*/
/**
- * @file libavcodec/dsputil.h
+ * @file
* DSP utils.
* note, many functions in here may use MMX which trashes the FPU state, it is
* absolutely necessary to call emms_c() between dsp & float/double code
//#define DEBUG
/* dct code */
typedef short DCTELEM;
-typedef int DWTELEM;
-typedef short IDWTELEM;
void fdct_ifast (DCTELEM *data);
void fdct_ifast248 (DCTELEM *data);
void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
+void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
const int16_t *h_weights, const int16_t *v_weights);
+/* Bink functions */
+void ff_bink_idct_c (DCTELEM *block);
+void ff_bink_idct_add_c(uint8_t *dest, int linesize, DCTELEM *block);
+void ff_bink_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
+
+/* CAVS functions */
+void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
+
+/* VC1 functions */
+void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
+void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
+
+/* EA functions */
+void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
+
/* 1/2^n downscaling functions from imgconvert.c */
void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
-typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
-typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
+
+typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
#define DEF_OLD_QPEL(name)\
void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
// although currently h<4 is not used as functions with width <8 are neither used nor implemented
typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
-
-// for snow slices
-typedef struct slice_buffer_s slice_buffer;
-
/**
* Scantable.
*/
uint8_t raster_end[64];
#if ARCH_PPC
/** Used by dct_quantize_altivec to find last-non-zero */
- DECLARE_ALIGNED(16, uint8_t, inverse[64]);
+ DECLARE_ALIGNED(16, uint8_t, inverse)[64];
#endif
} ScanTable;
void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
-void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize,
+void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize,
int block_w, int block_h,
int src_x, int src_y, int w, int h);
void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
+ void (*put_pixels_nonclamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
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];
* 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);
- int (*add_hfyu_left_prediction)(uint8_t *dst, uint8_t *src, int w, int acc);
- void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, uint8_t *src, int w, int *red, int *green, int *blue);
+ void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
+ void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
+ int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
+ void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
/* 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 (*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_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
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 (*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);
+ void (*lpc_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);
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);
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.
+ * 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
*/
- void (*sub_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
+ int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift);
+ /* ape functions */
/**
- * Calculate scalar product of two vectors.
+ * Calculate scalar product of v1 and v2,
+ * and v1[i] += v3[i] * mul
* @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);
+ int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
/* rv30 functions */
qpel_mc_func put_rv30_tpel_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];
+
+ /* bink functions */
+ op_fill_func fill_block_tab[2];
+ void (*scale_block)(const uint8_t src[64]/*align 8*/, uint8_t *dst/*align 8*/, int linesize);
} DSPContext;
void dsputil_static_init(void);
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)
-#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
+void ff_dsputil_init_dwt(DSPContext *c);
+void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
+void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
#if HAVE_MMX
# 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);
+#define LOCAL_ALIGNED(a, t, v, s, ...) \
+ uint8_t la_##v[sizeof(t s __VA_ARGS__) + (a)]; \
+ t (*v) __VA_ARGS__ = (void *)FFALIGN((uintptr_t)la_##v, a)
-/* FFT computation */
-
-/* NOTE: soon integer code will be added, so you must use the
- FFTSample type */
-typedef float FFTSample;
-
-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 */
- FFTComplex *tmp_buf;
- int mdct_size; /* size of MDCT (i.e. number of input data * 2) */
- int mdct_bits; /* n = 2^nbits */
- /* pre/post rotation tables */
- FFTSample *tcos;
- FFTSample *tsin;
- void (*fft_permute)(struct FFTContext *s, FFTComplex *z);
- void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
- void (*imdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
- void (*imdct_half)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
- void (*mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
- int split_radix;
- int permutation;
-#define FF_MDCT_PERM_NONE 0
-#define FF_MDCT_PERM_INTERLEAVE 1
-} FFTContext;
-
-#if CONFIG_HARDCODED_TABLES
-#define COSTABLE_CONST const
+#if HAVE_LOCAL_ALIGNED_8
+# define LOCAL_ALIGNED_8(t, v, s, ...) DECLARE_ALIGNED(8, t, v) s __VA_ARGS__
#else
-#define COSTABLE_CONST
+# define LOCAL_ALIGNED_8(t, v, s, ...) LOCAL_ALIGNED(8, t, v, s, __VA_ARGS__)
#endif
-#define COSTABLE(size) \
- COSTABLE_CONST DECLARE_ALIGNED_16(FFTSample, ff_cos_##size[size/2])
-#define SINTABLE(size) \
- DECLARE_ALIGNED_16(FFTSample, ff_sin_##size[size/2])
-extern COSTABLE(16);
-extern COSTABLE(32);
-extern COSTABLE(64);
-extern COSTABLE(128);
-extern COSTABLE(256);
-extern COSTABLE(512);
-extern COSTABLE(1024);
-extern COSTABLE(2048);
-extern COSTABLE(4096);
-extern COSTABLE(8192);
-extern COSTABLE(16384);
-extern COSTABLE(32768);
-extern COSTABLE(65536);
-extern COSTABLE_CONST FFTSample* const ff_cos_tabs[13];
-
-extern SINTABLE(16);
-extern SINTABLE(32);
-extern SINTABLE(64);
-extern SINTABLE(128);
-extern SINTABLE(256);
-extern SINTABLE(512);
-extern SINTABLE(1024);
-extern SINTABLE(2048);
-extern SINTABLE(4096);
-extern SINTABLE(8192);
-extern SINTABLE(16384);
-extern SINTABLE(32768);
-extern SINTABLE(65536);
-
-/**
- * Sets up a complex FFT.
- * @param nbits log2 of the length of the input array
- * @param inverse if 0 perform the forward transform, if 1 perform the inverse
- */
-int ff_fft_init(FFTContext *s, int nbits, int inverse);
-void ff_fft_permute_c(FFTContext *s, FFTComplex *z);
-void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
-
-void ff_fft_init_altivec(FFTContext *s);
-void ff_fft_init_mmx(FFTContext *s);
-void ff_fft_init_arm(FFTContext *s);
-
-/**
- * Do the permutation needed BEFORE calling ff_fft_calc().
- */
-static inline void ff_fft_permute(FFTContext *s, FFTComplex *z)
-{
- s->fft_permute(s, z);
-}
-/**
- * Do a complex FFT with the parameters defined in ff_fft_init(). The
- * input data must be permuted before. No 1.0/sqrt(n) normalization is done.
- */
-static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)
-{
- s->fft_calc(s, z);
-}
-void ff_fft_end(FFTContext *s);
-
-/* MDCT computation */
-
-static inline void ff_imdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input)
-{
- s->imdct_calc(s, output, input);
-}
-static inline void ff_imdct_half(FFTContext *s, FFTSample *output, const FFTSample *input)
-{
- s->imdct_half(s, output, input);
-}
-
-static inline void ff_mdct_calc(FFTContext *s, FFTSample *output,
- const FFTSample *input)
-{
- s->mdct_calc(s, output, input);
-}
-
-/**
- * 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);
-
-/**
- * Generate a sine window.
- * @param window pointer to half window
- * @param n size of half window
- */
-void ff_sine_window_init(float *window, int n);
-extern float ff_sine_32 [ 32];
-extern float ff_sine_64 [ 64];
-extern float ff_sine_128 [ 128];
-extern float ff_sine_256 [ 256];
-extern float ff_sine_512 [ 512];
-extern float ff_sine_1024[1024];
-extern float ff_sine_2048[2048];
-extern float ff_sine_4096[4096];
-extern float * const ff_sine_windows[13];
-
-int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale);
-void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
-void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input);
-void ff_mdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
-void ff_mdct_end(FFTContext *s);
-
-/* Real Discrete Fourier Transform */
-
-enum RDFTransformType {
- RDFT,
- IRDFT,
- RIDFT,
- IRIDFT,
-};
-
-typedef struct {
- int nbits;
- int inverse;
- int sign_convention;
-
- /* pre/post rotation tables */
- const FFTSample *tcos;
- FFTSample *tsin;
- FFTContext fft;
-} RDFTContext;
+#if HAVE_LOCAL_ALIGNED_16
+# define LOCAL_ALIGNED_16(t, v, s, ...) DECLARE_ALIGNED(16, t, v) s __VA_ARGS__
+#else
+# define LOCAL_ALIGNED_16(t, v, s, ...) LOCAL_ALIGNED(16, t, v, s, __VA_ARGS__)
+#endif
-/**
- * Sets up a real FFT.
- * @param nbits log2 of the length of the input array
- * @param trans the type of transform
- */
-int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
-void ff_rdft_calc(RDFTContext *s, FFTSample *data);
-void ff_rdft_end(RDFTContext *s);
+/* PSNR */
+void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
+ int orig_linesize[3], int coded_linesize,
+ AVCodecContext *avctx);
#define WRAPPER8_16(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\