30 #ifndef AVCODEC_DSPUTIL_H
31 #define AVCODEC_DSPUTIL_H
55 #define H264_IDCT(depth) \
56 void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
57 void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
58 void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
59 void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
60 void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
61 void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
62 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
63 void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
64 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
65 void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\
66 void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\
67 void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);
83 #define MAX_NEG_CROP 1024
89 #define PUTAVG_PIXELS(depth)\
90 void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
91 void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
92 void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
93 void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);
99 #define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
100 #define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
101 #define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
102 #define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c
116 int dxx,
int dxy,
int dyx,
int dyy,
int shift,
int r,
int width,
int height);
149 #define DEF_OLD_QPEL(name)\
150 void ff_put_ ## name (uint8_t *dst, uint8_t *src, int stride);\
151 void ff_put_no_rnd_ ## name (uint8_t *dst, uint8_t *src, int stride);\
152 void ff_avg_ ## name (uint8_t *dst, uint8_t *src, int stride);
167 #define CALL_2X_PIXELS(a, b, n)\
168 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
169 b(block , pixels , line_size, h);\
170 b(block+n, pixels+n, line_size, h);\
189 int idct_permutation_type);
217 int dxx,
int dxy,
int dyx,
int dyy,
int shift,
int r,
int width,
int height);
391 const float *src1,
int len);
427 #define FF_NO_IDCT_PERM 1
428 #define FF_LIBMPEG2_IDCT_PERM 2
429 #define FF_SIMPLE_IDCT_PERM 3
430 #define FF_TRANSPOSE_IDCT_PERM 4
431 #define FF_PARTTRANS_IDCT_PERM 5
432 #define FF_SSE2_IDCT_PERM 6
436 #define BASIS_SHIFT 16
437 #define RECON_SHIFT 6
440 #define EDGE_WIDTH 16
442 #define EDGE_BOTTOM 2
471 const int16_t *window,
unsigned int len);
516 #define BYTE_VEC32(c) ((c)*0x01010101UL)
517 #define BYTE_VEC64(c) ((c)*0x0001000100010001UL)
521 return (a | b) - (((a ^
b) & ~
BYTE_VEC32(0x01)) >> 1);
526 return (a & b) + (((a ^
b) & ~
BYTE_VEC32(0x01)) >> 1);
531 return (a | b) - (((a ^
b) & ~
BYTE_VEC64(0x01)) >> 1);
536 return (a & b) + (((a ^
b) & ~
BYTE_VEC64(0x01)) >> 1);
545 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
573 #if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMX
574 # define STRIDE_ALIGN 16
576 # define STRIDE_ALIGN 8
583 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...) \
584 uint8_t la_##v[sizeof(t s o) + (a)]; \
585 t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
587 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) \
588 DECLARE_ALIGNED(a, t, la_##v) s o; \
591 #define LOCAL_ALIGNED(a, t, v, ...) E(LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,))
593 #if HAVE_LOCAL_ALIGNED_8
594 # define LOCAL_ALIGNED_8(t, v, ...) E(LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,))
596 # define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
599 #if HAVE_LOCAL_ALIGNED_16
600 # define LOCAL_ALIGNED_16(t, v, ...) E(LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,))
602 # define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
605 #define WRAPPER8_16_SQ(name8, name16)\
606 static int name16(void *s, uint8_t *dst, uint8_t *src, int stride, int h){\
608 score +=name8(s, dst , src , stride, 8);\
609 score +=name8(s, dst+8 , src+8 , stride, 8);\
613 score +=name8(s, dst , src , stride, 8);\
614 score +=name8(s, dst+8 , src+8 , stride, 8);\
void(* vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len)
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last)
permute block according to permuatation.
void ff_dsputil_static_init(void)
op_pixels_func put_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
me_cmp_func frame_skip_cmp[6]
void(* add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha)
void(* shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
const uint8_t ff_alternate_vertical_scan[64]
void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx)
void(* idct_add)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
void(* vorbis_inverse_coupling)(float *mag, float *ang, int blocksize)
void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx)
void ff_fdct248_islow_10(DCTELEM *data)
int(* try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale)
void(* draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
int(* me_cmp_func)(void *s, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
const uint8_t ff_zigzag248_direct[64]
void(* apply_window_int16)(int16_t *output, const int16_t *input, const int16_t *window, unsigned int len)
Apply symmetric window in 16-bit fixed-point.
void ff_j_rev_dct(DCTELEM *data)
void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc)
int(* pix_sum)(uint8_t *pix, int line_size)
void ff_dsputil_init(DSPContext *p, AVCodecContext *avctx)
qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]
static uint64_t rnd_avg64(uint64_t a, uint64_t b)
void(* qpel_mc_func)(uint8_t *dst, uint8_t *src, int stride)
h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]
void ff_fdct_mmx(DCTELEM *block)
void(* get_pixels)(DCTELEM *block, const uint8_t *pixels, int line_size)
const uint8_t ff_alternate_horizontal_scan[64]
uint32_t ff_squareTbl[512]
#define DEF_OLD_QPEL(name)
me_cmp_func me_pre_cmp[6]
static void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
void(* add_bytes)(uint8_t *dst, uint8_t *src, int w)
op_pixels_func avg_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
uint8_t idct_permutation[64]
idct input permutation.
const uint8_t * scantable
#define AV_COPY128U(d, s)
static uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
void(* h263_h_loop_filter)(uint8_t *src, int stride, int qscale)
qpel_mc_func put_qpel_pixels_tab[2][16]
#define PUTAVG_PIXELS(depth)
qpel_mc_func put_mspel_pixels_tab[8]
void(* vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len)
void(* add_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
void ff_dsputil_init_alpha(DSPContext *c, AVCodecContext *avctx)
int32_t(* scalarproduct_int16)(const int16_t *v1, const int16_t *v2, int len)
Calculate scalar product of two vectors.
void(* put_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
void(* bswap16_buf)(uint16_t *dst, const uint16_t *src, int len)
void(* sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top)
subtract huffyuv's variant of median prediction note, this might read from src1[-1], src2[-1]
static uint32_t rnd_avg32(uint32_t a, uint32_t b)
tpel_mc_func avg_tpel_pixels_tab[11]
void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
int ff_check_alignment(void)
void(* bswap_buf)(uint32_t *dst, const uint32_t *src, int w)
qpel_mc_func avg_qpel_pixels_tab[2][16]
void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp)
void(* idct_put)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> clip to unsigned 8 bit -> dest.
void ff_fdct_sse2(DCTELEM *block)
int(* add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left)
void(* h264_chroma_mc_func)(uint8_t *dst, uint8_t *src, int srcStride, int h, int x, int y)
void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type)
op_pixels_func put_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
void ff_dsputil_init_bfin(DSPContext *c, AVCodecContext *avctx)
void(* clear_block)(DCTELEM *block)
void(* idct)(DCTELEM *block)
int dct_bits
Size of DCT coefficients.
static void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
h264_chroma_mc_func put_h264_chroma_pixels_tab[3]
h264 Chroma MC
void ff_dsputil_init_dwt(DSPContext *c)
void(* add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size)
float ff_scalarproduct_float_c(const float *v1, const float *v2, int len)
Return the scalar product of two vectors.
void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
void(* add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top)
void(* fdct248)(DCTELEM *block)
void ff_fdct_ifast(DCTELEM *data)
void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
qpel_mc_func avg_2tap_qpel_pixels_tab[4][16]
void ff_dsputil_init_vis(DSPContext *c, AVCodecContext *avctx)
static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby, const int size, const int h, int ref_index, int src_index, me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags)
compares a block (either a full macroblock or a partition thereof) against a proposed motion-compensa...
void(* op_pixels_func)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
void ff_jpeg_fdct_islow_10(DCTELEM *data)
op_pixels_func avg_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
int idct_permutation_type
void(* vector_clipf)(float *dst, const float *src, float min, float max, int len)
static int get_penalty_factor(int lambda, int lambda2, int type)
main external API structure.
static void(WINAPI *cond_broadcast)(pthread_cond_t *cond)
void(* butterflies_float)(float *restrict v1, float *restrict v2, int len)
Calculate the sum and difference of two vectors of floats.
me_cmp_func dct264_sad[6]
void(* diff_pixels)(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride)
op_fill_func fill_block_tab[2]
void(* gmc1)(uint8_t *dst, uint8_t *src, int srcStride, int h, int x16, int y16, int rounder)
translational global motion compensation.
qpel_mc_func put_2tap_qpel_pixels_tab[4][16]
void(* add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale)
static void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
int(* pix_norm1)(uint8_t *pix, int line_size)
int(* ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2, int size)
void(* h263_v_loop_filter)(uint8_t *src, int stride, int qscale)
struct ScanTable ScanTable
Scantable.
void ff_fdct248_islow_8(DCTELEM *data)
void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
void(* vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len)
void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable)
uint8_t ff_cropTbl[256+2 *MAX_NEG_CROP]
static uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
static const uint16_t scale[4]
qpel_mc_func put_h264_qpel_pixels_tab[4][16]
void(* put_no_rnd_pixels_l2[2])(uint8_t *block, const uint8_t *a, const uint8_t *b, int line_size, int h)
int(* sum_abs_dctelem)(DCTELEM *block)
void(* clear_blocks)(DCTELEM *blocks)
void(* h261_loop_filter)(uint8_t *src, int stride)
tpel_mc_func put_tpel_pixels_tab[11]
Thirdpel motion compensation with rounding (a+b+1)>>1.
void(* vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min, int32_t max, unsigned int len)
Clip each element in an array of int32_t to a given minimum and maximum value.
static void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
void(* diff_bytes)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w)
qpel_mc_func avg_h264_qpel_pixels_tab[4][16]
void(* put_signed_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
const uint8_t ff_zigzag_direct[64]
void ff_jpeg_fdct_islow_8(DCTELEM *data)
void(* fdct)(DCTELEM *block)
void(* gmc)(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
global motion compensation.
void ff_dsputil_init_sh4(DSPContext *c, AVCodecContext *avctx)
static int16_t basis[64][64]
void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
int32_t(* scalarproduct_and_madd_int16)(int16_t *v1, const int16_t *v2, const int16_t *v3, int len, int mul)
Calculate scalar product of v1 and v2, and v1[i] += v3[i] * mul.
void(* tpel_mc_func)(uint8_t *block, const uint8_t *pixels, int line_size, int w, int h)
float(* scalarproduct_float)(const float *v1, const float *v2, int len)
Calculate the scalar product of two vectors of floats.
me_cmp_func hadamard8_diff[6]
void(* butterflies_float_interleave)(float *dst, const float *src0, const float *src1, int len)
Calculate the sum and difference of two vectors of floats and interleave results into a separate outp...
void ff_fdct_mmxext(DCTELEM *block)
void(* add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size)
me_cmp_func quant_psnr[6]
static void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
void ff_fdct_ifast248(DCTELEM *data)
qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]
struct DSPContext DSPContext
DSPContext.
me_cmp_func me_sub_cmp[6]
void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
static const uint16_t rounder[4]
void ff_init_scantable_permutation(uint8_t *idct_permutation, int idct_permutation_type)
static void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
void ff_wmv2_idct_c(DCTELEM *data)
me_cmp_func pix_abs[2][4]
void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
void(* op_fill_func)(uint8_t *block, uint8_t value, int line_size, int h)