X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdsputil.h;h=350ce1c44a0843f65768bfbf828a571dc756bcfb;hb=2db953f84671997e936f91140ffb5143c1537844;hp=a4a9f87677294ade19572bb730ed510f7a6637ac;hpb=65d5d5865845f057cc6530a8d0f34db952d9009c;p=ffmpeg diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h index a4a9f876772..350ce1c44a0 100644 --- a/libavcodec/dsputil.h +++ b/libavcodec/dsputil.h @@ -32,65 +32,8 @@ #include "avcodec.h" -/* encoding scans */ -extern const uint8_t ff_alternate_horizontal_scan[64]; -extern const uint8_t ff_alternate_vertical_scan[64]; - extern uint32_t ff_square_tab[512]; -void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride); -void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride); -void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride); -void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride); - -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); - -/* minimum alignment rules ;) - * 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 libav-devel mailing list, ... - * - * !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 - */ - -/* add and put pixel (decoding) - * Block sizes 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 than 4. */ -typedef void (*qpel_mc_func)(uint8_t *dst /* align width (8 or 16) */, - uint8_t *src /* align 1 */, ptrdiff_t stride); - -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 */, ptrdiff_t stride); \ - void ff_put_no_rnd_ ## name(uint8_t *dst /* align width (8 or 16) */, \ - uint8_t *src /* align 1 */, ptrdiff_t stride); \ - void ff_avg_ ## name(uint8_t *dst /* align width (8 or 16) */, \ - uint8_t *src /* align 1 */, ptrdiff_t stride); - -DEF_OLD_QPEL(qpel16_mc11_old_c) -DEF_OLD_QPEL(qpel16_mc31_old_c) -DEF_OLD_QPEL(qpel16_mc12_old_c) -DEF_OLD_QPEL(qpel16_mc32_old_c) -DEF_OLD_QPEL(qpel16_mc13_old_c) -DEF_OLD_QPEL(qpel16_mc33_old_c) -DEF_OLD_QPEL(qpel8_mc11_old_c) -DEF_OLD_QPEL(qpel8_mc31_old_c) -DEF_OLD_QPEL(qpel8_mc12_old_c) -DEF_OLD_QPEL(qpel8_mc32_old_c) -DEF_OLD_QPEL(qpel8_mc13_old_c) -DEF_OLD_QPEL(qpel8_mc33_old_c) - struct MpegEncContext; /* Motion estimation: * h is limited to { width / 2, width, 2 * width }, @@ -101,59 +44,11 @@ typedef int (*me_cmp_func)(struct MpegEncContext *c, uint8_t *blk1 /* align width (8 or 16) */, uint8_t *blk2 /* align 1 */, int line_size, int h); -/** - * Scantable. - */ -typedef struct ScanTable { - const uint8_t *scantable; - uint8_t permutated[64]; - uint8_t raster_end[64]; -} ScanTable; - -void ff_init_scantable(uint8_t *permutation, ScanTable *st, - const uint8_t *src_scantable); -void ff_init_scantable_permutation(uint8_t *idct_permutation, - int idct_permutation_type); - /** * DSPContext. */ typedef struct DSPContext { - /* pixel ops : interface with DCT */ - void (*get_pixels)(int16_t *block /* align 16 */, - const uint8_t *pixels /* align 8 */, - int line_size); - void (*diff_pixels)(int16_t *block /* align 16 */, - const uint8_t *s1 /* align 8 */, - const uint8_t *s2 /* align 8 */, - int stride); - void (*put_pixels_clamped)(const int16_t *block /* align 16 */, - uint8_t *pixels /* align 8 */, - int line_size); - void (*put_signed_pixels_clamped)(const int16_t *block /* align 16 */, - uint8_t *pixels /* align 8 */, - int line_size); - void (*add_pixels_clamped)(const int16_t *block /* align 16 */, - uint8_t *pixels /* align 8 */, - int line_size); - void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size); int (*sum_abs_dctelem)(int16_t *block /* align 16 */); - /** - * translational global motion compensation. - */ - void (*gmc1)(uint8_t *dst /* align 8 */, uint8_t *src /* align 1 */, - int srcStride, int h, int x16, int y16, int rounder); - /** - * global motion compensation. - */ - 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)(int16_t *block /* align 16 */); - void (*clear_blocks)(int16_t *blocks /* align 16 */); - int (*pix_sum)(uint8_t *pix, int line_size); - int (*pix_norm1)(uint8_t *pix, int line_size); me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */ me_cmp_func sse[6]; @@ -175,116 +70,7 @@ typedef struct DSPContext { me_cmp_func ildct_cmp[6]; // only width 16 used me_cmp_func frame_skip_cmp[6]; // only width 8 used - qpel_mc_func put_qpel_pixels_tab[2][16]; - qpel_mc_func avg_qpel_pixels_tab[2][16]; - qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]; - qpel_mc_func put_mspel_pixels_tab[8]; - me_cmp_func pix_abs[2][4]; - - void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w); - void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len); - - /* assume len is a multiple of 8, and arrays are 16-byte aligned */ - void (*vector_clipf)(float *dst /* align 16 */, - const float *src /* align 16 */, - float min, float max, int len /* align 16 */); - - /* (I)DCT */ - void (*fdct)(int16_t *block /* align 16 */); - void (*fdct248)(int16_t *block /* align 16 */); - - /* IDCT really */ - void (*idct)(int16_t *block /* align 16 */); - - /** - * block -> idct -> clip to unsigned 8 bit -> dest. - * (-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, int16_t *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, int16_t *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
- * An example to avoid confusion: - * - (->decode coeffs -> zigzag reorder -> dequant -> reference IDCT -> ...) - * - (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]; - 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, int h, int sides); -#define EDGE_WIDTH 16 -#define EDGE_TOP 1 -#define EDGE_BOTTOM 2 - - void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, - int src_wrap, int width, int height); - - /** - * Calculate scalar product of two vectors. - * @param len length of vectors, should be multiple of 16 - */ - 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, - * and v1[i] += v3[i] * mul - * @param len length of vectors, should be multiple of 16 - */ - int32_t (*scalarproduct_and_madd_int16)(int16_t *v1 /* align 16 */, - const int16_t *v2, - const int16_t *v3, - int len, int mul); - - /** - * 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 be in the range [-(1 << 24), 1 << 24] - * @param max maximum value - * constraints: must be in 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); - - op_fill_func fill_block_tab[2]; } DSPContext; void ff_dsputil_static_init(void); @@ -292,13 +78,8 @@ void ff_dsputil_init(DSPContext *p, AVCodecContext *avctx); void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type); -void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx, - unsigned high_bit_depth); -void ff_dsputil_init_bfin(DSPContext *c, AVCodecContext *avctx, - unsigned high_bit_depth); -void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx, - unsigned high_bit_depth); -void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx, - unsigned high_bit_depth); +void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx); +void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx); +void ff_dsputil_init_x86(DSPContext *c, AVCodecContext *avctx); #endif /* AVCODEC_DSPUTIL_H */