X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdsputil.h;h=aa026e15f5a65aff4ffb118490c0746c812eb9ff;hb=71ea26811cbd8345cb10ab29406594e1fc19204e;hp=9b4aef758ac5ab69b6e29ed1ec93e26495741788;hpb=e39e3abad4ce667f06d41097428695a7aad095fb;p=ffmpeg diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h index 9b4aef758ac..aa026e15f5a 100644 --- a/libavcodec/dsputil.h +++ b/libavcodec/dsputil.h @@ -38,34 +38,41 @@ /* dct code */ typedef short DCTELEM; -void fdct_ifast (DCTELEM *data); -void fdct_ifast248 (DCTELEM *data); -void ff_jpeg_fdct_islow (DCTELEM *data); -void ff_fdct248_islow (DCTELEM *data); - -void j_rev_dct (DCTELEM *data); -void j_rev_dct4 (DCTELEM *data); -void j_rev_dct2 (DCTELEM *data); -void j_rev_dct1 (DCTELEM *data); +void ff_fdct_ifast (DCTELEM *data); +void ff_fdct_ifast248 (DCTELEM *data); +void ff_jpeg_fdct_islow_8(DCTELEM *data); +void ff_jpeg_fdct_islow_10(DCTELEM *data); +void ff_fdct248_islow_8(DCTELEM *data); +void ff_fdct248_islow_10(DCTELEM *data); + +void ff_j_rev_dct (DCTELEM *data); +void ff_j_rev_dct4 (DCTELEM *data); +void ff_j_rev_dct2 (DCTELEM *data); +void ff_j_rev_dct1 (DCTELEM *data); void ff_wmv2_idct_c(DCTELEM *data); void ff_fdct_mmx(DCTELEM *block); void ff_fdct_mmx2(DCTELEM *block); void ff_fdct_sse2(DCTELEM *block); -void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride); -void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride); -void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride); -void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride); -void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block); -void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block); -void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); -void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); -void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); -void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]); - -void ff_h264_chroma_dc_dequant_idct_c(DCTELEM *block, int qmul); -void ff_h264_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qmul); +#define H264_IDCT(depth) \ +void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\ +void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\ +void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\ +void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\ +void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\ +void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\ +void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\ +void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\ +void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\ +void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\ +void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\ +void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul); + +H264_IDCT( 8) +H264_IDCT( 9) +H264_IDCT(10) + void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp); void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc); @@ -82,10 +89,20 @@ extern const uint8_t ff_zigzag248_direct[64]; extern uint32_t ff_squareTbl[512]; extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP]; -void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride); -void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride); -void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride); -void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride); +#define PUTAVG_PIXELS(depth)\ +void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\ +void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\ +void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\ +void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride); + +PUTAVG_PIXELS( 8) +PUTAVG_PIXELS( 9) +PUTAVG_PIXELS(10) + +#define ff_put_pixels8x8_c ff_put_pixels8x8_8_c +#define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c +#define ff_put_pixels16x16_c ff_put_pixels16x16_8_c +#define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c /* VP3 DSP functions */ void ff_vp3_idct_c(DCTELEM *block/* align 16*/); @@ -96,14 +113,15 @@ void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM 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); -/* 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); - /* EA functions */ void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block); +/* RV40 functions */ +void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride); +void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride); +void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride); +void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride); + /* 1/2^n downscaling functions from imgconvert.c */ void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height); @@ -135,7 +153,7 @@ void clear_blocks_c(DCTELEM *blocks); /* add and put pixel (decoding) */ // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16 -//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4 +//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h); 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); @@ -168,7 +186,7 @@ static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ } /* motion estimation */ -// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2 +// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2 // 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))*/; @@ -186,11 +204,18 @@ typedef struct ScanTable{ } ScanTable; void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable); +void ff_init_scantable_permutation(uint8_t *idct_permutation, + int idct_permutation_type); -void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize, - int block_w, int block_h, +#define EMULATED_EDGE(depth) \ +void ff_emulated_edge_mc_ ## depth (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); +EMULATED_EDGE(8) +EMULATED_EDGE(9) +EMULATED_EDGE(10) + void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize); void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize); void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize); @@ -199,12 +224,16 @@ void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int lin * DSPContext. */ typedef struct DSPContext { + /** + * Size of DCT coefficients. + */ + int dct_bits; + /* pixel ops : interface with DCT */ void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size); 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); @@ -349,7 +378,6 @@ typedef struct DSPContext { /* 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 @@ -359,8 +387,6 @@ typedef struct DSPContext { 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 (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len); @@ -399,31 +425,16 @@ typedef struct DSPContext { void (*vector_fmul_scalar)(float *dst, const float *src, float mul, int len); /** - * Multiply a vector of floats by concatenated short vectors of - * floats and by a scalar float. Source and destination vectors - * must overlap exactly or not at all. - * [0]: short vectors of length 2, 8-byte aligned - * [1]: short vectors of length 4, 16-byte aligned - * @param dst output vector, 16-byte aligned + * Multiply a vector of floats by a scalar float and add to + * destination vector. Source and destination vectors must + * overlap exactly or not at all. + * @param dst result vector, 16-byte aligned * @param src input vector, 16-byte aligned - * @param sv array of pointers to short vectors * @param mul scalar value - * @param len number of elements in src and dst, multiple of 4 - */ - void (*vector_fmul_sv_scalar[2])(float *dst, const float *src, - const float **sv, float mul, int len); - /** - * Multiply short vectors of floats by a scalar float, store - * concatenated result. - * [0]: short vectors of length 2, 8-byte aligned - * [1]: short vectors of length 4, 16-byte aligned - * @param dst output vector, 16-byte aligned - * @param sv array of pointers to short vectors - * @param mul scalar value - * @param len number of output elements, multiple of 4 + * @param len length of vector, multiple of 4 */ - void (*sv_fmul_scalar[2])(float *dst, const float **sv, - float mul, int len); + void (*vector_fmac_scalar)(float *dst, const float *src, float mul, + int len); /** * Calculate the scalar product of two vectors of floats. * @param v1 first vector, 16-byte aligned @@ -439,6 +450,23 @@ typedef struct DSPContext { */ void (*butterflies_float)(float *restrict v1, float *restrict v2, int len); + /** + * Calculate the sum and difference of two vectors of floats and interleave + * results into a separate output vector of floats, with each sum + * positioned before the corresponding difference. + * + * @param dst output vector + * constraints: 16-byte aligned + * @param src0 first input vector + * constraints: 32-byte aligned + * @param src1 second input vector + * constraints: 32-byte aligned + * @param len number of elements in the input + * constraints: multiple of 8 + */ + void (*butterflies_float_interleave)(float *dst, const float *src0, + const float *src1, int len); + /* (I)DCT */ void (*fdct)(DCTELEM *block/* align 16*/); void (*fdct248)(DCTELEM *block/* align 16*/); @@ -467,8 +495,8 @@ typedef struct DSPContext { * with the zigzag/alternate scan
* 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) + * - (x -> reference dct -> reference idct -> x) + * - (x -> reference 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]; @@ -485,7 +513,7 @@ typedef struct DSPContext { #define BASIS_SHIFT 16 #define RECON_SHIFT 6 - void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides); + void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides); #define EDGE_WIDTH 16 #define EDGE_TOP 1 #define EDGE_BOTTOM 2 @@ -508,9 +536,8 @@ typedef struct DSPContext { /** * 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)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift); + int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len); /* ape functions */ /** * Calculate scalar product of v1 and v2, @@ -533,23 +560,27 @@ typedef struct DSPContext { void (*apply_window_int16)(int16_t *output, const int16_t *input, const int16_t *window, unsigned int len); - /* 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]; + /** + * Clip each element in an array of int32_t to a given minimum and maximum value. + * @param dst destination array + * constraints: 16-byte aligned + * @param src source array + * constraints: 16-byte aligned + * @param min minimum value + * constraints: must in the the range [-(1<<24), 1<<24] + * @param max maximum value + * constraints: must in the the range [-(1<<24), 1<<24] + * @param len number of elements in the array + * constraints: multiple of 32 greater than zero + */ + void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); - /* 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(DSPContext* p, AVCodecContext *avctx); +void ff_dsputil_static_init(void); +void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx); int ff_check_alignment(void); @@ -562,6 +593,7 @@ void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scant void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type); #define BYTE_VEC32(c) ((c)*0x01010101UL) +#define BYTE_VEC64(c) ((c)*0x0001000100010001UL) static inline uint32_t rnd_avg32(uint32_t a, uint32_t b) { @@ -573,6 +605,16 @@ static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b) return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1); } +static inline uint64_t rnd_avg64(uint64_t a, uint64_t b) +{ + return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1); +} + +static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b) +{ + return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1); +} + static inline int get_penalty_factor(int lambda, int lambda2, int type){ switch(type&0xFF){ default: @@ -597,58 +639,23 @@ static inline int get_penalty_factor(int lambda, int lambda2, int type){ } } -/** - * 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() - -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); +void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx); +void ff_dsputil_init_vis(DSPContext* c, AVCodecContext *avctx); void ff_dsputil_init_dwt(DSPContext *c); -void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx); -void ff_rv40dsp_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 - -#undef emms_c - -static inline void emms(void) -{ - __asm__ volatile ("emms;":::"memory"); -} - -#define emms_c() emms() - -#elif ARCH_ARM - -#if HAVE_NEON +#if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMI || HAVE_MMX # define STRIDE_ALIGN 16 -#endif - -#elif ARCH_PPC - -#define STRIDE_ALIGN 16 - -#elif HAVE_MMI - -#define STRIDE_ALIGN 16 - -#endif - -#ifndef STRIDE_ALIGN +#else # define STRIDE_ALIGN 8 #endif @@ -672,11 +679,6 @@ static inline void emms(void) # define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__) #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 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)\