/*
* DSP utils
- * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
+ * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*/
/**
- * @file dsputil.h
+ * @file libavcodec/dsputil.h
* 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
*/
-#ifndef FFMPEG_DSPUTIL_H
-#define FFMPEG_DSPUTIL_H
+#ifndef AVCODEC_DSPUTIL_H
+#define AVCODEC_DSPUTIL_H
+#include "libavutil/intreadwrite.h"
#include "avcodec.h"
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_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
- const float *src2, int src3, int blocksize, int step);
-void ff_float_to_int16_c(int16_t *dst, const float *src, int len);
+void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
+ const float *win, float add_bias, int len);
+void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
+void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels);
/* encoding scans */
extern const uint8_t ff_alternate_horizontal_scan[64];
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_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);
+
+/* VP6 DSP functions */
+void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
+ const int16_t *h_weights, const int16_t *v_weights);
+
/* 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);
int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
/* minimum alignment rules ;)
-if u notice errors in the align stuff, need more alignment for some asm code for some cpu
-or need to use a function with less aligned data then send a mail to the ffmpeg-dev list, ...
+If you notice errors in the align stuff, need more alignment for some ASM code
+for some CPU or need to use a function with less aligned data then send a mail
+to the ffmpeg-devel mailing list, ...
-!warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
-i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
+!warning These alignments might not match reality, (missing attribute((align))
+stuff somewhere possible).
+I (Michael) did not check them, these are just the alignments which I think
+could be reached easily ...
!future video codecs might need functions with less strict alignment
*/
// for snow slices
typedef struct slice_buffer_s slice_buffer;
+/**
+ * Scantable.
+ */
+typedef struct ScanTable{
+ const uint8_t *scantable;
+ uint8_t permutated[64];
+ 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]);
+#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,
+ int block_w, int block_h,
+ int src_x, int src_y, int w, int h);
+
/**
* DSPContext.
*/
*/
void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
+ void (*clear_block)(DCTELEM *block/*align 16*/);
void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
int (*pix_sum)(uint8_t * pix, int line_size);
int (*pix_norm1)(uint8_t * pix, int line_size);
// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
- me_cmp_func sad[5]; /* identical to pix_absAxA except additional void * */
- me_cmp_func sse[5];
- me_cmp_func hadamard8_diff[5];
- me_cmp_func dct_sad[5];
- me_cmp_func quant_psnr[5];
- me_cmp_func bit[5];
- me_cmp_func rd[5];
- me_cmp_func vsad[5];
- me_cmp_func vsse[5];
- me_cmp_func nsse[5];
- me_cmp_func w53[5];
- me_cmp_func w97[5];
- me_cmp_func dct_max[5];
- me_cmp_func dct264_sad[5];
-
- me_cmp_func me_pre_cmp[5];
- me_cmp_func me_cmp[5];
- me_cmp_func me_sub_cmp[5];
- me_cmp_func mb_cmp[5];
- me_cmp_func ildct_cmp[5]; //only width 16 used
- me_cmp_func frame_skip_cmp[5]; //only width 8 used
+ me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
+ me_cmp_func sse[6];
+ me_cmp_func hadamard8_diff[6];
+ me_cmp_func dct_sad[6];
+ me_cmp_func quant_psnr[6];
+ me_cmp_func bit[6];
+ me_cmp_func rd[6];
+ me_cmp_func vsad[6];
+ me_cmp_func vsse[6];
+ me_cmp_func nsse[6];
+ me_cmp_func w53[6];
+ me_cmp_func w97[6];
+ me_cmp_func dct_max[6];
+ me_cmp_func dct264_sad[6];
+
+ me_cmp_func me_pre_cmp[6];
+ me_cmp_func me_cmp[6];
+ me_cmp_func me_sub_cmp[6];
+ me_cmp_func mb_cmp[6];
+ me_cmp_func ildct_cmp[6]; //only width 16 used
+ me_cmp_func frame_skip_cmp[6]; //only width 8 used
int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
int size);
* 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];
+ /* This is really one func used in VC-1 decoding */
+ h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3];
+ h264_chroma_mc_func avg_no_rnd_vc1_chroma_pixels_tab[3];
qpel_mc_func put_h264_qpel_pixels_tab[4][16];
qpel_mc_func avg_h264_qpel_pixels_tab[4][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 (*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);
/* 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, int stride, int alpha, int beta, int8_t *tc0);
- void (*h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
- void (*h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
- void (*h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0);
- void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
- void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta);
+ 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 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_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);
+ 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);
/* 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);
+ void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
+ /* 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);
+ void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
+ /**
+ * Multiply a vector of floats by a scalar float. 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 mul scalar value
+ * @param len length of vector, multiple of 4
+ */
+ 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
+ * @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
+ */
+ void (*sv_fmul_scalar[2])(float *dst, const float **sv,
+ float mul, int len);
+ /**
+ * Calculate the scalar product of two vectors of floats.
+ * @param v1 first vector, 16-byte aligned
+ * @param v2 second vector, 16-byte aligned
+ * @param len length of vectors, multiple of 4
+ */
+ float (*scalarproduct_float)(const float *v1, const float *v2, int len);
+ /**
+ * Calculate the sum and difference of two vectors of floats.
+ * @param v1 first input vector, sum output, 16-byte aligned
+ * @param v2 second input vector, difference output, 16-byte aligned
+ * @param len length of vectors, multiple of 4
+ */
+ void (*butterflies_float)(float *restrict v1, float *restrict v2, 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, int len);
+ 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*/);
#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);
+ /* 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 (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
+ /* mlp/truehd functions */
+ void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
+ int firorder, int iirorder,
+ unsigned int filter_shift, int32_t mask, int blocksize,
+ int32_t *sample_buffer);
+
/* 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_inv_trans_8x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
+ void (*vc1_inv_trans_8x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
+ void (*vc1_inv_trans_4x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
+ void (*vc1_inv_trans_4x4_dc)(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);
+ void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq);
+ void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq);
+ void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq);
+ void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq);
+ void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq);
+ void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
/* 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];
+ op_pixels_func avg_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);
/* should be defined by architectures supporting
one or more MultiMedia extension */
int mm_support(void);
+extern int mm_flags;
void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
-void dsputil_init_armv4l(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_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)
-#if defined(HAVE_MMX)
+#if 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 inline void emms(void)
{
- __asm __volatile ("emms;":::"memory");
+ __asm__ volatile ("emms;":::"memory");
}
#define emms_c() \
{\
- if (mm_flags & MM_MMX)\
+ if (mm_flags & FF_MM_MMX)\
emms();\
}
-void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
-
-#elif defined(ARCH_ARMV4L)
+#elif ARCH_ARM
-#define MM_IWMMXT 0x0100 /* XScale IWMMXT */
-
-extern int mm_flags;
-
-#elif defined(ARCH_POWERPC)
-
-#define MM_ALTIVEC 0x0001 /* standard AltiVec */
+#if HAVE_NEON
+# define STRIDE_ALIGN 16
+#endif
-extern int mm_flags;
+#elif ARCH_PPC
-#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
#define STRIDE_ALIGN 16
-#elif defined(HAVE_MMI)
+#elif HAVE_MMI
-#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
#define STRIDE_ALIGN 16
-#endif
+#else
+
+#define mm_flags 0
+#define mm_support() 0
-#ifndef DECLARE_ALIGNED_8
-# define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
#endif
#ifndef STRIDE_ALIGN
FFTSample type */
typedef float FFTSample;
-struct MDCTContext;
-
typedef struct FFTComplex {
FFTSample re, im;
} FFTComplex;
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 MDCTContext *s, FFTSample *output,
- const FFTSample *input, FFTSample *tmp);
+ 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
+#define SINTABLE_CONST const
+#else
+#define COSTABLE_CONST
+#define SINTABLE_CONST
+#endif
+
+#define COSTABLE(size) \
+ COSTABLE_CONST DECLARE_ALIGNED_16(FFTSample, ff_cos_##size[size/2])
+#define SINTABLE(size) \
+ SINTABLE_CONST 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[17];
+
+/**
+ * Initializes the cosine table in ff_cos_tabs[index]
+ * \param index index in ff_cos_tabs array of the table to initialize
+ */
+void ff_init_ff_cos_tabs(int index);
+
+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(FFTContext *s, FFTComplex *z);
+void ff_fft_permute_c(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);
+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);
/* 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;
+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.
*/
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);
+/**
+ * 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;
+ SINTABLE_CONST FFTSample *tsin;
+ FFTContext fft;
+} RDFTContext;
+
+/**
+ * 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);
#define WRAPPER8_16(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
}
-static inline void copy_block2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-static inline void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-static inline void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-static inline void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
+static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
int i;
for(i=0; i<h; i++)
}
}
-#endif /* FFMPEG_DSPUTIL_H */
+#endif /* AVCODEC_DSPUTIL_H */
projects / ffmpeg / blobdiff