dsputil.h

64-bits H.264 from ffmpeg 2008 version Build in VC++ 2008 no error warning, Jesse Stone, Taiwan

     * (-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 */
    void (*h264_idct_add)(uint8_t *dst, DCTELEM *block, int stride);
    void (*h264_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
    void (*h264_idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
    void (*h264_idct8_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
    void (*h264_dct)(DCTELEM block[4][4]);

    /* 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);

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

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

/*static  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_armv4l(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 defined(HAVE_MMX)

#undef emms_c

#define MM_MMX    0x0001 /* standard MMX */
#define MM_3DNOW  0x0004 /* AMD 3DNOW */
#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
#define MM_SSE    0x0008 /* SSE functions */
#define MM_SSE2   0x0010 /* PIV SSE2 functions */
#define MM_3DNOWEXT  0x0020 /* AMD 3DNowExt */
#define MM_SSE3   0x0040 /* Prescott SSE3 functions */
#define MM_SSSE3  0x0080 /* Conroe SSSE3 functions */

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);

static  void emms(void)
{
    asm volatile ("emms;":::"memory");
}


#define emms_c() \
{\
    if (mm_flags & MM_MMX)\
        emms();\
}

void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);

#elif defined(ARCH_ARMV4L)

#define MM_IWMMXT    0x0100 /* XScale IWMMXT */

extern int mm_flags;

#elif defined(ARCH_POWERPC)

#define MM_ALTIVEC    0x0001 /* standard AltiVec */

extern int mm_flags;

#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
#define STRIDE_ALIGN 16

#elif defined(HAVE_MMI)

#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
#define STRIDE_ALIGN 16

#endif

#ifndef DECLARE_ALIGNED_8
#   define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
#endif

#ifndef STRIDE_ALIGN
#   define STRIDE_ALIGN 8
#endif

/* PSNR */
void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
              int orig_linesize[3], int coded_linesize,
              AVCodecContext *avctx);

/* FFT computation */

/* NOTE: soon integer code will be added, so you must use the
   FFTSample type */
typedef float FFTSample;

struct MDCTContext;

typedef struct FFTComplex {
    FFTSample re, im;
} FFTComplex;

typedef struct FFTContext {
    int nbits;
    int inverse;
    uint16_t *revtab;
    FFTComplex *exptab;
    FFTComplex *exptab1; /* only used by SSE code */
    void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
    void (*imdct_calc)(struct MDCTContext *s, FFTSample *output,
                       const FFTSample *input, FFTSample *tmp);
} FFTContext;

int ff_fft_init(FFTContext *s, int nbits, int inverse);
void ff_fft_permute(FFTContext *s, FFTComplex *z);
void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z);
void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z);

/*static  void ff_fft_calc(FFTContext *s, FFTComplex *z)
{
    s->fft_calc(s, z);
}*/
//void ff_fft_end(FFTContext *s);

/* MDCT computation */

typedef struct MDCTContext {
    int n;  /* size of MDCT (i.e. number of input data * 2) */
    int nbits; /* n = 2^nbits */
    /* pre/post rotation tables */
    FFTSample *tcos;
    FFTSample *tsin;
    FFTContext fft;
} MDCTContext;

/**
 * Generate a Kaiser-Bessel Derived Window.
 * @param   window  pointer to half window
 * @param   alpha   determines window shape
 * @param   n       size of half window
 */
void ff_kbd_window_init(float *window, float alpha, int n);

int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
void ff_imdct_calc(MDCTContext *s, FFTSample *output,
                const FFTSample *input, FFTSample *tmp);
void ff_imdct_calc_3dn2(MDCTContext *s, FFTSample *output,
                        const FFTSample *input, FFTSample *tmp);
void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output,
                       const FFTSample *input, FFTSample *tmp);
void ff_mdct_calc(MDCTContext *s, FFTSample *out,
               const FFTSample *input, FFTSample *tmp);
void ff_mdct_end(MDCTContext *s);

#define WRAPPER8_16(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
    return name8(s, dst           , src           , stride, h)\
          +name8(s, dst+8         , src+8         , stride, h);\
}

#define WRAPPER8_16_SQ(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
    int score=0;\
    score +=name8(s, dst           , src           , stride, 8);\
    score +=name8(s, dst+8         , src+8         , stride, 8);\
    if(h==16){\
        dst += 8*stride;\
        src += 8*stride;\
        score +=name8(s, dst           , src           , stride, 8);\
        score +=name8(s, dst+8         , src+8         , stride, 8);\
    }\
    return score;\
}


static  void copy_block2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN16(dst   , AV_RN16(src   ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static  void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static  void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static  void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        dst[8]= src[8];
        dst+=dstStride;
        src+=srcStride;
    }
}

static  void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        AV_WN32(dst+8 , AV_RN32(src+8 ));
        AV_WN32(dst+12, AV_RN32(src+12));
        dst+=dstStride;
        src+=srcStride;
    }
}

static  void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        AV_WN32(dst+8 , AV_RN32(src+8 ));
        AV_WN32(dst+12, AV_RN32(src+12));
        dst[16]= src[16];
        dst+=dstStride;
        src+=srcStride;
    }
}

#endif /* FFMPEG_DSPUTIL_H */