X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fvp56.h;h=0607e0d4ce6f7dbdeb42acef8f778f4651aee2fa;hb=55b9ef18e4a139fc24a3b695cb3c176f3ced09b8;hp=4eb414bbbe760eaa356033f3d6b80eed7768efc4;hpb=3b636f21daa6996e20ef97131a1c3649c3043304;p=ffmpeg diff --git a/libavcodec/vp56.h b/libavcodec/vp56.h index 4eb414bbbe7..0607e0d4ce6 100644 --- a/libavcodec/vp56.h +++ b/libavcodec/vp56.h @@ -1,26 +1,28 @@ -/** - * @file - * VP5 and VP6 compatible video decoder (common features) - * +/* * Copyright (C) 2006 Aurelien Jacobs * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg is free software; you can redistribute it and/or + * Libav is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * FFmpeg is distributed in the hope that it will be useful, + * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with FFmpeg; if not, write to the Free Software + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ +/** + * @file + * VP5 and VP6 compatible video decoder (common features) + */ + #ifndef AVCODEC_VP56_H #define AVCODEC_VP56_H @@ -31,7 +33,11 @@ #include "vp56dsp.h" typedef struct vp56_context VP56Context; -typedef struct vp56_mv VP56mv; + +typedef struct { + int16_t x; + int16_t y; +} DECLARE_ALIGNED(4, , VP56mv); typedef void (*VP56ParseVectorAdjustment)(VP56Context *s, VP56mv *vect); @@ -41,16 +47,17 @@ typedef void (*VP56Filter)(VP56Context *s, uint8_t *dst, uint8_t *src, typedef void (*VP56ParseCoeff)(VP56Context *s); typedef void (*VP56DefaultModelsInit)(VP56Context *s); typedef void (*VP56ParseVectorModels)(VP56Context *s); -typedef void (*VP56ParseCoeffModels)(VP56Context *s); +typedef int (*VP56ParseCoeffModels)(VP56Context *s); typedef int (*VP56ParseHeader)(VP56Context *s, const uint8_t *buf, int buf_size, int *golden_frame); typedef struct { int high; - int bits; + int bits; /* stored negated (i.e. negative "bits" is a positive number of + bits left) in order to eliminate a negate in cache refilling */ const uint8_t *buffer; const uint8_t *end; - unsigned long code_word; + unsigned int code_word; } VP56RangeCoder; typedef struct { @@ -59,11 +66,6 @@ typedef struct { DCTELEM dc_coeff; } VP56RefDc; -struct vp56_mv { - int x; - int y; -}; - typedef struct { uint8_t type; VP56mv mv; @@ -169,81 +171,104 @@ struct vp56_context { }; -void vp56_init(AVCodecContext *avctx, int flip, int has_alpha); -int vp56_free(AVCodecContext *avctx); -void vp56_init_dequant(VP56Context *s, int quantizer); -int vp56_decode_frame(AVCodecContext *avctx, void *data, int *data_size, - AVPacket *avpkt); +void ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha); +int ff_vp56_free(AVCodecContext *avctx); +void ff_vp56_init_dequant(VP56Context *s, int quantizer); +int ff_vp56_decode_frame(AVCodecContext *avctx, void *data, int *data_size, + AVPacket *avpkt); /** * vp56 specific range coder implementation */ -static inline void vp56_init_range_decoder(VP56RangeCoder *c, - const uint8_t *buf, int buf_size) +extern const uint8_t ff_vp56_norm_shift[256]; +void ff_vp56_init_range_decoder(VP56RangeCoder *c, const uint8_t *buf, int buf_size); + +static av_always_inline unsigned int vp56_rac_renorm(VP56RangeCoder *c) { - c->high = 255; - c->bits = 8; - c->buffer = buf; - c->end = buf + buf_size; - c->code_word = bytestream_get_be16(&c->buffer); + int shift = ff_vp56_norm_shift[c->high]; + int bits = c->bits; + unsigned int code_word = c->code_word; + + c->high <<= shift; + code_word <<= shift; + bits += shift; + if(bits >= 0 && c->buffer < c->end) { + code_word |= bytestream_get_be16(&c->buffer) << bits; + bits -= 16; + } + c->bits = bits; + return code_word; } -static inline int vp56_rac_get_prob(VP56RangeCoder *c, uint8_t prob) +#if ARCH_ARM +#include "arm/vp56_arith.h" +#elif ARCH_X86 +#include "x86/vp56_arith.h" +#endif + +#ifndef vp56_rac_get_prob +#define vp56_rac_get_prob vp56_rac_get_prob +static av_always_inline int vp56_rac_get_prob(VP56RangeCoder *c, uint8_t prob) { + unsigned int code_word = vp56_rac_renorm(c); unsigned int low = 1 + (((c->high - 1) * prob) >> 8); - unsigned int low_shift = low << 8; - int bit = c->code_word >= low_shift; + unsigned int low_shift = low << 16; + int bit = code_word >= low_shift; - if (bit) { - c->high -= low; - c->code_word -= low_shift; - } else { - c->high = low; - } + c->high = bit ? c->high - low : low; + c->code_word = bit ? code_word - low_shift : code_word; - /* normalize */ - while (c->high < 128) { - c->high <<= 1; - c->code_word <<= 1; - if (--c->bits == 0 && c->buffer < c->end) { - c->bits = 8; - c->code_word |= *c->buffer++; - } - } return bit; } +#endif -static inline int vp56_rac_get(VP56RangeCoder *c) +#ifndef vp56_rac_get_prob_branchy +// branchy variant, to be used where there's a branch based on the bit decoded +static av_always_inline int vp56_rac_get_prob_branchy(VP56RangeCoder *c, int prob) { + unsigned long code_word = vp56_rac_renorm(c); + unsigned low = 1 + (((c->high - 1) * prob) >> 8); + unsigned low_shift = low << 16; + + if (code_word >= low_shift) { + c->high -= low; + c->code_word = code_word - low_shift; + return 1; + } + + c->high = low; + c->code_word = code_word; + return 0; +} +#endif + +static av_always_inline int vp56_rac_get(VP56RangeCoder *c) +{ + unsigned int code_word = vp56_rac_renorm(c); /* equiprobable */ int low = (c->high + 1) >> 1; - unsigned int low_shift = low << 8; - int bit = c->code_word >= low_shift; + unsigned int low_shift = low << 16; + int bit = code_word >= low_shift; if (bit) { - c->high = (c->high - low) << 1; - c->code_word -= low_shift; + c->high -= low; + code_word -= low_shift; } else { - c->high = low << 1; + c->high = low; } - /* normalize */ - c->code_word <<= 1; - if (--c->bits == 0 && c->buffer < c->end) { - c->bits = 8; - c->code_word |= *c->buffer++; - } + c->code_word = code_word; return bit; } // rounding is different than vp56_rac_get, is vp56_rac_get wrong? -static inline int vp8_rac_get(VP56RangeCoder *c) +static av_always_inline int vp8_rac_get(VP56RangeCoder *c) { return vp56_rac_get_prob(c, 128); } -static inline int vp56_rac_gets(VP56RangeCoder *c, int bits) +static av_unused int vp56_rac_gets(VP56RangeCoder *c, int bits) { int value = 0; @@ -254,7 +279,7 @@ static inline int vp56_rac_gets(VP56RangeCoder *c, int bits) return value; } -static inline int vp8_rac_get_uint(VP56RangeCoder *c, int bits) +static av_unused int vp8_rac_get_uint(VP56RangeCoder *c, int bits) { int value = 0; @@ -266,7 +291,7 @@ static inline int vp8_rac_get_uint(VP56RangeCoder *c, int bits) } // fixme: add 1 bit to all the calls to this? -static inline int vp8_rac_get_sint(VP56RangeCoder *c, int bits) +static av_unused int vp8_rac_get_sint(VP56RangeCoder *c, int bits) { int v; @@ -282,21 +307,22 @@ static inline int vp8_rac_get_sint(VP56RangeCoder *c, int bits) } // P(7) -static inline int vp56_rac_gets_nn(VP56RangeCoder *c, int bits) +static av_unused int vp56_rac_gets_nn(VP56RangeCoder *c, int bits) { int v = vp56_rac_gets(c, 7) << 1; return v + !v; } -static inline int vp8_rac_get_nn(VP56RangeCoder *c) +static av_unused int vp8_rac_get_nn(VP56RangeCoder *c) { int v = vp8_rac_get_uint(c, 7) << 1; return v + !v; } -static inline int vp56_rac_get_tree(VP56RangeCoder *c, - const VP56Tree *tree, - const uint8_t *probs) +static av_always_inline +int vp56_rac_get_tree(VP56RangeCoder *c, + const VP56Tree *tree, + const uint8_t *probs) { while (tree->val > 0) { if (vp56_rac_get_prob(c, probs[tree->prob_idx])) @@ -312,8 +338,9 @@ static inline int vp56_rac_get_tree(VP56RangeCoder *c, * on a node other than the root node, needed for coeff decode where this is * used to save a bit after a 0 token (by disallowing EOB to immediately follow.) */ -static inline int vp8_rac_get_tree_with_offset(VP56RangeCoder *c, const int8_t (*tree)[2], - const uint8_t *probs, int i) +static av_always_inline +int vp8_rac_get_tree_with_offset(VP56RangeCoder *c, const int8_t (*tree)[2], + const uint8_t *probs, int i) { do { i = tree[i][vp56_rac_get_prob(c, probs[i])]; @@ -324,14 +351,15 @@ static inline int vp8_rac_get_tree_with_offset(VP56RangeCoder *c, const int8_t ( // how probabilities are associated with decisions is different I think // well, the new scheme fits in the old but this way has one fewer branches per decision -static inline int vp8_rac_get_tree(VP56RangeCoder *c, const int8_t (*tree)[2], - const uint8_t *probs) +static av_always_inline +int vp8_rac_get_tree(VP56RangeCoder *c, const int8_t (*tree)[2], + const uint8_t *probs) { return vp8_rac_get_tree_with_offset(c, tree, probs, 0); } // DCTextra -static inline int vp8_rac_get_coeff(VP56RangeCoder *c, const uint8_t *prob) +static av_always_inline int vp8_rac_get_coeff(VP56RangeCoder *c, const uint8_t *prob) { int v = 0;