights);

    /* assume len is a multiple of 4, and arrays are 16-byte aligned */
    void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
    void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
    /* no alignment needed */
    void (*flac_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
    /* assume len is a multiple of 8, and arrays are 16-byte aligned */
    void (*vector_fmul)(float *dst, const float *src, int len);
    void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
    /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
    void (*vector_fmul_add_add)(float *dst, const float *src0, const float *src1, const float *src2, int src3, int len, int step);
    /* assume len is a multiple of 4, and arrays are 16-byte aligned */
    void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
    /* assume len is a multiple of 8, and arrays are 16-byte aligned */
    void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);

    /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
     * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
    void (*float_to_int16)(int16_t *dst, const float *src, long len);
    void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);

    /* (I)DCT */
    void (*fdct)(DCTELEM *block/* align 16*/);
    void (*fdct248)(DCTELEM *block/* align 16*/);

    /* IDCT really*/
    void (*idct)(DCTELEM *block/* align 16*/);

    /**
     * block -> idct -> clip to unsigned 8 bit -> dest.
     * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);

    /**
     * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);

    /**
     * idct input permutation.
     * several optimized IDCTs need a permutated input (relative to the normal order of the reference
     * IDCT)
     * this permutation must be performed before the idct_put/add, note, normally this can be merged
     * with the zigzag/alternate scan<br>
     * an example to avoid confusion:
     * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
     * - (x -> referece dct -> reference idct -> x)
     * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
     * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
     */
    uint8_t idct_permutation[64];
    int idct_permutation_type;
#define FF_NO_IDCT_PERM 1
#define FF_LIBMPEG2_IDCT_PERM 2
#define FF_SIMPLE_IDCT_PERM 3
#define FF_TRANSPOSE_IDCT_PERM 4
#define FF_PARTTRANS_IDCT_PERM 5
#define FF_SSE2_IDCT_PERM 6

    int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
    void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
#define BASIS_SHIFT 16
#define RECON_SHIFT 6

    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
#define EDGE_WIDTH 16

    /* h264 functions */
    /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
       NOTE!!! if you implement any of h264_idct_add, h264_idct_add16, h264_idct_add16intra, h264_idct_add8 then you must implement all of them
        The reason for above, is that no 2 out of one list may use a different permutation.
    */
    void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
    void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
    void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
    void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
    void (*h264_dct)(DCTELEM block[4][4]);
    void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
    void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
    void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
    void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);

    /* snow wavelet */
    void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
    void (*horizontal_compose97i)(IDWTELEM *b, int width);
    void (*inner_add_yblock)(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);

    void (*prefetch)(void *mem, int stride, int h);

    void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);

    /* vc1 functions */
    void (*vc1_inv_trans_8x8)(DCTELEM *b);
    void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
    void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
    void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
    void (*vc1_v_overlap)(uint8_t* src, int stride);
    void (*vc1_h_overlap)(uint8_t* src, int stride);
    /* put 8x8 block with bicubic interpolation and quarterpel precision
     * last argument is actually round value instead of height
     */
    op_pixels_func put_vc1_mspel_pixels_tab[16];

    /* intrax8 functions */
    void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
    void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
           int * range, int * sum,  int edges);

    /* ape functions */
    /**
     * Add contents of the second vector to the first one.
     * @param len length of vectors, should be multiple of 16
     */
    void (*add_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
    /**
     * Add contents of the second vector to the first one.
     * @param len length of vectors, should be multiple of 16
     */
    void (*sub_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
    /**
     * Calculate scalar product of two vectors.
     * @param len length of vectors, should be multiple of 16
     * @param shift number of bits to discard from product
     */
    int32_t (*scalarproduct_int16)(int16_t *v1, int16_t *v2/*align 16*/, int len, int shift);

    /* rv30 functions */
    qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
    qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];

    /* rv40 functions */
    qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
    qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
    h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
    h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
} DSPContext;

void dsputil_static_init(void);
void dsputil_init(DSPContext* p, AVCodecContext *avctx);

int ff_check_alignment(void);

/**
 * permute block according to permuatation.
 * @param last last non zero element in scantable order
 */
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);

void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);

#define         BYTE_VEC32(c)   ((c)*0x01010101UL)

static _inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
{
    return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
}

static _inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
{
    return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
}

static _inline int get_penalty_factor(int lambda, int lambda2, int type){
    switch(type&0xFF){
    default:
    case FF_CMP_SAD:
        return lambda>>FF_LAMBDA_SHIFT;
    case FF_CMP_DCT:
        return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
    case FF_CMP_W53:
        return (4*lambda)>>(FF_LAMBDA_SHIFT);
    case FF_CMP_W97:
        return (2*lambda)>>(FF_LAMBDA_SHIFT);
    case FF_CMP_SATD:
    case FF_CMP_DCT264:
        return (2*lambda)>>FF_LAMBDA_SHIFT;
    case FF_CMP_RD:
    case FF_CMP_PSNR:
    case FF_CMP_SSE:
    case FF_CMP_NSSE:
        return lambda2>>FF_LAMBDA_SHIFT;
    case FF_CMP_BIT:
        return 1;
    }
}

/**
 * Empty mmx state.
 * this must be called between any dsp function and float/double code.
 * for example sin(); dsp->idct_put(); emms_c(); cos()
 */
#define emms_c()

/* should be defined by architectures supporting
   one or more MultiMedia extension */
int mm_support(void);

void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);

#define DECLARE_ALIGNED_16(t, v) DECLARE_ALIGNED(16, t, v)

#if HAVE_MMX

#undef emms_c

extern int mm_flags;

void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);