You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

4391 lines
177 KiB

/*
* VP9 compatible video decoder
*
* Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
* Copyright (C) 2013 Clément Bœsch <u pkh me>
*
* This file is part of FFmpeg.
*
* FFmpeg 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,
* 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
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "avcodec.h"
#include "get_bits.h"
#include "internal.h"
#include "profiles.h"
#include "thread.h"
#include "videodsp.h"
#include "vp56.h"
#include "vp9.h"
#include "vp9data.h"
#include "vp9dsp.h"
#include "libavutil/avassert.h"
#include "libavutil/pixdesc.h"
#define VP9_SYNCCODE 0x498342
struct VP9Filter {
uint8_t level[8 * 8];
uint8_t /* bit=col */ mask[2 /* 0=y, 1=uv */][2 /* 0=col, 1=row */]
[8 /* rows */][4 /* 0=16, 1=8, 2=4, 3=inner4 */];
};
typedef struct VP9Block {
uint8_t seg_id, intra, comp, ref[2], mode[4], uvmode, skip;
enum FilterMode filter;
VP56mv mv[4 /* b_idx */][2 /* ref */];
enum BlockSize bs;
enum TxfmMode tx, uvtx;
enum BlockLevel bl;
enum BlockPartition bp;
} VP9Block;
typedef struct VP9Context {
VP9SharedContext s;
VP9DSPContext dsp;
VideoDSPContext vdsp;
GetBitContext gb;
VP56RangeCoder c;
VP56RangeCoder *c_b;
unsigned c_b_size;
VP9Block *b_base, *b;
int pass;
int row, row7, col, col7;
uint8_t *dst[3];
ptrdiff_t y_stride, uv_stride;
uint8_t ss_h, ss_v;
uint8_t last_bpp, bpp_index, bytesperpixel;
uint8_t last_keyframe;
// sb_cols/rows, rows/cols and last_fmt are used for allocating all internal
// arrays, and are thus per-thread. w/h and gf_fmt are synced between threads
// and are therefore per-stream. pix_fmt represents the value in the header
// of the currently processed frame.
int w, h;
enum AVPixelFormat pix_fmt, last_fmt, gf_fmt;
unsigned sb_cols, sb_rows, rows, cols;
ThreadFrame next_refs[8];
struct {
uint8_t lim_lut[64];
uint8_t mblim_lut[64];
} filter_lut;
unsigned tile_row_start, tile_row_end, tile_col_start, tile_col_end;
struct {
prob_context p;
uint8_t coef[4][2][2][6][6][3];
} prob_ctx[4];
struct {
prob_context p;
uint8_t coef[4][2][2][6][6][11];
} prob;
struct {
unsigned y_mode[4][10];
unsigned uv_mode[10][10];
unsigned filter[4][3];
unsigned mv_mode[7][4];
unsigned intra[4][2];
unsigned comp[5][2];
unsigned single_ref[5][2][2];
unsigned comp_ref[5][2];
unsigned tx32p[2][4];
unsigned tx16p[2][3];
unsigned tx8p[2][2];
unsigned skip[3][2];
unsigned mv_joint[4];
struct {
unsigned sign[2];
unsigned classes[11];
unsigned class0[2];
unsigned bits[10][2];
unsigned class0_fp[2][4];
unsigned fp[4];
unsigned class0_hp[2];
unsigned hp[2];
} mv_comp[2];
unsigned partition[4][4][4];
unsigned coef[4][2][2][6][6][3];
unsigned eob[4][2][2][6][6][2];
} counts;
// contextual (left/above) cache
DECLARE_ALIGNED(16, uint8_t, left_y_nnz_ctx)[16];
DECLARE_ALIGNED(16, uint8_t, left_mode_ctx)[16];
DECLARE_ALIGNED(16, VP56mv, left_mv_ctx)[16][2];
DECLARE_ALIGNED(16, uint8_t, left_uv_nnz_ctx)[2][16];
DECLARE_ALIGNED(8, uint8_t, left_partition_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_skip_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_txfm_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_segpred_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_intra_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_comp_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_ref_ctx)[8];
DECLARE_ALIGNED(8, uint8_t, left_filter_ctx)[8];
uint8_t *above_partition_ctx;
uint8_t *above_mode_ctx;
// FIXME maybe merge some of the below in a flags field?
uint8_t *above_y_nnz_ctx;
uint8_t *above_uv_nnz_ctx[2];
uint8_t *above_skip_ctx; // 1bit
uint8_t *above_txfm_ctx; // 2bit
uint8_t *above_segpred_ctx; // 1bit
uint8_t *above_intra_ctx; // 1bit
uint8_t *above_comp_ctx; // 1bit
uint8_t *above_ref_ctx; // 2bit
uint8_t *above_filter_ctx;
VP56mv (*above_mv_ctx)[2];
// whole-frame cache
uint8_t *intra_pred_data[3];
struct VP9Filter *lflvl;
DECLARE_ALIGNED(32, uint8_t, edge_emu_buffer)[135 * 144 * 2];
// block reconstruction intermediates
int block_alloc_using_2pass;
int16_t *block_base, *block, *uvblock_base[2], *uvblock[2];
uint8_t *eob_base, *uveob_base[2], *eob, *uveob[2];
struct { int x, y; } min_mv, max_mv;
DECLARE_ALIGNED(32, uint8_t, tmp_y)[64 * 64 * 2];
DECLARE_ALIGNED(32, uint8_t, tmp_uv)[2][64 * 64 * 2];
uint16_t mvscale[3][2];
uint8_t mvstep[3][2];
} VP9Context;
static const uint8_t bwh_tab[2][N_BS_SIZES][2] = {
{
{ 16, 16 }, { 16, 8 }, { 8, 16 }, { 8, 8 }, { 8, 4 }, { 4, 8 },
{ 4, 4 }, { 4, 2 }, { 2, 4 }, { 2, 2 }, { 2, 1 }, { 1, 2 }, { 1, 1 },
}, {
{ 8, 8 }, { 8, 4 }, { 4, 8 }, { 4, 4 }, { 4, 2 }, { 2, 4 },
{ 2, 2 }, { 2, 1 }, { 1, 2 }, { 1, 1 }, { 1, 1 }, { 1, 1 }, { 1, 1 },
}
};
static void vp9_unref_frame(AVCodecContext *ctx, VP9Frame *f)
{
ff_thread_release_buffer(ctx, &f->tf);
av_buffer_unref(&f->extradata);
av_buffer_unref(&f->hwaccel_priv_buf);
f->segmentation_map = NULL;
f->hwaccel_picture_private = NULL;
}
static int vp9_alloc_frame(AVCodecContext *ctx, VP9Frame *f)
{
VP9Context *s = ctx->priv_data;
int ret, sz;
if ((ret = ff_thread_get_buffer(ctx, &f->tf, AV_GET_BUFFER_FLAG_REF)) < 0)
return ret;
sz = 64 * s->sb_cols * s->sb_rows;
if (!(f->extradata = av_buffer_allocz(sz * (1 + sizeof(struct VP9mvrefPair))))) {
goto fail;
}
f->segmentation_map = f->extradata->data;
f->mv = (struct VP9mvrefPair *) (f->extradata->data + sz);
if (ctx->hwaccel) {
const AVHWAccel *hwaccel = ctx->hwaccel;
av_assert0(!f->hwaccel_picture_private);
if (hwaccel->frame_priv_data_size) {
f->hwaccel_priv_buf = av_buffer_allocz(hwaccel->frame_priv_data_size);
if (!f->hwaccel_priv_buf)
goto fail;
f->hwaccel_picture_private = f->hwaccel_priv_buf->data;
}
}
return 0;
fail:
vp9_unref_frame(ctx, f);
return AVERROR(ENOMEM);
}
static int vp9_ref_frame(AVCodecContext *ctx, VP9Frame *dst, VP9Frame *src)
{
int res;
if ((res = ff_thread_ref_frame(&dst->tf, &src->tf)) < 0) {
return res;
} else if (!(dst->extradata = av_buffer_ref(src->extradata))) {
goto fail;
}
dst->segmentation_map = src->segmentation_map;
dst->mv = src->mv;
dst->uses_2pass = src->uses_2pass;
if (src->hwaccel_picture_private) {
dst->hwaccel_priv_buf = av_buffer_ref(src->hwaccel_priv_buf);
if (!dst->hwaccel_priv_buf)
goto fail;
dst->hwaccel_picture_private = dst->hwaccel_priv_buf->data;
}
return 0;
fail:
vp9_unref_frame(ctx, dst);
return AVERROR(ENOMEM);
}
static int update_size(AVCodecContext *ctx, int w, int h)
{
#define HWACCEL_MAX (CONFIG_VP9_DXVA2_HWACCEL + CONFIG_VP9_D3D11VA_HWACCEL + CONFIG_VP9_VAAPI_HWACCEL)
enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmtp = pix_fmts;
VP9Context *s = ctx->priv_data;
uint8_t *p;
int bytesperpixel = s->bytesperpixel, res, cols, rows;
av_assert0(w > 0 && h > 0);
if (!(s->pix_fmt == s->gf_fmt && w == s->w && h == s->h)) {
if ((res = ff_set_dimensions(ctx, w, h)) < 0)
return res;
switch (s->pix_fmt) {
case AV_PIX_FMT_YUV420P:
#if CONFIG_VP9_DXVA2_HWACCEL
*fmtp++ = AV_PIX_FMT_DXVA2_VLD;
#endif
#if CONFIG_VP9_D3D11VA_HWACCEL
*fmtp++ = AV_PIX_FMT_D3D11VA_VLD;
#endif
#if CONFIG_VP9_VAAPI_HWACCEL
*fmtp++ = AV_PIX_FMT_VAAPI;
#endif
break;
case AV_PIX_FMT_YUV420P10:
case AV_PIX_FMT_YUV420P12:
#if CONFIG_VP9_VAAPI_HWACCEL
*fmtp++ = AV_PIX_FMT_VAAPI;
#endif
break;
}
*fmtp++ = s->pix_fmt;
*fmtp = AV_PIX_FMT_NONE;
res = ff_thread_get_format(ctx, pix_fmts);
if (res < 0)
return res;
ctx->pix_fmt = res;
s->gf_fmt = s->pix_fmt;
s->w = w;
s->h = h;
}
cols = (w + 7) >> 3;
rows = (h + 7) >> 3;
if (s->intra_pred_data[0] && cols == s->cols && rows == s->rows && s->pix_fmt == s->last_fmt)
return 0;
s->last_fmt = s->pix_fmt;
s->sb_cols = (w + 63) >> 6;
s->sb_rows = (h + 63) >> 6;
s->cols = (w + 7) >> 3;
s->rows = (h + 7) >> 3;
#define assign(var, type, n) var = (type) p; p += s->sb_cols * (n) * sizeof(*var)
av_freep(&s->intra_pred_data[0]);
// FIXME we slightly over-allocate here for subsampled chroma, but a little
// bit of padding shouldn't affect performance...
p = av_malloc(s->sb_cols * (128 + 192 * bytesperpixel +
sizeof(*s->lflvl) + 16 * sizeof(*s->above_mv_ctx)));
if (!p)
return AVERROR(ENOMEM);
assign(s->intra_pred_data[0], uint8_t *, 64 * bytesperpixel);
assign(s->intra_pred_data[1], uint8_t *, 64 * bytesperpixel);
assign(s->intra_pred_data[2], uint8_t *, 64 * bytesperpixel);
assign(s->above_y_nnz_ctx, uint8_t *, 16);
assign(s->above_mode_ctx, uint8_t *, 16);
assign(s->above_mv_ctx, VP56mv(*)[2], 16);
assign(s->above_uv_nnz_ctx[0], uint8_t *, 16);
assign(s->above_uv_nnz_ctx[1], uint8_t *, 16);
assign(s->above_partition_ctx, uint8_t *, 8);
assign(s->above_skip_ctx, uint8_t *, 8);
assign(s->above_txfm_ctx, uint8_t *, 8);
assign(s->above_segpred_ctx, uint8_t *, 8);
assign(s->above_intra_ctx, uint8_t *, 8);
assign(s->above_comp_ctx, uint8_t *, 8);
assign(s->above_ref_ctx, uint8_t *, 8);
assign(s->above_filter_ctx, uint8_t *, 8);
assign(s->lflvl, struct VP9Filter *, 1);
#undef assign
// these will be re-allocated a little later
av_freep(&s->b_base);
av_freep(&s->block_base);
if (s->s.h.bpp != s->last_bpp) {
ff_vp9dsp_init(&s->dsp, s->s.h.bpp, ctx->flags & AV_CODEC_FLAG_BITEXACT);
ff_videodsp_init(&s->vdsp, s->s.h.bpp);
s->last_bpp = s->s.h.bpp;
}
return 0;
}
static int update_block_buffers(AVCodecContext *ctx)
{
VP9Context *s = ctx->priv_data;
int chroma_blocks, chroma_eobs, bytesperpixel = s->bytesperpixel;
if (s->b_base && s->block_base && s->block_alloc_using_2pass == s->s.frames[CUR_FRAME].uses_2pass)
return 0;
av_free(s->b_base);
av_free(s->block_base);
chroma_blocks = 64 * 64 >> (s->ss_h + s->ss_v);
chroma_eobs = 16 * 16 >> (s->ss_h + s->ss_v);
if (s->s.frames[CUR_FRAME].uses_2pass) {
int sbs = s->sb_cols * s->sb_rows;
s->b_base = av_malloc_array(s->cols * s->rows, sizeof(VP9Block));
s->block_base = av_mallocz(((64 * 64 + 2 * chroma_blocks) * bytesperpixel * sizeof(int16_t) +
16 * 16 + 2 * chroma_eobs) * sbs);
if (!s->b_base || !s->block_base)
return AVERROR(ENOMEM);
s->uvblock_base[0] = s->block_base + sbs * 64 * 64 * bytesperpixel;
s->uvblock_base[1] = s->uvblock_base[0] + sbs * chroma_blocks * bytesperpixel;
s->eob_base = (uint8_t *) (s->uvblock_base[1] + sbs * chroma_blocks * bytesperpixel);
s->uveob_base[0] = s->eob_base + 16 * 16 * sbs;
s->uveob_base[1] = s->uveob_base[0] + chroma_eobs * sbs;
} else {
s->b_base = av_malloc(sizeof(VP9Block));
s->block_base = av_mallocz((64 * 64 + 2 * chroma_blocks) * bytesperpixel * sizeof(int16_t) +
16 * 16 + 2 * chroma_eobs);
if (!s->b_base || !s->block_base)
return AVERROR(ENOMEM);
s->uvblock_base[0] = s->block_base + 64 * 64 * bytesperpixel;
s->uvblock_base[1] = s->uvblock_base[0] + chroma_blocks * bytesperpixel;
s->eob_base = (uint8_t *) (s->uvblock_base[1] + chroma_blocks * bytesperpixel);
s->uveob_base[0] = s->eob_base + 16 * 16;
s->uveob_base[1] = s->uveob_base[0] + chroma_eobs;
}
s->block_alloc_using_2pass = s->s.frames[CUR_FRAME].uses_2pass;
return 0;
}
// for some reason the sign bit is at the end, not the start, of a bit sequence
static av_always_inline int get_sbits_inv(GetBitContext *gb, int n)
{
int v = get_bits(gb, n);
return get_bits1(gb) ? -v : v;
}
static av_always_inline int inv_recenter_nonneg(int v, int m)
{
return v > 2 * m ? v : v & 1 ? m - ((v + 1) >> 1) : m + (v >> 1);
}
// differential forward probability updates
static int update_prob(VP56RangeCoder *c, int p)
{
static const int inv_map_table[255] = {
7, 20, 33, 46, 59, 72, 85, 98, 111, 124, 137, 150, 163, 176,
189, 202, 215, 228, 241, 254, 1, 2, 3, 4, 5, 6, 8, 9,
10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 21, 22, 23, 24,
25, 26, 27, 28, 29, 30, 31, 32, 34, 35, 36, 37, 38, 39,
40, 41, 42, 43, 44, 45, 47, 48, 49, 50, 51, 52, 53, 54,
55, 56, 57, 58, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69,
70, 71, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84,
86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 99, 100,
101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 112, 113, 114, 115,
116, 117, 118, 119, 120, 121, 122, 123, 125, 126, 127, 128, 129, 130,
131, 132, 133, 134, 135, 136, 138, 139, 140, 141, 142, 143, 144, 145,
146, 147, 148, 149, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160,
161, 162, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175,
177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 190, 191,
192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 203, 204, 205, 206,
207, 208, 209, 210, 211, 212, 213, 214, 216, 217, 218, 219, 220, 221,
222, 223, 224, 225, 226, 227, 229, 230, 231, 232, 233, 234, 235, 236,
237, 238, 239, 240, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251,
252, 253, 253,
};
int d;
/* This code is trying to do a differential probability update. For a
* current probability A in the range [1, 255], the difference to a new
* probability of any value can be expressed differentially as 1-A,255-A
* where some part of this (absolute range) exists both in positive as
* well as the negative part, whereas another part only exists in one
* half. We're trying to code this shared part differentially, i.e.
* times two where the value of the lowest bit specifies the sign, and
* the single part is then coded on top of this. This absolute difference
* then again has a value of [0,254], but a bigger value in this range
* indicates that we're further away from the original value A, so we
* can code this as a VLC code, since higher values are increasingly
* unlikely. The first 20 values in inv_map_table[] allow 'cheap, rough'
* updates vs. the 'fine, exact' updates further down the range, which
* adds one extra dimension to this differential update model. */
if (!vp8_rac_get(c)) {
d = vp8_rac_get_uint(c, 4) + 0;
} else if (!vp8_rac_get(c)) {
d = vp8_rac_get_uint(c, 4) + 16;
} else if (!vp8_rac_get(c)) {
d = vp8_rac_get_uint(c, 5) + 32;
} else {
d = vp8_rac_get_uint(c, 7);
if (d >= 65)
d = (d << 1) - 65 + vp8_rac_get(c);
d += 64;
av_assert2(d < FF_ARRAY_ELEMS(inv_map_table));
}
return p <= 128 ? 1 + inv_recenter_nonneg(inv_map_table[d], p - 1) :
255 - inv_recenter_nonneg(inv_map_table[d], 255 - p);
}
static int read_colorspace_details(AVCodecContext *ctx)
{
static const enum AVColorSpace colorspaces[8] = {
AVCOL_SPC_UNSPECIFIED, AVCOL_SPC_BT470BG, AVCOL_SPC_BT709, AVCOL_SPC_SMPTE170M,
AVCOL_SPC_SMPTE240M, AVCOL_SPC_BT2020_NCL, AVCOL_SPC_RESERVED, AVCOL_SPC_RGB,
};
VP9Context *s = ctx->priv_data;
int bits = ctx->profile <= 1 ? 0 : 1 + get_bits1(&s->gb); // 0:8, 1:10, 2:12
s->bpp_index = bits;
s->s.h.bpp = 8 + bits * 2;
s->bytesperpixel = (7 + s->s.h.bpp) >> 3;
ctx->colorspace = colorspaces[get_bits(&s->gb, 3)];
if (ctx->colorspace == AVCOL_SPC_RGB) { // RGB = profile 1
static const enum AVPixelFormat pix_fmt_rgb[3] = {
AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12
};
s->ss_h = s->ss_v = 0;
ctx->color_range = AVCOL_RANGE_JPEG;
s->pix_fmt = pix_fmt_rgb[bits];
if (ctx->profile & 1) {
if (get_bits1(&s->gb)) {
av_log(ctx, AV_LOG_ERROR, "Reserved bit set in RGB\n");
return AVERROR_INVALIDDATA;
}
} else {
av_log(ctx, AV_LOG_ERROR, "RGB not supported in profile %d\n",
ctx->profile);
return AVERROR_INVALIDDATA;
}
} else {
static const enum AVPixelFormat pix_fmt_for_ss[3][2 /* v */][2 /* h */] = {
{ { AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P },
{ AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV420P } },
{ { AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV422P10 },
{ AV_PIX_FMT_YUV440P10, AV_PIX_FMT_YUV420P10 } },
{ { AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV422P12 },
{ AV_PIX_FMT_YUV440P12, AV_PIX_FMT_YUV420P12 } }
};
ctx->color_range = get_bits1(&s->gb) ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
if (ctx->profile & 1) {
s->ss_h = get_bits1(&s->gb);
s->ss_v = get_bits1(&s->gb);
s->pix_fmt = pix_fmt_for_ss[bits][s->ss_v][s->ss_h];
if (s->pix_fmt == AV_PIX_FMT_YUV420P) {
av_log(ctx, AV_LOG_ERROR, "YUV 4:2:0 not supported in profile %d\n",
ctx->profile);
return AVERROR_INVALIDDATA;
} else if (get_bits1(&s->gb)) {
av_log(ctx, AV_LOG_ERROR, "Profile %d color details reserved bit set\n",
ctx->profile);
return AVERROR_INVALIDDATA;
}
} else {
s->ss_h = s->ss_v = 1;
s->pix_fmt = pix_fmt_for_ss[bits][1][1];
}
}
return 0;
}
static int decode_frame_header(AVCodecContext *ctx,
const uint8_t *data, int size, int *ref)
{
VP9Context *s = ctx->priv_data;
int c, i, j, k, l, m, n, w, h, max, size2, res, sharp;
int last_invisible;
const uint8_t *data2;
/* general header */
if ((res = init_get_bits8(&s->gb, data, size)) < 0) {
av_log(ctx, AV_LOG_ERROR, "Failed to initialize bitstream reader\n");
return res;
}
if (get_bits(&s->gb, 2) != 0x2) { // frame marker
av_log(ctx, AV_LOG_ERROR, "Invalid frame marker\n");
return AVERROR_INVALIDDATA;
}
ctx->profile = get_bits1(&s->gb);
ctx->profile |= get_bits1(&s->gb) << 1;
if (ctx->profile == 3) ctx->profile += get_bits1(&s->gb);
if (ctx->profile > 3) {
av_log(ctx, AV_LOG_ERROR, "Profile %d is not yet supported\n", ctx->profile);
return AVERROR_INVALIDDATA;
}
s->s.h.profile = ctx->profile;
if (get_bits1(&s->gb)) {
*ref = get_bits(&s->gb, 3);
return 0;
}
s->last_keyframe = s->s.h.keyframe;
s->s.h.keyframe = !get_bits1(&s->gb);
last_invisible = s->s.h.invisible;
s->s.h.invisible = !get_bits1(&s->gb);
s->s.h.errorres = get_bits1(&s->gb);
s->s.h.use_last_frame_mvs = !s->s.h.errorres && !last_invisible;
if (s->s.h.keyframe) {
if (get_bits_long(&s->gb, 24) != VP9_SYNCCODE) { // synccode
av_log(ctx, AV_LOG_ERROR, "Invalid sync code\n");
return AVERROR_INVALIDDATA;
}
if ((res = read_colorspace_details(ctx)) < 0)
return res;
// for profile 1, here follows the subsampling bits
s->s.h.refreshrefmask = 0xff;
w = get_bits(&s->gb, 16) + 1;
h = get_bits(&s->gb, 16) + 1;
if (get_bits1(&s->gb)) // display size
skip_bits(&s->gb, 32);
} else {
s->s.h.intraonly = s->s.h.invisible ? get_bits1(&s->gb) : 0;
s->s.h.resetctx = s->s.h.errorres ? 0 : get_bits(&s->gb, 2);
if (s->s.h.intraonly) {
if (get_bits_long(&s->gb, 24) != VP9_SYNCCODE) { // synccode
av_log(ctx, AV_LOG_ERROR, "Invalid sync code\n");
return AVERROR_INVALIDDATA;
}
if (ctx->profile >= 1) {
if ((res = read_colorspace_details(ctx)) < 0)
return res;
} else {
s->ss_h = s->ss_v = 1;
s->s.h.bpp = 8;
s->bpp_index = 0;
s->bytesperpixel = 1;
s->pix_fmt = AV_PIX_FMT_YUV420P;
ctx->colorspace = AVCOL_SPC_BT470BG;
ctx->color_range = AVCOL_RANGE_JPEG;
}
s->s.h.refreshrefmask = get_bits(&s->gb, 8);
w = get_bits(&s->gb, 16) + 1;
h = get_bits(&s->gb, 16) + 1;
if (get_bits1(&s->gb)) // display size
skip_bits(&s->gb, 32);
} else {
s->s.h.refreshrefmask = get_bits(&s->gb, 8);
s->s.h.refidx[0] = get_bits(&s->gb, 3);
s->s.h.signbias[0] = get_bits1(&s->gb) && !s->s.h.errorres;
s->s.h.refidx[1] = get_bits(&s->gb, 3);
s->s.h.signbias[1] = get_bits1(&s->gb) && !s->s.h.errorres;
s->s.h.refidx[2] = get_bits(&s->gb, 3);
s->s.h.signbias[2] = get_bits1(&s->gb) && !s->s.h.errorres;
if (!s->s.refs[s->s.h.refidx[0]].f->buf[0] ||
!s->s.refs[s->s.h.refidx[1]].f->buf[0] ||
!s->s.refs[s->s.h.refidx[2]].f->buf[0]) {
av_log(ctx, AV_LOG_ERROR, "Not all references are available\n");
return AVERROR_INVALIDDATA;
}
if (get_bits1(&s->gb)) {
w = s->s.refs[s->s.h.refidx[0]].f->width;
h = s->s.refs[s->s.h.refidx[0]].f->height;
} else if (get_bits1(&s->gb)) {
w = s->s.refs[s->s.h.refidx[1]].f->width;
h = s->s.refs[s->s.h.refidx[1]].f->height;
} else if (get_bits1(&s->gb)) {
w = s->s.refs[s->s.h.refidx[2]].f->width;
h = s->s.refs[s->s.h.refidx[2]].f->height;
} else {
w = get_bits(&s->gb, 16) + 1;
h = get_bits(&s->gb, 16) + 1;
}
// Note that in this code, "CUR_FRAME" is actually before we
// have formally allocated a frame, and thus actually represents
// the _last_ frame
s->s.h.use_last_frame_mvs &= s->s.frames[CUR_FRAME].tf.f->width == w &&
s->s.frames[CUR_FRAME].tf.f->height == h;
if (get_bits1(&s->gb)) // display size
skip_bits(&s->gb, 32);
s->s.h.highprecisionmvs = get_bits1(&s->gb);
s->s.h.filtermode = get_bits1(&s->gb) ? FILTER_SWITCHABLE :
get_bits(&s->gb, 2);
s->s.h.allowcompinter = s->s.h.signbias[0] != s->s.h.signbias[1] ||
s->s.h.signbias[0] != s->s.h.signbias[2];
if (s->s.h.allowcompinter) {
if (s->s.h.signbias[0] == s->s.h.signbias[1]) {
s->s.h.fixcompref = 2;
s->s.h.varcompref[0] = 0;
s->s.h.varcompref[1] = 1;
} else if (s->s.h.signbias[0] == s->s.h.signbias[2]) {
s->s.h.fixcompref = 1;
s->s.h.varcompref[0] = 0;
s->s.h.varcompref[1] = 2;
} else {
s->s.h.fixcompref = 0;
s->s.h.varcompref[0] = 1;
s->s.h.varcompref[1] = 2;
}
}
}
}
s->s.h.refreshctx = s->s.h.errorres ? 0 : get_bits1(&s->gb);
s->s.h.parallelmode = s->s.h.errorres ? 1 : get_bits1(&s->gb);
s->s.h.framectxid = c = get_bits(&s->gb, 2);
if (s->s.h.keyframe || s->s.h.intraonly)
s->s.h.framectxid = 0; // BUG: libvpx ignores this field in keyframes
/* loopfilter header data */
if (s->s.h.keyframe || s->s.h.errorres || s->s.h.intraonly) {
// reset loopfilter defaults
s->s.h.lf_delta.ref[0] = 1;
s->s.h.lf_delta.ref[1] = 0;
s->s.h.lf_delta.ref[2] = -1;
s->s.h.lf_delta.ref[3] = -1;
s->s.h.lf_delta.mode[0] = 0;
s->s.h.lf_delta.mode[1] = 0;
memset(s->s.h.segmentation.feat, 0, sizeof(s->s.h.segmentation.feat));
}
s->s.h.filter.level = get_bits(&s->gb, 6);
sharp = get_bits(&s->gb, 3);
// if sharpness changed, reinit lim/mblim LUTs. if it didn't change, keep
// the old cache values since they are still valid
if (s->s.h.filter.sharpness != sharp)
memset(s->filter_lut.lim_lut, 0, sizeof(s->filter_lut.lim_lut));
s->s.h.filter.sharpness = sharp;
if ((s->s.h.lf_delta.enabled = get_bits1(&s->gb))) {
if ((s->s.h.lf_delta.updated = get_bits1(&s->gb))) {
for (i = 0; i < 4; i++)
if (get_bits1(&s->gb))
s->s.h.lf_delta.ref[i] = get_sbits_inv(&s->gb, 6);
for (i = 0; i < 2; i++)
if (get_bits1(&s->gb))
s->s.h.lf_delta.mode[i] = get_sbits_inv(&s->gb, 6);
}
}
/* quantization header data */
s->s.h.yac_qi = get_bits(&s->gb, 8);
s->s.h.ydc_qdelta = get_bits1(&s->gb) ? get_sbits_inv(&s->gb, 4) : 0;
s->s.h.uvdc_qdelta = get_bits1(&s->gb) ? get_sbits_inv(&s->gb, 4) : 0;
s->s.h.uvac_qdelta = get_bits1(&s->gb) ? get_sbits_inv(&s->gb, 4) : 0;
s->s.h.lossless = s->s.h.yac_qi == 0 && s->s.h.ydc_qdelta == 0 &&
s->s.h.uvdc_qdelta == 0 && s->s.h.uvac_qdelta == 0;
if (s->s.h.lossless)
ctx->properties |= FF_CODEC_PROPERTY_LOSSLESS;
/* segmentation header info */
if ((s->s.h.segmentation.enabled = get_bits1(&s->gb))) {
if ((s->s.h.segmentation.update_map = get_bits1(&s->gb))) {
for (i = 0; i < 7; i++)
s->s.h.segmentation.prob[i] = get_bits1(&s->gb) ?
get_bits(&s->gb, 8) : 255;
if ((s->s.h.segmentation.temporal = get_bits1(&s->gb))) {
for (i = 0; i < 3; i++)
s->s.h.segmentation.pred_prob[i] = get_bits1(&s->gb) ?
get_bits(&s->gb, 8) : 255;
}
}
if (get_bits1(&s->gb)) {
s->s.h.segmentation.absolute_vals = get_bits1(&s->gb);
for (i = 0; i < 8; i++) {
if ((s->s.h.segmentation.feat[i].q_enabled = get_bits1(&s->gb)))
s->s.h.segmentation.feat[i].q_val = get_sbits_inv(&s->gb, 8);
if ((s->s.h.segmentation.feat[i].lf_enabled = get_bits1(&s->gb)))
s->s.h.segmentation.feat[i].lf_val = get_sbits_inv(&s->gb, 6);
if ((s->s.h.segmentation.feat[i].ref_enabled = get_bits1(&s->gb)))
s->s.h.segmentation.feat[i].ref_val = get_bits(&s->gb, 2);
s->s.h.segmentation.feat[i].skip_enabled = get_bits1(&s->gb);
}
}
}
// set qmul[] based on Y/UV, AC/DC and segmentation Q idx deltas
for (i = 0; i < (s->s.h.segmentation.enabled ? 8 : 1); i++) {
int qyac, qydc, quvac, quvdc, lflvl, sh;
if (s->s.h.segmentation.enabled && s->s.h.segmentation.feat[i].q_enabled) {
if (s->s.h.segmentation.absolute_vals)
qyac = av_clip_uintp2(s->s.h.segmentation.feat[i].q_val, 8);
else
qyac = av_clip_uintp2(s->s.h.yac_qi + s->s.h.segmentation.feat[i].q_val, 8);
} else {
qyac = s->s.h.yac_qi;
}
qydc = av_clip_uintp2(qyac + s->s.h.ydc_qdelta, 8);
quvdc = av_clip_uintp2(qyac + s->s.h.uvdc_qdelta, 8);
quvac = av_clip_uintp2(qyac + s->s.h.uvac_qdelta, 8);
qyac = av_clip_uintp2(qyac, 8);
s->s.h.segmentation.feat[i].qmul[0][0] = vp9_dc_qlookup[s->bpp_index][qydc];
s->s.h.segmentation.feat[i].qmul[0][1] = vp9_ac_qlookup[s->bpp_index][qyac];
s->s.h.segmentation.feat[i].qmul[1][0] = vp9_dc_qlookup[s->bpp_index][quvdc];
s->s.h.segmentation.feat[i].qmul[1][1] = vp9_ac_qlookup[s->bpp_index][quvac];
sh = s->s.h.filter.level >= 32;
if (s->s.h.segmentation.enabled && s->s.h.segmentation.feat[i].lf_enabled) {
if (s->s.h.segmentation.absolute_vals)
lflvl = av_clip_uintp2(s->s.h.segmentation.feat[i].lf_val, 6);
else
lflvl = av_clip_uintp2(s->s.h.filter.level + s->s.h.segmentation.feat[i].lf_val, 6);
} else {
lflvl = s->s.h.filter.level;
}
if (s->s.h.lf_delta.enabled) {
s->s.h.segmentation.feat[i].lflvl[0][0] =
s->s.h.segmentation.feat[i].lflvl[0][1] =
av_clip_uintp2(lflvl + (s->s.h.lf_delta.ref[0] * (1 << sh)), 6);
for (j = 1; j < 4; j++) {
s->s.h.segmentation.feat[i].lflvl[j][0] =
av_clip_uintp2(lflvl + ((s->s.h.lf_delta.ref[j] +
s->s.h.lf_delta.mode[0]) * (1 << sh)), 6);
s->s.h.segmentation.feat[i].lflvl[j][1] =
av_clip_uintp2(lflvl + ((s->s.h.lf_delta.ref[j] +
s->s.h.lf_delta.mode[1]) * (1 << sh)), 6);
}
} else {
memset(s->s.h.segmentation.feat[i].lflvl, lflvl,
sizeof(s->s.h.segmentation.feat[i].lflvl));
}
}
/* tiling info */
if ((res = update_size(ctx, w, h)) < 0) {
av_log(ctx, AV_LOG_ERROR, "Failed to initialize decoder for %dx%d @ %d\n",
w, h, s->pix_fmt);
return res;
}
for (s->s.h.tiling.log2_tile_cols = 0;
s->sb_cols > (64 << s->s.h.tiling.log2_tile_cols);
s->s.h.tiling.log2_tile_cols++) ;
for (max = 0; (s->sb_cols >> max) >= 4; max++) ;
max = FFMAX(0, max - 1);
while (max > s->s.h.tiling.log2_tile_cols) {
if (get_bits1(&s->gb))
s->s.h.tiling.log2_tile_cols++;
else
break;
}
s->s.h.tiling.log2_tile_rows = decode012(&s->gb);
s->s.h.tiling.tile_rows = 1 << s->s.h.tiling.log2_tile_rows;
if (s->s.h.tiling.tile_cols != (1 << s->s.h.tiling.log2_tile_cols)) {
s->s.h.tiling.tile_cols = 1 << s->s.h.tiling.log2_tile_cols;
s->c_b = av_fast_realloc(s->c_b, &s->c_b_size,
sizeof(VP56RangeCoder) * s->s.h.tiling.tile_cols);
if (!s->c_b) {
av_log(ctx, AV_LOG_ERROR, "Ran out of memory during range coder init\n");
return AVERROR(ENOMEM);
}
}
/* check reference frames */
if (!s->s.h.keyframe && !s->s.h.intraonly) {
for (i = 0; i < 3; i++) {
AVFrame *ref = s->s.refs[s->s.h.refidx[i]].f;
int refw = ref->width, refh = ref->height;
if (ref->format != ctx->pix_fmt) {
av_log(ctx, AV_LOG_ERROR,
"Ref pixfmt (%s) did not match current frame (%s)",
av_get_pix_fmt_name(ref->format),
av_get_pix_fmt_name(ctx->pix_fmt));
return AVERROR_INVALIDDATA;
} else if (refw == w && refh == h) {
s->mvscale[i][0] = s->mvscale[i][1] = 0;
} else {
if (w * 2 < refw || h * 2 < refh || w > 16 * refw || h > 16 * refh) {
av_log(ctx, AV_LOG_ERROR,
"Invalid ref frame dimensions %dx%d for frame size %dx%d\n",
refw, refh, w, h);
return AVERROR_INVALIDDATA;
}
s->mvscale[i][0] = (refw << 14) / w;
s->mvscale[i][1] = (refh << 14) / h;
s->mvstep[i][0] = 16 * s->mvscale[i][0] >> 14;
s->mvstep[i][1] = 16 * s->mvscale[i][1] >> 14;
}
}
}
if (s->s.h.keyframe || s->s.h.errorres || (s->s.h.intraonly && s->s.h.resetctx == 3)) {
s->prob_ctx[0].p = s->prob_ctx[1].p = s->prob_ctx[2].p =
s->prob_ctx[3].p = vp9_default_probs;
memcpy(s->prob_ctx[0].coef, vp9_default_coef_probs,
sizeof(vp9_default_coef_probs));
memcpy(s->prob_ctx[1].coef, vp9_default_coef_probs,
sizeof(vp9_default_coef_probs));
memcpy(s->prob_ctx[2].coef, vp9_default_coef_probs,
sizeof(vp9_default_coef_probs));
memcpy(s->prob_ctx[3].coef, vp9_default_coef_probs,
sizeof(vp9_default_coef_probs));
} else if (s->s.h.intraonly && s->s.h.resetctx == 2) {
s->prob_ctx[c].p = vp9_default_probs;
memcpy(s->prob_ctx[c].coef, vp9_default_coef_probs,
sizeof(vp9_default_coef_probs));
}
// next 16 bits is size of the rest of the header (arith-coded)
s->s.h.compressed_header_size = size2 = get_bits(&s->gb, 16);
s->s.h.uncompressed_header_size = (get_bits_count(&s->gb) + 7) / 8;
data2 = align_get_bits(&s->gb);
if (size2 > size - (data2 - data)) {
av_log(ctx, AV_LOG_ERROR, "Invalid compressed header size\n");
return AVERROR_INVALIDDATA;
}
ff_vp56_init_range_decoder(&s->c, data2, size2);
if (vp56_rac_get_prob_branchy(&s->c, 128)) { // marker bit
av_log(ctx, AV_LOG_ERROR, "Marker bit was set\n");
return AVERROR_INVALIDDATA;
}
if (s->s.h.keyframe || s->s.h.intraonly) {
memset(s->counts.coef, 0, sizeof(s->counts.coef));
memset(s->counts.eob, 0, sizeof(s->counts.eob));
} else {
memset(&s->counts, 0, sizeof(s->counts));
}
// FIXME is it faster to not copy here, but do it down in the fw updates
// as explicit copies if the fw update is missing (and skip the copy upon
// fw update)?
s->prob.p = s->prob_ctx[c].p;
// txfm updates
if (s->s.h.lossless) {
s->s.h.txfmmode = TX_4X4;
} else {
s->s.h.txfmmode = vp8_rac_get_uint(&s->c, 2);
if (s->s.h.txfmmode == 3)
s->s.h.txfmmode += vp8_rac_get(&s->c);
if (s->s.h.txfmmode == TX_SWITCHABLE) {
for (i = 0; i < 2; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.tx8p[i] = update_prob(&s->c, s->prob.p.tx8p[i]);
for (i = 0; i < 2; i++)
for (j = 0; j < 2; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.tx16p[i][j] =
update_prob(&s->c, s->prob.p.tx16p[i][j]);
for (i = 0; i < 2; i++)
for (j = 0; j < 3; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.tx32p[i][j] =
update_prob(&s->c, s->prob.p.tx32p[i][j]);
}
}
// coef updates
for (i = 0; i < 4; i++) {
uint8_t (*ref)[2][6][6][3] = s->prob_ctx[c].coef[i];
if (vp8_rac_get(&s->c)) {
for (j = 0; j < 2; j++)
for (k = 0; k < 2; k++)
for (l = 0; l < 6; l++)
for (m = 0; m < 6; m++) {
uint8_t *p = s->prob.coef[i][j][k][l][m];
uint8_t *r = ref[j][k][l][m];
if (m >= 3 && l == 0) // dc only has 3 pt
break;
for (n = 0; n < 3; n++) {
if (vp56_rac_get_prob_branchy(&s->c, 252)) {
p[n] = update_prob(&s->c, r[n]);
} else {
p[n] = r[n];
}
}
p[3] = 0;
}
} else {
for (j = 0; j < 2; j++)
for (k = 0; k < 2; k++)
for (l = 0; l < 6; l++)
for (m = 0; m < 6; m++) {
uint8_t *p = s->prob.coef[i][j][k][l][m];
uint8_t *r = ref[j][k][l][m];
if (m > 3 && l == 0) // dc only has 3 pt
break;
memcpy(p, r, 3);
p[3] = 0;
}
}
if (s->s.h.txfmmode == i)
break;
}
// mode updates
for (i = 0; i < 3; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.skip[i] = update_prob(&s->c, s->prob.p.skip[i]);
if (!s->s.h.keyframe && !s->s.h.intraonly) {
for (i = 0; i < 7; i++)
for (j = 0; j < 3; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_mode[i][j] =
update_prob(&s->c, s->prob.p.mv_mode[i][j]);
if (s->s.h.filtermode == FILTER_SWITCHABLE)
for (i = 0; i < 4; i++)
for (j = 0; j < 2; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.filter[i][j] =
update_prob(&s->c, s->prob.p.filter[i][j]);
for (i = 0; i < 4; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.intra[i] = update_prob(&s->c, s->prob.p.intra[i]);
if (s->s.h.allowcompinter) {
s->s.h.comppredmode = vp8_rac_get(&s->c);
if (s->s.h.comppredmode)
s->s.h.comppredmode += vp8_rac_get(&s->c);
if (s->s.h.comppredmode == PRED_SWITCHABLE)
for (i = 0; i < 5; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.comp[i] =
update_prob(&s->c, s->prob.p.comp[i]);
} else {
s->s.h.comppredmode = PRED_SINGLEREF;
}
if (s->s.h.comppredmode != PRED_COMPREF) {
for (i = 0; i < 5; i++) {
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.single_ref[i][0] =
update_prob(&s->c, s->prob.p.single_ref[i][0]);
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.single_ref[i][1] =
update_prob(&s->c, s->prob.p.single_ref[i][1]);
}
}
if (s->s.h.comppredmode != PRED_SINGLEREF) {
for (i = 0; i < 5; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.comp_ref[i] =
update_prob(&s->c, s->prob.p.comp_ref[i]);
}
for (i = 0; i < 4; i++)
for (j = 0; j < 9; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.y_mode[i][j] =
update_prob(&s->c, s->prob.p.y_mode[i][j]);
for (i = 0; i < 4; i++)
for (j = 0; j < 4; j++)
for (k = 0; k < 3; k++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.partition[3 - i][j][k] =
update_prob(&s->c, s->prob.p.partition[3 - i][j][k]);
// mv fields don't use the update_prob subexp model for some reason
for (i = 0; i < 3; i++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_joint[i] = (vp8_rac_get_uint(&s->c, 7) << 1) | 1;
for (i = 0; i < 2; i++) {
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].sign = (vp8_rac_get_uint(&s->c, 7) << 1) | 1;
for (j = 0; j < 10; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].classes[j] =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].class0 = (vp8_rac_get_uint(&s->c, 7) << 1) | 1;
for (j = 0; j < 10; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].bits[j] =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
}
for (i = 0; i < 2; i++) {
for (j = 0; j < 2; j++)
for (k = 0; k < 3; k++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].class0_fp[j][k] =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
for (j = 0; j < 3; j++)
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].fp[j] =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
}
if (s->s.h.highprecisionmvs) {
for (i = 0; i < 2; i++) {
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].class0_hp =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
if (vp56_rac_get_prob_branchy(&s->c, 252))
s->prob.p.mv_comp[i].hp =
(vp8_rac_get_uint(&s->c, 7) << 1) | 1;
}
}
}
return (data2 - data) + size2;
}
static av_always_inline void clamp_mv(VP56mv *dst, const VP56mv *src,
VP9Context *s)
{
dst->x = av_clip(src->x, s->min_mv.x, s->max_mv.x);
dst->y = av_clip(src->y, s->min_mv.y, s->max_mv.y);
}
static void find_ref_mvs(VP9Context *s,
VP56mv *pmv, int ref, int z, int idx, int sb)
{
static const int8_t mv_ref_blk_off[N_BS_SIZES][8][2] = {
[BS_64x64] = {{ 3, -1 }, { -1, 3 }, { 4, -1 }, { -1, 4 },
{ -1, -1 }, { 0, -1 }, { -1, 0 }, { 6, -1 }},
[BS_64x32] = {{ 0, -1 }, { -1, 0 }, { 4, -1 }, { -1, 2 },
{ -1, -1 }, { 0, -3 }, { -3, 0 }, { 2, -1 }},
[BS_32x64] = {{ -1, 0 }, { 0, -1 }, { -1, 4 }, { 2, -1 },
{ -1, -1 }, { -3, 0 }, { 0, -3 }, { -1, 2 }},
[BS_32x32] = {{ 1, -1 }, { -1, 1 }, { 2, -1 }, { -1, 2 },
{ -1, -1 }, { 0, -3 }, { -3, 0 }, { -3, -3 }},
[BS_32x16] = {{ 0, -1 }, { -1, 0 }, { 2, -1 }, { -1, -1 },
{ -1, 1 }, { 0, -3 }, { -3, 0 }, { -3, -3 }},
[BS_16x32] = {{ -1, 0 }, { 0, -1 }, { -1, 2 }, { -1, -1 },
{ 1, -1 }, { -3, 0 }, { 0, -3 }, { -3, -3 }},
[BS_16x16] = {{ 0, -1 }, { -1, 0 }, { 1, -1 }, { -1, 1 },
{ -1, -1 }, { 0, -3 }, { -3, 0 }, { -3, -3 }},
[BS_16x8] = {{ 0, -1 }, { -1, 0 }, { 1, -1 }, { -1, -1 },
{ 0, -2 }, { -2, 0 }, { -2, -1 }, { -1, -2 }},
[BS_8x16] = {{ -1, 0 }, { 0, -1 }, { -1, 1 }, { -1, -1 },
{ -2, 0 }, { 0, -2 }, { -1, -2 }, { -2, -1 }},
[BS_8x8] = {{ 0, -1 }, { -1, 0 }, { -1, -1 }, { 0, -2 },
{ -2, 0 }, { -1, -2 }, { -2, -1 }, { -2, -2 }},
[BS_8x4] = {{ 0, -1 }, { -1, 0 }, { -1, -1 }, { 0, -2 },
{ -2, 0 }, { -1, -2 }, { -2, -1 }, { -2, -2 }},
[BS_4x8] = {{ 0, -1 }, { -1, 0 }, { -1, -1 }, { 0, -2 },
{ -2, 0 }, { -1, -2 }, { -2, -1 }, { -2, -2 }},
[BS_4x4] = {{ 0, -1 }, { -1, 0 }, { -1, -1 }, { 0, -2 },
{ -2, 0 }, { -1, -2 }, { -2, -1 }, { -2, -2 }},
};
VP9Block *b = s->b;
int row = s->row, col = s->col, row7 = s->row7;
const int8_t (*p)[2] = mv_ref_blk_off[b->bs];
#define INVALID_MV 0x80008000U
uint32_t mem = INVALID_MV, mem_sub8x8 = INVALID_MV;
int i;
#define RETURN_DIRECT_MV(mv) \
do { \
uint32_t m = AV_RN32A(&mv); \
if (!idx) { \
AV_WN32A(pmv, m); \
return; \
} else if (mem == INVALID_MV) { \
mem = m; \
} else if (m != mem) { \
AV_WN32A(pmv, m); \
return; \
} \
} while (0)
if (sb >= 0) {
if (sb == 2 || sb == 1) {
RETURN_DIRECT_MV(b->mv[0][z]);
} else if (sb == 3) {
RETURN_DIRECT_MV(b->mv[2][z]);
RETURN_DIRECT_MV(b->mv[1][z]);
RETURN_DIRECT_MV(b->mv[0][z]);
}
#define RETURN_MV(mv) \
do { \
if (sb > 0) { \
VP56mv tmp; \
uint32_t m; \
av_assert2(idx == 1); \
av_assert2(mem != INVALID_MV); \
if (mem_sub8x8 == INVALID_MV) { \
clamp_mv(&tmp, &mv, s); \
m = AV_RN32A(&tmp); \
if (m != mem) { \
AV_WN32A(pmv, m); \
return; \
} \
mem_sub8x8 = AV_RN32A(&mv); \
} else if (mem_sub8x8 != AV_RN32A(&mv)) { \
clamp_mv(&tmp, &mv, s); \
m = AV_RN32A(&tmp); \
if (m != mem) { \
AV_WN32A(pmv, m); \
} else { \
/* BUG I'm pretty sure this isn't the intention */ \
AV_WN32A(pmv, 0); \
} \
return; \
} \
} else { \
uint32_t m = AV_RN32A(&mv); \
if (!idx) { \
clamp_mv(pmv, &mv, s); \
return; \
} else if (mem == INVALID_MV) { \
mem = m; \
} else if (m != mem) { \
clamp_mv(pmv, &mv, s); \
return; \
} \
} \
} while (0)
if (row > 0) {
struct VP9mvrefPair *mv = &s->s.frames[CUR_FRAME].mv[(row - 1) * s->sb_cols * 8 + col];
if (mv->ref[0] == ref) {
RETURN_MV(s->above_mv_ctx[2 * col + (sb & 1)][0]);
} else if (mv->ref[1] == ref) {
RETURN_MV(s->above_mv_ctx[2 * col + (sb & 1)][1]);
}
}
if (col > s->tile_col_start) {
struct VP9mvrefPair *mv = &s->s.frames[CUR_FRAME].mv[row * s->sb_cols * 8 + col - 1];
if (mv->ref[0] == ref) {
RETURN_MV(s->left_mv_ctx[2 * row7 + (sb >> 1)][0]);
} else if (mv->ref[1] == ref) {
RETURN_MV(s->left_mv_ctx[2 * row7 + (sb >> 1)][1]);
}
}
i = 2;
} else {
i = 0;
}
// previously coded MVs in this neighbourhood, using same reference frame
for (; i < 8; i++) {
int c = p[i][0] + col, r = p[i][1] + row;
if (c >= s->tile_col_start && c < s->cols && r >= 0 && r < s->rows) {
struct VP9mvrefPair *mv = &s->s.frames[CUR_FRAME].mv[r * s->sb_cols * 8 + c];
if (mv->ref[0] == ref) {
RETURN_MV(mv->mv[0]);
} else if (mv->ref[1] == ref) {
RETURN_MV(mv->mv[1]);
}
}
}
// MV at this position in previous frame, using same reference frame
if (s->s.h.use_last_frame_mvs) {
struct VP9mvrefPair *mv = &s->s.frames[REF_FRAME_MVPAIR].mv[row * s->sb_cols * 8 + col];
if (!s->s.frames[REF_FRAME_MVPAIR].uses_2pass)
ff_thread_await_progress(&s->s.frames[REF_FRAME_MVPAIR].tf, row >> 3, 0);
if (mv->ref[0] == ref) {
RETURN_MV(mv->mv[0]);
} else if (mv->ref[1] == ref) {
RETURN_MV(mv->mv[1]);
}
}
#define RETURN_SCALE_MV(mv, scale) \
do { \
if (scale) { \
VP56mv mv_temp = { -mv.x, -mv.y }; \
RETURN_MV(mv_temp); \
} else { \
RETURN_MV(mv); \
} \
} while (0)
// previously coded MVs in this neighbourhood, using different reference frame
for (i = 0; i < 8; i++) {
int c = p[i][0] + col, r = p[i][1] + row;
if (c >= s->tile_col_start && c < s->cols && r >= 0 && r < s->rows) {
struct VP9mvrefPair *mv = &s->s.frames[CUR_FRAME].mv[r * s->sb_cols * 8 + c];
if (mv->ref[0] != ref && mv->ref[0] >= 0) {
RETURN_SCALE_MV(mv->mv[0], s->s.h.signbias[mv->ref[0]] != s->s.h.signbias[ref]);
}
if (mv->ref[1] != ref && mv->ref[1] >= 0 &&
// BUG - libvpx has this condition regardless of whether
// we used the first ref MV and pre-scaling
AV_RN32A(&mv->mv[0]) != AV_RN32A(&mv->mv[1])) {
RETURN_SCALE_MV(mv->mv[1], s->s.h.signbias[mv->ref[1]] != s->s.h.signbias[ref]);
}
}
}
// MV at this position in previous frame, using different reference frame
if (s->s.h.use_last_frame_mvs) {
struct VP9mvrefPair *mv = &s->s.frames[REF_FRAME_MVPAIR].mv[row * s->sb_cols * 8 + col];
// no need to await_progress, because we already did that above
if (mv->ref[0] != ref && mv->ref[0] >= 0) {
RETURN_SCALE_MV(mv->mv[0], s->s.h.signbias[mv->ref[0]] != s->s.h.signbias[ref]);
}
if (mv->ref[1] != ref && mv->ref[1] >= 0 &&
// BUG - libvpx has this condition regardless of whether
// we used the first ref MV and pre-scaling
AV_RN32A(&mv->mv[0]) != AV_RN32A(&mv->mv[1])) {
RETURN_SCALE_MV(mv->mv[1], s->s.h.signbias[mv->ref[1]] != s->s.h.signbias[ref]);
}
}
AV_ZERO32(pmv);
clamp_mv(pmv, pmv, s);
#undef INVALID_MV
#undef RETURN_MV
#undef RETURN_SCALE_MV
}
static av_always_inline int read_mv_component(VP9Context *s, int idx, int hp)
{
int bit, sign = vp56_rac_get_prob(&s->c, s->prob.p.mv_comp[idx].sign);
int n, c = vp8_rac_get_tree(&s->c, vp9_mv_class_tree,
s->prob.p.mv_comp[idx].classes);
s->counts.mv_comp[idx].sign[sign]++;
s->counts.mv_comp[idx].classes[c]++;
if (c) {
int m;
for (n = 0, m = 0; m < c; m++) {
bit = vp56_rac_get_prob(&s->c, s->prob.p.mv_comp[idx].bits[m]);
n |= bit << m;
s->counts.mv_comp[idx].bits[m][bit]++;
}
n <<= 3;
bit = vp8_rac_get_tree(&s->c, vp9_mv_fp_tree, s->prob.p.mv_comp[idx].fp);
n |= bit << 1;
s->counts.mv_comp[idx].fp[bit]++;
if (hp) {
bit = vp56_rac_get_prob(&s->c, s->prob.p.mv_comp[idx].hp);
s->counts.mv_comp[idx].hp[bit]++;
n |= bit;
} else {
n |= 1;
// bug in libvpx - we count for bw entropy purposes even if the
// bit wasn't coded
s->counts.mv_comp[idx].hp[1]++;
}
n += 8 << c;
} else {
n = vp56_rac_get_prob(&s->c, s->prob.p.mv_comp[idx].class0);
s->counts.mv_comp[idx].class0[n]++;
bit = vp8_rac_get_tree(&s->c, vp9_mv_fp_tree,
s->prob.p.mv_comp[idx].class0_fp[n]);
s->counts.mv_comp[idx].class0_fp[n][bit]++;
n = (n << 3) | (bit << 1);
if (hp) {
bit = vp56_rac_get_prob(&s->c, s->prob.p.mv_comp[idx].class0_hp);
s->counts.mv_comp[idx].class0_hp[bit]++;
n |= bit;
} else {
n |= 1;
// bug in libvpx - we count for bw entropy purposes even if the
// bit wasn't coded
s->counts.mv_comp[idx].class0_hp[1]++;
}
}
return sign ? -(n + 1) : (n + 1);
}
static void fill_mv(VP9Context *s,
VP56mv *mv, int mode, int sb)
{
VP9Block *b = s->b;
if (mode == ZEROMV) {
AV_ZERO64(mv);
} else {
int hp;
// FIXME cache this value and reuse for other subblocks
find_ref_mvs(s, &mv[0], b->ref[0], 0, mode == NEARMV,
mode == NEWMV ? -1 : sb);
// FIXME maybe move this code into find_ref_mvs()
if ((mode == NEWMV || sb == -1) &&
!(hp = s->s.h.highprecisionmvs && abs(mv[0].x) < 64 && abs(mv[0].y) < 64)) {
if (mv[0].y & 1) {
if (mv[0].y < 0)
mv[0].y++;
else
mv[0].y--;
}
if (mv[0].x & 1) {
if (mv[0].x < 0)
mv[0].x++;
else
mv[0].x--;
}
}
if (mode == NEWMV) {
enum MVJoint j = vp8_rac_get_tree(&s->c, vp9_mv_joint_tree,
s->prob.p.mv_joint);
s->counts.mv_joint[j]++;
if (j >= MV_JOINT_V)
mv[0].y += read_mv_component(s, 0, hp);
if (j & 1)
mv[0].x += read_mv_component(s, 1, hp);
}
if (b->comp) {
// FIXME cache this value and reuse for other subblocks
find_ref_mvs(s, &mv[1], b->ref[1], 1, mode == NEARMV,
mode == NEWMV ? -1 : sb);
if ((mode == NEWMV || sb == -1) &&
!(hp = s->s.h.highprecisionmvs && abs(mv[1].x) < 64 && abs(mv[1].y) < 64)) {
if (mv[1].y & 1) {
if (mv[1].y < 0)
mv[1].y++;
else
mv[1].y--;
}
if (mv[1].x & 1) {
if (mv[1].x < 0)
mv[1].x++;
else
mv[1].x--;
}
}
if (mode == NEWMV) {
enum MVJoint j = vp8_rac_get_tree(&s->c, vp9_mv_joint_tree,
s->prob.p.mv_joint);
s->counts.mv_joint[j]++;
if (j >= MV_JOINT_V)
mv[1].y += read_mv_component(s, 0, hp);
if (j & 1)
mv[1].x += read_mv_component(s, 1, hp);
}
}
}
}
static av_always_inline void setctx_2d(uint8_t *ptr, int w, int h,
ptrdiff_t stride, int v)
{
switch (w) {
case 1:
do {
*ptr = v;
ptr += stride;
} while (--h);
break;
case 2: {
int v16 = v * 0x0101;
do {
AV_WN16A(ptr, v16);
ptr += stride;
} while (--h);
break;
}
case 4: {
uint32_t v32 = v * 0x01010101;
do {
AV_WN32A(ptr, v32);
ptr += stride;
} while (--h);
break;
}
case 8: {
#if HAVE_FAST_64BIT
uint64_t v64 = v * 0x0101010101010101ULL;
do {
AV_WN64A(ptr, v64);
ptr += stride;
} while (--h);
#else
uint32_t v32 = v * 0x01010101;
do {
AV_WN32A(ptr, v32);
AV_WN32A(ptr + 4, v32);
ptr += stride;
} while (--h);
#endif
break;
}
}
}
static void decode_mode(AVCodecContext *ctx)
{
static const uint8_t left_ctx[N_BS_SIZES] = {
0x0, 0x8, 0x0, 0x8, 0xc, 0x8, 0xc, 0xe, 0xc, 0xe, 0xf, 0xe, 0xf
};
static const uint8_t above_ctx[N_BS_SIZES] = {
0x0, 0x0, 0x8, 0x8, 0x8, 0xc, 0xc, 0xc, 0xe, 0xe, 0xe, 0xf, 0xf
};
static const uint8_t max_tx_for_bl_bp[N_BS_SIZES] = {
TX_32X32, TX_32X32, TX_32X32, TX_32X32, TX_16X16, TX_16X16,
TX_16X16, TX_8X8, TX_8X8, TX_8X8, TX_4X4, TX_4X4, TX_4X4
};
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
int row = s->row, col = s->col, row7 = s->row7;
enum TxfmMode max_tx = max_tx_for_bl_bp[b->bs];
int bw4 = bwh_tab[1][b->bs][0], w4 = FFMIN(s->cols - col, bw4);
int bh4 = bwh_tab[1][b->bs][1], h4 = FFMIN(s->rows - row, bh4), y;
int have_a = row > 0, have_l = col > s->tile_col_start;
int vref, filter_id;
if (!s->s.h.segmentation.enabled) {
b->seg_id = 0;
} else if (s->s.h.keyframe || s->s.h.intraonly) {
b->seg_id = !s->s.h.segmentation.update_map ? 0 :
vp8_rac_get_tree(&s->c, vp9_segmentation_tree, s->s.h.segmentation.prob);
} else if (!s->s.h.segmentation.update_map ||
(s->s.h.segmentation.temporal &&
vp56_rac_get_prob_branchy(&s->c,
s->s.h.segmentation.pred_prob[s->above_segpred_ctx[col] +
s->left_segpred_ctx[row7]]))) {
if (!s->s.h.errorres && s->s.frames[REF_FRAME_SEGMAP].segmentation_map) {
int pred = 8, x;
uint8_t *refsegmap = s->s.frames[REF_FRAME_SEGMAP].segmentation_map;
if (!s->s.frames[REF_FRAME_SEGMAP].uses_2pass)
ff_thread_await_progress(&s->s.frames[REF_FRAME_SEGMAP].tf, row >> 3, 0);
for (y = 0; y < h4; y++) {
int idx_base = (y + row) * 8 * s->sb_cols + col;
for (x = 0; x < w4; x++)
pred = FFMIN(pred, refsegmap[idx_base + x]);
}
av_assert1(pred < 8);
b->seg_id = pred;
} else {
b->seg_id = 0;
}
memset(&s->above_segpred_ctx[col], 1, w4);
memset(&s->left_segpred_ctx[row7], 1, h4);
} else {
b->seg_id = vp8_rac_get_tree(&s->c, vp9_segmentation_tree,
s->s.h.segmentation.prob);
memset(&s->above_segpred_ctx[col], 0, w4);
memset(&s->left_segpred_ctx[row7], 0, h4);
}
if (s->s.h.segmentation.enabled &&
(s->s.h.segmentation.update_map || s->s.h.keyframe || s->s.h.intraonly)) {
setctx_2d(&s->s.frames[CUR_FRAME].segmentation_map[row * 8 * s->sb_cols + col],
bw4, bh4, 8 * s->sb_cols, b->seg_id);
}
b->skip = s->s.h.segmentation.enabled &&
s->s.h.segmentation.feat[b->seg_id].skip_enabled;
if (!b->skip) {
int c = s->left_skip_ctx[row7] + s->above_skip_ctx[col];
b->skip = vp56_rac_get_prob(&s->c, s->prob.p.skip[c]);
s->counts.skip[c][b->skip]++;
}
if (s->s.h.keyframe || s->s.h.intraonly) {
b->intra = 1;
} else if (s->s.h.segmentation.enabled && s->s.h.segmentation.feat[b->seg_id].ref_enabled) {
b->intra = !s->s.h.segmentation.feat[b->seg_id].ref_val;
} else {
int c, bit;
if (have_a && have_l) {
c = s->above_intra_ctx[col] + s->left_intra_ctx[row7];
c += (c == 2);
} else {
c = have_a ? 2 * s->above_intra_ctx[col] :
have_l ? 2 * s->left_intra_ctx[row7] : 0;
}
bit = vp56_rac_get_prob(&s->c, s->prob.p.intra[c]);
s->counts.intra[c][bit]++;
b->intra = !bit;
}
if ((b->intra || !b->skip) && s->s.h.txfmmode == TX_SWITCHABLE) {
int c;
if (have_a) {
if (have_l) {
c = (s->above_skip_ctx[col] ? max_tx :
s->above_txfm_ctx[col]) +
(s->left_skip_ctx[row7] ? max_tx :
s->left_txfm_ctx[row7]) > max_tx;
} else {
c = s->above_skip_ctx[col] ? 1 :
(s->above_txfm_ctx[col] * 2 > max_tx);
}
} else if (have_l) {
c = s->left_skip_ctx[row7] ? 1 :
(s->left_txfm_ctx[row7] * 2 > max_tx);
} else {
c = 1;
}
switch (max_tx) {
case TX_32X32:
b->tx = vp56_rac_get_prob(&s->c, s->prob.p.tx32p[c][0]);
if (b->tx) {
b->tx += vp56_rac_get_prob(&s->c, s->prob.p.tx32p[c][1]);
if (b->tx == 2)
b->tx += vp56_rac_get_prob(&s->c, s->prob.p.tx32p[c][2]);
}
s->counts.tx32p[c][b->tx]++;
break;
case TX_16X16:
b->tx = vp56_rac_get_prob(&s->c, s->prob.p.tx16p[c][0]);
if (b->tx)
b->tx += vp56_rac_get_prob(&s->c, s->prob.p.tx16p[c][1]);
s->counts.tx16p[c][b->tx]++;
break;
case TX_8X8:
b->tx = vp56_rac_get_prob(&s->c, s->prob.p.tx8p[c]);
s->counts.tx8p[c][b->tx]++;
break;
case TX_4X4:
b->tx = TX_4X4;
break;
}
} else {
b->tx = FFMIN(max_tx, s->s.h.txfmmode);
}
if (s->s.h.keyframe || s->s.h.intraonly) {
uint8_t *a = &s->above_mode_ctx[col * 2];
uint8_t *l = &s->left_mode_ctx[(row7) << 1];
b->comp = 0;
if (b->bs > BS_8x8) {
// FIXME the memory storage intermediates here aren't really
// necessary, they're just there to make the code slightly
// simpler for now
b->mode[0] = a[0] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_ymode_probs[a[0]][l[0]]);
if (b->bs != BS_8x4) {
b->mode[1] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_ymode_probs[a[1]][b->mode[0]]);
l[0] = a[1] = b->mode[1];
} else {
l[0] = a[1] = b->mode[1] = b->mode[0];
}
if (b->bs != BS_4x8) {
b->mode[2] = a[0] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_ymode_probs[a[0]][l[1]]);
if (b->bs != BS_8x4) {
b->mode[3] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_ymode_probs[a[1]][b->mode[2]]);
l[1] = a[1] = b->mode[3];
} else {
l[1] = a[1] = b->mode[3] = b->mode[2];
}
} else {
b->mode[2] = b->mode[0];
l[1] = a[1] = b->mode[3] = b->mode[1];
}
} else {
b->mode[0] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_ymode_probs[*a][*l]);
b->mode[3] = b->mode[2] = b->mode[1] = b->mode[0];
// FIXME this can probably be optimized
memset(a, b->mode[0], bwh_tab[0][b->bs][0]);
memset(l, b->mode[0], bwh_tab[0][b->bs][1]);
}
b->uvmode = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
vp9_default_kf_uvmode_probs[b->mode[3]]);
} else if (b->intra) {
b->comp = 0;
if (b->bs > BS_8x8) {
b->mode[0] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.y_mode[0]);
s->counts.y_mode[0][b->mode[0]]++;
if (b->bs != BS_8x4) {
b->mode[1] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.y_mode[0]);
s->counts.y_mode[0][b->mode[1]]++;
} else {
b->mode[1] = b->mode[0];
}
if (b->bs != BS_4x8) {
b->mode[2] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.y_mode[0]);
s->counts.y_mode[0][b->mode[2]]++;
if (b->bs != BS_8x4) {
b->mode[3] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.y_mode[0]);
s->counts.y_mode[0][b->mode[3]]++;
} else {
b->mode[3] = b->mode[2];
}
} else {
b->mode[2] = b->mode[0];
b->mode[3] = b->mode[1];
}
} else {
static const uint8_t size_group[10] = {
3, 3, 3, 3, 2, 2, 2, 1, 1, 1
};
int sz = size_group[b->bs];
b->mode[0] = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.y_mode[sz]);
b->mode[1] = b->mode[2] = b->mode[3] = b->mode[0];
s->counts.y_mode[sz][b->mode[3]]++;
}
b->uvmode = vp8_rac_get_tree(&s->c, vp9_intramode_tree,
s->prob.p.uv_mode[b->mode[3]]);
s->counts.uv_mode[b->mode[3]][b->uvmode]++;
} else {
static const uint8_t inter_mode_ctx_lut[14][14] = {
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5 },
{ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 2, 2, 1, 3 },
{ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 2, 2, 1, 3 },
{ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 1, 1, 0, 3 },
{ 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 3, 3, 3, 4 },
};
if (s->s.h.segmentation.enabled && s->s.h.segmentation.feat[b->seg_id].ref_enabled) {
av_assert2(s->s.h.segmentation.feat[b->seg_id].ref_val != 0);
b->comp = 0;
b->ref[0] = s->s.h.segmentation.feat[b->seg_id].ref_val - 1;
} else {
// read comp_pred flag
if (s->s.h.comppredmode != PRED_SWITCHABLE) {
b->comp = s->s.h.comppredmode == PRED_COMPREF;
} else {
int c;
// FIXME add intra as ref=0xff (or -1) to make these easier?
if (have_a) {
if (have_l) {
if (s->above_comp_ctx[col] && s->left_comp_ctx[row7]) {
c = 4;
} else if (s->above_comp_ctx[col]) {
c = 2 + (s->left_intra_ctx[row7] ||
s->left_ref_ctx[row7] == s->s.h.fixcompref);
} else if (s->left_comp_ctx[row7]) {
c = 2 + (s->above_intra_ctx[col] ||
s->above_ref_ctx[col] == s->s.h.fixcompref);
} else {
c = (!s->above_intra_ctx[col] &&
s->above_ref_ctx[col] == s->s.h.fixcompref) ^
(!s->left_intra_ctx[row7] &&
s->left_ref_ctx[row & 7] == s->s.h.fixcompref);
}
} else {
c = s->above_comp_ctx[col] ? 3 :
(!s->above_intra_ctx[col] && s->above_ref_ctx[col] == s->s.h.fixcompref);
}
} else if (have_l) {
c = s->left_comp_ctx[row7] ? 3 :
(!s->left_intra_ctx[row7] && s->left_ref_ctx[row7] == s->s.h.fixcompref);
} else {
c = 1;
}
b->comp = vp56_rac_get_prob(&s->c, s->prob.p.comp[c]);
s->counts.comp[c][b->comp]++;
}
// read actual references
// FIXME probably cache a few variables here to prevent repetitive
// memory accesses below
if (b->comp) /* two references */ {
int fix_idx = s->s.h.signbias[s->s.h.fixcompref], var_idx = !fix_idx, c, bit;
b->ref[fix_idx] = s->s.h.fixcompref;
// FIXME can this codeblob be replaced by some sort of LUT?
if (have_a) {
if (have_l) {
if (s->above_intra_ctx[col]) {
if (s->left_intra_ctx[row7]) {
c = 2;
} else {
c = 1 + 2 * (s->left_ref_ctx[row7] != s->s.h.varcompref[1]);
}
} else if (s->left_intra_ctx[row7]) {
c = 1 + 2 * (s->above_ref_ctx[col] != s->s.h.varcompref[1]);
} else {
int refl = s->left_ref_ctx[row7], refa = s->above_ref_ctx[col];
if (refl == refa && refa == s->s.h.varcompref[1]) {
c = 0;
} else if (!s->left_comp_ctx[row7] && !s->above_comp_ctx[col]) {
if ((refa == s->s.h.fixcompref && refl == s->s.h.varcompref[0]) ||
(refl == s->s.h.fixcompref && refa == s->s.h.varcompref[0])) {
c = 4;
} else {
c = (refa == refl) ? 3 : 1;
}
} else if (!s->left_comp_ctx[row7]) {
if (refa == s->s.h.varcompref[1] && refl != s->s.h.varcompref[1]) {
c = 1;
} else {
c = (refl == s->s.h.varcompref[1] &&
refa != s->s.h.varcompref[1]) ? 2 : 4;
}
} else if (!s->above_comp_ctx[col]) {
if (refl == s->s.h.varcompref[1] && refa != s->s.h.varcompref[1]) {
c = 1;
} else {
c = (refa == s->s.h.varcompref[1] &&
refl != s->s.h.varcompref[1]) ? 2 : 4;
}
} else {
c = (refl == refa) ? 4 : 2;
}
}
} else {
if (s->above_intra_ctx[col]) {
c = 2;
} else if (s->above_comp_ctx[col]) {
c = 4 * (s->above_ref_ctx[col] != s->s.h.varcompref[1]);
} else {
c = 3 * (s->above_ref_ctx[col] != s->s.h.varcompref[1]);
}
}
} else if (have_l) {
if (s->left_intra_ctx[row7]) {
c = 2;
} else if (s->left_comp_ctx[row7]) {
c = 4 * (s->left_ref_ctx[row7] != s->s.h.varcompref[1]);
} else {
c = 3 * (s->left_ref_ctx[row7] != s->s.h.varcompref[1]);
}
} else {
c = 2;
}
bit = vp56_rac_get_prob(&s->c, s->prob.p.comp_ref[c]);
b->ref[var_idx] = s->s.h.varcompref[bit];
s->counts.comp_ref[c][bit]++;
} else /* single reference */ {
int bit, c;
if (have_a && !s->above_intra_ctx[col]) {
if (have_l && !s->left_intra_ctx[row7]) {
if (s->left_comp_ctx[row7]) {
if (s->above_comp_ctx[col]) {
c = 1 + (!s->s.h.fixcompref || !s->left_ref_ctx[row7] ||
!s->above_ref_ctx[col]);
} else {
c = (3 * !s->above_ref_ctx[col]) +
(!s->s.h.fixcompref || !s->left_ref_ctx[row7]);
}
} else if (s->above_comp_ctx[col]) {
c = (3 * !s->left_ref_ctx[row7]) +
(!s->s.h.fixcompref || !s->above_ref_ctx[col]);
} else {
c = 2 * !s->left_ref_ctx[row7] + 2 * !s->above_ref_ctx[col];
}
} else if (s->above_intra_ctx[col]) {
c = 2;
} else if (s->above_comp_ctx[col]) {
c = 1 + (!s->s.h.fixcompref || !s->above_ref_ctx[col]);
} else {
c = 4 * (!s->above_ref_ctx[col]);
}
} else if (have_l && !s->left_intra_ctx[row7]) {
if (s->left_intra_ctx[row7]) {
c = 2;
} else if (s->left_comp_ctx[row7]) {
c = 1 + (!s->s.h.fixcompref || !s->left_ref_ctx[row7]);
} else {
c = 4 * (!s->left_ref_ctx[row7]);
}
} else {
c = 2;
}
bit = vp56_rac_get_prob(&s->c, s->prob.p.single_ref[c][0]);
s->counts.single_ref[c][0][bit]++;
if (!bit) {
b->ref[0] = 0;
} else {
// FIXME can this codeblob be replaced by some sort of LUT?
if (have_a) {
if (have_l) {
if (s->left_intra_ctx[row7]) {
if (s->above_intra_ctx[col]) {
c = 2;
} else if (s->above_comp_ctx[col]) {
c = 1 + 2 * (s->s.h.fixcompref == 1 ||
s->above_ref_ctx[col] == 1);
} else if (!s->above_ref_ctx[col]) {
c = 3;
} else {
c = 4 * (s->above_ref_ctx[col] == 1);
}
} else if (s->above_intra_ctx[col]) {
if (s->left_intra_ctx[row7]) {
c = 2;
} else if (s->left_comp_ctx[row7]) {
c = 1 + 2 * (s->s.h.fixcompref == 1 ||
s->left_ref_ctx[row7] == 1);
} else if (!s->left_ref_ctx[row7]) {
c = 3;
} else {
c = 4 * (s->left_ref_ctx[row7] == 1);
}
} else if (s->above_comp_ctx[col]) {
if (s->left_comp_ctx[row7]) {
if (s->left_ref_ctx[row7] == s->above_ref_ctx[col]) {
c = 3 * (s->s.h.fixcompref == 1 ||
s->left_ref_ctx[row7] == 1);
} else {
c = 2;
}
} else if (!s->left_ref_ctx[row7]) {
c = 1 + 2 * (s->s.h.fixcompref == 1 ||
s->above_ref_ctx[col] == 1);
} else {
c = 3 * (s->left_ref_ctx[row7] == 1) +
(s->s.h.fixcompref == 1 || s->above_ref_ctx[col] == 1);
}
} else if (s->left_comp_ctx[row7]) {
if (!s->above_ref_ctx[col]) {
c = 1 + 2 * (s->s.h.fixcompref == 1 ||
s->left_ref_ctx[row7] == 1);
} else {
c = 3 * (s->above_ref_ctx[col] == 1) +
(s->s.h.fixcompref == 1 || s->left_ref_ctx[row7] == 1);
}
} else if (!s->above_ref_ctx[col]) {
if (!s->left_ref_ctx[row7]) {
c = 3;
} else {
c = 4 * (s->left_ref_ctx[row7] == 1);
}
} else if (!s->left_ref_ctx[row7]) {
c = 4 * (s->above_ref_ctx[col] == 1);
} else {
c = 2 * (s->left_ref_ctx[row7] == 1) +
2 * (s->above_ref_ctx[col] == 1);
}
} else {
if (s->above_intra_ctx[col] ||
(!s->above_comp_ctx[col] && !s->above_ref_ctx[col])) {
c = 2;
} else if (s->above_comp_ctx[col]) {
c = 3 * (s->s.h.fixcompref == 1 || s->above_ref_ctx[col] == 1);
} else {
c = 4 * (s->above_ref_ctx[col] == 1);
}
}
} else if (have_l) {
if (s->left_intra_ctx[row7] ||
(!s->left_comp_ctx[row7] && !s->left_ref_ctx[row7])) {
c = 2;
} else if (s->left_comp_ctx[row7]) {
c = 3 * (s->s.h.fixcompref == 1 || s->left_ref_ctx[row7] == 1);
} else {
c = 4 * (s->left_ref_ctx[row7] == 1);
}
} else {
c = 2;
}
bit = vp56_rac_get_prob(&s->c, s->prob.p.single_ref[c][1]);
s->counts.single_ref[c][1][bit]++;
b->ref[0] = 1 + bit;
}
}
}
if (b->bs <= BS_8x8) {
if (s->s.h.segmentation.enabled && s->s.h.segmentation.feat[b->seg_id].skip_enabled) {
b->mode[0] = b->mode[1] = b->mode[2] = b->mode[3] = ZEROMV;
} else {
static const uint8_t off[10] = {
3, 0, 0, 1, 0, 0, 0, 0, 0, 0
};
// FIXME this needs to use the LUT tables from find_ref_mvs
// because not all are -1,0/0,-1
int c = inter_mode_ctx_lut[s->above_mode_ctx[col + off[b->bs]]]
[s->left_mode_ctx[row7 + off[b->bs]]];
b->mode[0] = vp8_rac_get_tree(&s->c, vp9_inter_mode_tree,
s->prob.p.mv_mode[c]);
b->mode[1] = b->mode[2] = b->mode[3] = b->mode[0];
s->counts.mv_mode[c][b->mode[0] - 10]++;
}
}
if (s->s.h.filtermode == FILTER_SWITCHABLE) {
int c;
if (have_a && s->above_mode_ctx[col] >= NEARESTMV) {
if (have_l && s->left_mode_ctx[row7] >= NEARESTMV) {
c = s->above_filter_ctx[col] == s->left_filter_ctx[row7] ?
s->left_filter_ctx[row7] : 3;
} else {
c = s->above_filter_ctx[col];
}
} else if (have_l && s->left_mode_ctx[row7] >= NEARESTMV) {
c = s->left_filter_ctx[row7];
} else {
c = 3;
}
filter_id = vp8_rac_get_tree(&s->c, vp9_filter_tree,
s->prob.p.filter[c]);
s->counts.filter[c][filter_id]++;
b->filter = vp9_filter_lut[filter_id];
} else {
b->filter = s->s.h.filtermode;
}
if (b->bs > BS_8x8) {
int c = inter_mode_ctx_lut[s->above_mode_ctx[col]][s->left_mode_ctx[row7]];
b->mode[0] = vp8_rac_get_tree(&s->c, vp9_inter_mode_tree,
s->prob.p.mv_mode[c]);
s->counts.mv_mode[c][b->mode[0] - 10]++;
fill_mv(s, b->mv[0], b->mode[0], 0);
if (b->bs != BS_8x4) {
b->mode[1] = vp8_rac_get_tree(&s->c, vp9_inter_mode_tree,
s->prob.p.mv_mode[c]);
s->counts.mv_mode[c][b->mode[1] - 10]++;
fill_mv(s, b->mv[1], b->mode[1], 1);
} else {
b->mode[1] = b->mode[0];
AV_COPY32(&b->mv[1][0], &b->mv[0][0]);
AV_COPY32(&b->mv[1][1], &b->mv[0][1]);
}
if (b->bs != BS_4x8) {
b->mode[2] = vp8_rac_get_tree(&s->c, vp9_inter_mode_tree,
s->prob.p.mv_mode[c]);
s->counts.mv_mode[c][b->mode[2] - 10]++;
fill_mv(s, b->mv[2], b->mode[2], 2);
if (b->bs != BS_8x4) {
b->mode[3] = vp8_rac_get_tree(&s->c, vp9_inter_mode_tree,
s->prob.p.mv_mode[c]);
s->counts.mv_mode[c][b->mode[3] - 10]++;
fill_mv(s, b->mv[3], b->mode[3], 3);
} else {
b->mode[3] = b->mode[2];
AV_COPY32(&b->mv[3][0], &b->mv[2][0]);
AV_COPY32(&b->mv[3][1], &b->mv[2][1]);
}
} else {
b->mode[2] = b->mode[0];
AV_COPY32(&b->mv[2][0], &b->mv[0][0]);
AV_COPY32(&b->mv[2][1], &b->mv[0][1]);
b->mode[3] = b->mode[1];
AV_COPY32(&b->mv[3][0], &b->mv[1][0]);
AV_COPY32(&b->mv[3][1], &b->mv[1][1]);
}
} else {
fill_mv(s, b->mv[0], b->mode[0], -1);
AV_COPY32(&b->mv[1][0], &b->mv[0][0]);
AV_COPY32(&b->mv[2][0], &b->mv[0][0]);
AV_COPY32(&b->mv[3][0], &b->mv[0][0]);
AV_COPY32(&b->mv[1][1], &b->mv[0][1]);
AV_COPY32(&b->mv[2][1], &b->mv[0][1]);
AV_COPY32(&b->mv[3][1], &b->mv[0][1]);
}
vref = b->ref[b->comp ? s->s.h.signbias[s->s.h.varcompref[0]] : 0];
}
#if HAVE_FAST_64BIT
#define SPLAT_CTX(var, val, n) \
switch (n) { \
case 1: var = val; break; \
case 2: AV_WN16A(&var, val * 0x0101); break; \
case 4: AV_WN32A(&var, val * 0x01010101); break; \
case 8: AV_WN64A(&var, val * 0x0101010101010101ULL); break; \
case 16: { \
uint64_t v64 = val * 0x0101010101010101ULL; \
AV_WN64A( &var, v64); \
AV_WN64A(&((uint8_t *) &var)[8], v64); \
break; \
} \
}
#else
#define SPLAT_CTX(var, val, n) \
switch (n) { \
case 1: var = val; break; \
case 2: AV_WN16A(&var, val * 0x0101); break; \
case 4: AV_WN32A(&var, val * 0x01010101); break; \
case 8: { \
uint32_t v32 = val * 0x01010101; \
AV_WN32A( &var, v32); \
AV_WN32A(&((uint8_t *) &var)[4], v32); \
break; \
} \
case 16: { \
uint32_t v32 = val * 0x01010101; \
AV_WN32A( &var, v32); \
AV_WN32A(&((uint8_t *) &var)[4], v32); \
AV_WN32A(&((uint8_t *) &var)[8], v32); \
AV_WN32A(&((uint8_t *) &var)[12], v32); \
break; \
} \
}
#endif
switch (bwh_tab[1][b->bs][0]) {
#define SET_CTXS(dir, off, n) \
do { \
SPLAT_CTX(s->dir##_skip_ctx[off], b->skip, n); \
SPLAT_CTX(s->dir##_txfm_ctx[off], b->tx, n); \
SPLAT_CTX(s->dir##_partition_ctx[off], dir##_ctx[b->bs], n); \
if (!s->s.h.keyframe && !s->s.h.intraonly) { \
SPLAT_CTX(s->dir##_intra_ctx[off], b->intra, n); \
SPLAT_CTX(s->dir##_comp_ctx[off], b->comp, n); \
SPLAT_CTX(s->dir##_mode_ctx[off], b->mode[3], n); \
if (!b->intra) { \
SPLAT_CTX(s->dir##_ref_ctx[off], vref, n); \
if (s->s.h.filtermode == FILTER_SWITCHABLE) { \
SPLAT_CTX(s->dir##_filter_ctx[off], filter_id, n); \
} \
} \
} \
} while (0)
case 1: SET_CTXS(above, col, 1); break;
case 2: SET_CTXS(above, col, 2); break;
case 4: SET_CTXS(above, col, 4); break;
case 8: SET_CTXS(above, col, 8); break;
}
switch (bwh_tab[1][b->bs][1]) {
case 1: SET_CTXS(left, row7, 1); break;
case 2: SET_CTXS(left, row7, 2); break;
case 4: SET_CTXS(left, row7, 4); break;
case 8: SET_CTXS(left, row7, 8); break;
}
#undef SPLAT_CTX
#undef SET_CTXS
if (!s->s.h.keyframe && !s->s.h.intraonly) {
if (b->bs > BS_8x8) {
int mv0 = AV_RN32A(&b->mv[3][0]), mv1 = AV_RN32A(&b->mv[3][1]);
AV_COPY32(&s->left_mv_ctx[row7 * 2 + 0][0], &b->mv[1][0]);
AV_COPY32(&s->left_mv_ctx[row7 * 2 + 0][1], &b->mv[1][1]);
AV_WN32A(&s->left_mv_ctx[row7 * 2 + 1][0], mv0);
AV_WN32A(&s->left_mv_ctx[row7 * 2 + 1][1], mv1);
AV_COPY32(&s->above_mv_ctx[col * 2 + 0][0], &b->mv[2][0]);
AV_COPY32(&s->above_mv_ctx[col * 2 + 0][1], &b->mv[2][1]);
AV_WN32A(&s->above_mv_ctx[col * 2 + 1][0], mv0);
AV_WN32A(&s->above_mv_ctx[col * 2 + 1][1], mv1);
} else {
int n, mv0 = AV_RN32A(&b->mv[3][0]), mv1 = AV_RN32A(&b->mv[3][1]);
for (n = 0; n < w4 * 2; n++) {
AV_WN32A(&s->above_mv_ctx[col * 2 + n][0], mv0);
AV_WN32A(&s->above_mv_ctx[col * 2 + n][1], mv1);
}
for (n = 0; n < h4 * 2; n++) {
AV_WN32A(&s->left_mv_ctx[row7 * 2 + n][0], mv0);
AV_WN32A(&s->left_mv_ctx[row7 * 2 + n][1], mv1);
}
}
}
// FIXME kinda ugly
for (y = 0; y < h4; y++) {
int x, o = (row + y) * s->sb_cols * 8 + col;
struct VP9mvrefPair *mv = &s->s.frames[CUR_FRAME].mv[o];
if (b->intra) {
for (x = 0; x < w4; x++) {
mv[x].ref[0] =
mv[x].ref[1] = -1;
}
} else if (b->comp) {
for (x = 0; x < w4; x++) {
mv[x].ref[0] = b->ref[0];
mv[x].ref[1] = b->ref[1];
AV_COPY32(&mv[x].mv[0], &b->mv[3][0]);
AV_COPY32(&mv[x].mv[1], &b->mv[3][1]);
}
} else {
for (x = 0; x < w4; x++) {
mv[x].ref[0] = b->ref[0];
mv[x].ref[1] = -1;
AV_COPY32(&mv[x].mv[0], &b->mv[3][0]);
}
}
}
}
// FIXME merge cnt/eob arguments?
static av_always_inline int
decode_coeffs_b_generic(VP56RangeCoder *c, int16_t *coef, int n_coeffs,
int is_tx32x32, int is8bitsperpixel, int bpp, unsigned (*cnt)[6][3],
unsigned (*eob)[6][2], uint8_t (*p)[6][11],
int nnz, const int16_t *scan, const int16_t (*nb)[2],
const int16_t *band_counts, const int16_t *qmul)
{
int i = 0, band = 0, band_left = band_counts[band];
uint8_t *tp = p[0][nnz];
uint8_t cache[1024];
do {
int val, rc;
val = vp56_rac_get_prob_branchy(c, tp[0]); // eob
eob[band][nnz][val]++;
if (!val)
break;
skip_eob:
if (!vp56_rac_get_prob_branchy(c, tp[1])) { // zero
cnt[band][nnz][0]++;
if (!--band_left)
band_left = band_counts[++band];
cache[scan[i]] = 0;
nnz = (1 + cache[nb[i][0]] + cache[nb[i][1]]) >> 1;
tp = p[band][nnz];
if (++i == n_coeffs)
break; //invalid input; blocks should end with EOB
goto skip_eob;
}
rc = scan[i];
if (!vp56_rac_get_prob_branchy(c, tp[2])) { // one
cnt[band][nnz][1]++;
val = 1;
cache[rc] = 1;
} else {
// fill in p[3-10] (model fill) - only once per frame for each pos
if (!tp[3])
memcpy(&tp[3], vp9_model_pareto8[tp[2]], 8);
cnt[band][nnz][2]++;
if (!vp56_rac_get_prob_branchy(c, tp[3])) { // 2, 3, 4
if (!vp56_rac_get_prob_branchy(c, tp[4])) {
cache[rc] = val = 2;
} else {
val = 3 + vp56_rac_get_prob(c, tp[5]);
cache[rc] = 3;
}
} else if (!vp56_rac_get_prob_branchy(c, tp[6])) { // cat1/2
cache[rc] = 4;
if (!vp56_rac_get_prob_branchy(c, tp[7])) {
val = 5 + vp56_rac_get_prob(c, 159);
} else {
val = 7 + (vp56_rac_get_prob(c, 165) << 1);
val += vp56_rac_get_prob(c, 145);
}
} else { // cat 3-6
cache[rc] = 5;
if (!vp56_rac_get_prob_branchy(c, tp[8])) {
if (!vp56_rac_get_prob_branchy(c, tp[9])) {
val = 11 + (vp56_rac_get_prob(c, 173) << 2);
val += (vp56_rac_get_prob(c, 148) << 1);
val += vp56_rac_get_prob(c, 140);
} else {
val = 19 + (vp56_rac_get_prob(c, 176) << 3);
val += (vp56_rac_get_prob(c, 155) << 2);
val += (vp56_rac_get_prob(c, 140) << 1);
val += vp56_rac_get_prob(c, 135);
}
} else if (!vp56_rac_get_prob_branchy(c, tp[10])) {
val = 35 + (vp56_rac_get_prob(c, 180) << 4);
val += (vp56_rac_get_prob(c, 157) << 3);
val += (vp56_rac_get_prob(c, 141) << 2);
val += (vp56_rac_get_prob(c, 134) << 1);
val += vp56_rac_get_prob(c, 130);
} else {
val = 67;
if (!is8bitsperpixel) {
if (bpp == 12) {
val += vp56_rac_get_prob(c, 255) << 17;
val += vp56_rac_get_prob(c, 255) << 16;
}
val += (vp56_rac_get_prob(c, 255) << 15);
val += (vp56_rac_get_prob(c, 255) << 14);
}
val += (vp56_rac_get_prob(c, 254) << 13);
val += (vp56_rac_get_prob(c, 254) << 12);
val += (vp56_rac_get_prob(c, 254) << 11);
val += (vp56_rac_get_prob(c, 252) << 10);
val += (vp56_rac_get_prob(c, 249) << 9);
val += (vp56_rac_get_prob(c, 243) << 8);
val += (vp56_rac_get_prob(c, 230) << 7);
val += (vp56_rac_get_prob(c, 196) << 6);
val += (vp56_rac_get_prob(c, 177) << 5);
val += (vp56_rac_get_prob(c, 153) << 4);
val += (vp56_rac_get_prob(c, 140) << 3);
val += (vp56_rac_get_prob(c, 133) << 2);
val += (vp56_rac_get_prob(c, 130) << 1);
val += vp56_rac_get_prob(c, 129);
}
}
}
#define STORE_COEF(c, i, v) do { \
if (is8bitsperpixel) { \
c[i] = v; \
} else { \
AV_WN32A(&c[i * 2], v); \
} \
} while (0)
if (!--band_left)
band_left = band_counts[++band];
if (is_tx32x32)
STORE_COEF(coef, rc, ((vp8_rac_get(c) ? -val : val) * qmul[!!i]) / 2);
else
STORE_COEF(coef, rc, (vp8_rac_get(c) ? -val : val) * qmul[!!i]);
nnz = (1 + cache[nb[i][0]] + cache[nb[i][1]]) >> 1;
tp = p[band][nnz];
} while (++i < n_coeffs);
return i;
}
static int decode_coeffs_b_8bpp(VP9Context *s, int16_t *coef, int n_coeffs,
unsigned (*cnt)[6][3], unsigned (*eob)[6][2],
uint8_t (*p)[6][11], int nnz, const int16_t *scan,
const int16_t (*nb)[2], const int16_t *band_counts,
const int16_t *qmul)
{
return decode_coeffs_b_generic(&s->c, coef, n_coeffs, 0, 1, 8, cnt, eob, p,
nnz, scan, nb, band_counts, qmul);
}
static int decode_coeffs_b32_8bpp(VP9Context *s, int16_t *coef, int n_coeffs,
unsigned (*cnt)[6][3], unsigned (*eob)[6][2],
uint8_t (*p)[6][11], int nnz, const int16_t *scan,
const int16_t (*nb)[2], const int16_t *band_counts,
const int16_t *qmul)
{
return decode_coeffs_b_generic(&s->c, coef, n_coeffs, 1, 1, 8, cnt, eob, p,
nnz, scan, nb, band_counts, qmul);
}
static int decode_coeffs_b_16bpp(VP9Context *s, int16_t *coef, int n_coeffs,
unsigned (*cnt)[6][3], unsigned (*eob)[6][2],
uint8_t (*p)[6][11], int nnz, const int16_t *scan,
const int16_t (*nb)[2], const int16_t *band_counts,
const int16_t *qmul)
{
return decode_coeffs_b_generic(&s->c, coef, n_coeffs, 0, 0, s->s.h.bpp, cnt, eob, p,
nnz, scan, nb, band_counts, qmul);
}
static int decode_coeffs_b32_16bpp(VP9Context *s, int16_t *coef, int n_coeffs,
unsigned (*cnt)[6][3], unsigned (*eob)[6][2],
uint8_t (*p)[6][11], int nnz, const int16_t *scan,
const int16_t (*nb)[2], const int16_t *band_counts,
const int16_t *qmul)
{
return decode_coeffs_b_generic(&s->c, coef, n_coeffs, 1, 0, s->s.h.bpp, cnt, eob, p,
nnz, scan, nb, band_counts, qmul);
}
static av_always_inline int decode_coeffs(AVCodecContext *ctx, int is8bitsperpixel)
{
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
int row = s->row, col = s->col;
uint8_t (*p)[6][11] = s->prob.coef[b->tx][0 /* y */][!b->intra];
unsigned (*c)[6][3] = s->counts.coef[b->tx][0 /* y */][!b->intra];
unsigned (*e)[6][2] = s->counts.eob[b->tx][0 /* y */][!b->intra];
int w4 = bwh_tab[1][b->bs][0] << 1, h4 = bwh_tab[1][b->bs][1] << 1;
int end_x = FFMIN(2 * (s->cols - col), w4);
int end_y = FFMIN(2 * (s->rows - row), h4);
int n, pl, x, y, res;
int16_t (*qmul)[2] = s->s.h.segmentation.feat[b->seg_id].qmul;
int tx = 4 * s->s.h.lossless + b->tx;
const int16_t * const *yscans = vp9_scans[tx];
const int16_t (* const *ynbs)[2] = vp9_scans_nb[tx];
const int16_t *uvscan = vp9_scans[b->uvtx][DCT_DCT];
const int16_t (*uvnb)[2] = vp9_scans_nb[b->uvtx][DCT_DCT];
uint8_t *a = &s->above_y_nnz_ctx[col * 2];
uint8_t *l = &s->left_y_nnz_ctx[(row & 7) << 1];
static const int16_t band_counts[4][8] = {
{ 1, 2, 3, 4, 3, 16 - 13 },
{ 1, 2, 3, 4, 11, 64 - 21 },
{ 1, 2, 3, 4, 11, 256 - 21 },
{ 1, 2, 3, 4, 11, 1024 - 21 },
};
const int16_t *y_band_counts = band_counts[b->tx];
const int16_t *uv_band_counts = band_counts[b->uvtx];
int bytesperpixel = is8bitsperpixel ? 1 : 2;
int total_coeff = 0;
#define MERGE(la, end, step, rd) \
for (n = 0; n < end; n += step) \
la[n] = !!rd(&la[n])
#define MERGE_CTX(step, rd) \
do { \
MERGE(l, end_y, step, rd); \
MERGE(a, end_x, step, rd); \
} while (0)
#define DECODE_Y_COEF_LOOP(step, mode_index, v) \
for (n = 0, y = 0; y < end_y; y += step) { \
for (x = 0; x < end_x; x += step, n += step * step) { \
enum TxfmType txtp = vp9_intra_txfm_type[b->mode[mode_index]]; \
res = (is8bitsperpixel ? decode_coeffs_b##v##_8bpp : decode_coeffs_b##v##_16bpp) \
(s, s->block + 16 * n * bytesperpixel, 16 * step * step, \
c, e, p, a[x] + l[y], yscans[txtp], \
ynbs[txtp], y_band_counts, qmul[0]); \
a[x] = l[y] = !!res; \
total_coeff |= !!res; \
if (step >= 4) { \
AV_WN16A(&s->eob[n], res); \
} else { \
s->eob[n] = res; \
} \
} \
}
#define SPLAT(la, end, step, cond) \
if (step == 2) { \
for (n = 1; n < end; n += step) \
la[n] = la[n - 1]; \
} else if (step == 4) { \
if (cond) { \
for (n = 0; n < end; n += step) \
AV_WN32A(&la[n], la[n] * 0x01010101); \
} else { \
for (n = 0; n < end; n += step) \
memset(&la[n + 1], la[n], FFMIN(end - n - 1, 3)); \
} \
} else /* step == 8 */ { \
if (cond) { \
if (HAVE_FAST_64BIT) { \
for (n = 0; n < end; n += step) \
AV_WN64A(&la[n], la[n] * 0x0101010101010101ULL); \
} else { \
for (n = 0; n < end; n += step) { \
uint32_t v32 = la[n] * 0x01010101; \
AV_WN32A(&la[n], v32); \
AV_WN32A(&la[n + 4], v32); \
} \
} \
} else { \
for (n = 0; n < end; n += step) \
memset(&la[n + 1], la[n], FFMIN(end - n - 1, 7)); \
} \
}
#define SPLAT_CTX(step) \
do { \
SPLAT(a, end_x, step, end_x == w4); \
SPLAT(l, end_y, step, end_y == h4); \
} while (0)
/* y tokens */
switch (b->tx) {
case TX_4X4:
DECODE_Y_COEF_LOOP(1, b->bs > BS_8x8 ? n : 0,);
break;
case TX_8X8:
MERGE_CTX(2, AV_RN16A);
DECODE_Y_COEF_LOOP(2, 0,);
SPLAT_CTX(2);
break;
case TX_16X16:
MERGE_CTX(4, AV_RN32A);
DECODE_Y_COEF_LOOP(4, 0,);
SPLAT_CTX(4);
break;
case TX_32X32:
MERGE_CTX(8, AV_RN64A);
DECODE_Y_COEF_LOOP(8, 0, 32);
SPLAT_CTX(8);
break;
}
#define DECODE_UV_COEF_LOOP(step, v) \
for (n = 0, y = 0; y < end_y; y += step) { \
for (x = 0; x < end_x; x += step, n += step * step) { \
res = (is8bitsperpixel ? decode_coeffs_b##v##_8bpp : decode_coeffs_b##v##_16bpp) \
(s, s->uvblock[pl] + 16 * n * bytesperpixel, \
16 * step * step, c, e, p, a[x] + l[y], \
uvscan, uvnb, uv_band_counts, qmul[1]); \
a[x] = l[y] = !!res; \
total_coeff |= !!res; \
if (step >= 4) { \
AV_WN16A(&s->uveob[pl][n], res); \
} else { \
s->uveob[pl][n] = res; \
} \
} \
}
p = s->prob.coef[b->uvtx][1 /* uv */][!b->intra];
c = s->counts.coef[b->uvtx][1 /* uv */][!b->intra];
e = s->counts.eob[b->uvtx][1 /* uv */][!b->intra];
w4 >>= s->ss_h;
end_x >>= s->ss_h;
h4 >>= s->ss_v;
end_y >>= s->ss_v;
for (pl = 0; pl < 2; pl++) {
a = &s->above_uv_nnz_ctx[pl][col << !s->ss_h];
l = &s->left_uv_nnz_ctx[pl][(row & 7) << !s->ss_v];
switch (b->uvtx) {
case TX_4X4:
DECODE_UV_COEF_LOOP(1,);
break;
case TX_8X8:
MERGE_CTX(2, AV_RN16A);
DECODE_UV_COEF_LOOP(2,);
SPLAT_CTX(2);
break;
case TX_16X16:
MERGE_CTX(4, AV_RN32A);
DECODE_UV_COEF_LOOP(4,);
SPLAT_CTX(4);
break;
case TX_32X32:
MERGE_CTX(8, AV_RN64A);
DECODE_UV_COEF_LOOP(8, 32);
SPLAT_CTX(8);
break;
}
}
return total_coeff;
}
static int decode_coeffs_8bpp(AVCodecContext *ctx)
{
return decode_coeffs(ctx, 1);
}
static int decode_coeffs_16bpp(AVCodecContext *ctx)
{
return decode_coeffs(ctx, 0);
}
static av_always_inline int check_intra_mode(VP9Context *s, int mode, uint8_t **a,
uint8_t *dst_edge, ptrdiff_t stride_edge,
uint8_t *dst_inner, ptrdiff_t stride_inner,
uint8_t *l, int col, int x, int w,
int row, int y, enum TxfmMode tx,
int p, int ss_h, int ss_v, int bytesperpixel)
{
int have_top = row > 0 || y > 0;
int have_left = col > s->tile_col_start || x > 0;
int have_right = x < w - 1;
int bpp = s->s.h.bpp;
static const uint8_t mode_conv[10][2 /* have_left */][2 /* have_top */] = {
[VERT_PRED] = { { DC_127_PRED, VERT_PRED },
{ DC_127_PRED, VERT_PRED } },
[HOR_PRED] = { { DC_129_PRED, DC_129_PRED },
{ HOR_PRED, HOR_PRED } },
[DC_PRED] = { { DC_128_PRED, TOP_DC_PRED },
{ LEFT_DC_PRED, DC_PRED } },
[DIAG_DOWN_LEFT_PRED] = { { DC_127_PRED, DIAG_DOWN_LEFT_PRED },
{ DC_127_PRED, DIAG_DOWN_LEFT_PRED } },
[DIAG_DOWN_RIGHT_PRED] = { { DIAG_DOWN_RIGHT_PRED, DIAG_DOWN_RIGHT_PRED },
{ DIAG_DOWN_RIGHT_PRED, DIAG_DOWN_RIGHT_PRED } },
[VERT_RIGHT_PRED] = { { VERT_RIGHT_PRED, VERT_RIGHT_PRED },
{ VERT_RIGHT_PRED, VERT_RIGHT_PRED } },
[HOR_DOWN_PRED] = { { HOR_DOWN_PRED, HOR_DOWN_PRED },
{ HOR_DOWN_PRED, HOR_DOWN_PRED } },
[VERT_LEFT_PRED] = { { DC_127_PRED, VERT_LEFT_PRED },
{ DC_127_PRED, VERT_LEFT_PRED } },
[HOR_UP_PRED] = { { DC_129_PRED, DC_129_PRED },
{ HOR_UP_PRED, HOR_UP_PRED } },
[TM_VP8_PRED] = { { DC_129_PRED, VERT_PRED },
{ HOR_PRED, TM_VP8_PRED } },
};
static const struct {
uint8_t needs_left:1;
uint8_t needs_top:1;
uint8_t needs_topleft:1;
uint8_t needs_topright:1;
uint8_t invert_left:1;
} edges[N_INTRA_PRED_MODES] = {
[VERT_PRED] = { .needs_top = 1 },
[HOR_PRED] = { .needs_left = 1 },
[DC_PRED] = { .needs_top = 1, .needs_left = 1 },
[DIAG_DOWN_LEFT_PRED] = { .needs_top = 1, .needs_topright = 1 },
[DIAG_DOWN_RIGHT_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 },
[VERT_RIGHT_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 },
[HOR_DOWN_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 },
[VERT_LEFT_PRED] = { .needs_top = 1, .needs_topright = 1 },
[HOR_UP_PRED] = { .needs_left = 1, .invert_left = 1 },
[TM_VP8_PRED] = { .needs_left = 1, .needs_top = 1, .needs_topleft = 1 },
[LEFT_DC_PRED] = { .needs_left = 1 },
[TOP_DC_PRED] = { .needs_top = 1 },
[DC_128_PRED] = { 0 },
[DC_127_PRED] = { 0 },
[DC_129_PRED] = { 0 }
};
av_assert2(mode >= 0 && mode < 10);
mode = mode_conv[mode][have_left][have_top];
if (edges[mode].needs_top) {
uint8_t *top, *topleft;
int n_px_need = 4 << tx, n_px_have = (((s->cols - col) << !ss_h) - x) * 4;
int n_px_need_tr = 0;
if (tx == TX_4X4 && edges[mode].needs_topright && have_right)
n_px_need_tr = 4;
// if top of sb64-row, use s->intra_pred_data[] instead of
// dst[-stride] for intra prediction (it contains pre- instead of
// post-loopfilter data)
if (have_top) {
top = !(row & 7) && !y ?
s->intra_pred_data[p] + (col * (8 >> ss_h) + x * 4) * bytesperpixel :
y == 0 ? &dst_edge[-stride_edge] : &dst_inner[-stride_inner];
if (have_left)
topleft = !(row & 7) && !y ?
s->intra_pred_data[p] + (col * (8 >> ss_h) + x * 4) * bytesperpixel :
y == 0 || x == 0 ? &dst_edge[-stride_edge] :
&dst_inner[-stride_inner];
}
if (have_top &&
(!edges[mode].needs_topleft || (have_left && top == topleft)) &&
(tx != TX_4X4 || !edges[mode].needs_topright || have_right) &&
n_px_need + n_px_need_tr <= n_px_have) {
*a = top;
} else {
if (have_top) {
if (n_px_need <= n_px_have) {
memcpy(*a, top, n_px_need * bytesperpixel);
} else {
#define memset_bpp(c, i1, v, i2, num) do { \
if (bytesperpixel == 1) { \
memset(&(c)[(i1)], (v)[(i2)], (num)); \
} else { \
int n, val = AV_RN16A(&(v)[(i2) * 2]); \
for (n = 0; n < (num); n++) { \
AV_WN16A(&(c)[((i1) + n) * 2], val); \
} \
} \
} while (0)
memcpy(*a, top, n_px_have * bytesperpixel);
memset_bpp(*a, n_px_have, (*a), n_px_have - 1, n_px_need - n_px_have);
}
} else {
#define memset_val(c, val, num) do { \
if (bytesperpixel == 1) { \
memset((c), (val), (num)); \
} else { \
int n; \
for (n = 0; n < (num); n++) { \
AV_WN16A(&(c)[n * 2], (val)); \
} \
} \
} while (0)
memset_val(*a, (128 << (bpp - 8)) - 1, n_px_need);
}
if (edges[mode].needs_topleft) {
if (have_left && have_top) {
#define assign_bpp(c, i1, v, i2) do { \
if (bytesperpixel == 1) { \
(c)[(i1)] = (v)[(i2)]; \
} else { \
AV_COPY16(&(c)[(i1) * 2], &(v)[(i2) * 2]); \
} \
} while (0)
assign_bpp(*a, -1, topleft, -1);
} else {
#define assign_val(c, i, v) do { \
if (bytesperpixel == 1) { \
(c)[(i)] = (v); \
} else { \
AV_WN16A(&(c)[(i) * 2], (v)); \
} \
} while (0)
assign_val((*a), -1, (128 << (bpp - 8)) + (have_top ? +1 : -1));
}
}
if (tx == TX_4X4 && edges[mode].needs_topright) {
if (have_top && have_right &&
n_px_need + n_px_need_tr <= n_px_have) {
memcpy(&(*a)[4 * bytesperpixel], &top[4 * bytesperpixel], 4 * bytesperpixel);
} else {
memset_bpp(*a, 4, *a, 3, 4);
}
}
}
}
if (edges[mode].needs_left) {
if (have_left) {
int n_px_need = 4 << tx, i, n_px_have = (((s->rows - row) << !ss_v) - y) * 4;
uint8_t *dst = x == 0 ? dst_edge : dst_inner;
ptrdiff_t stride = x == 0 ? stride_edge : stride_inner;
if (edges[mode].invert_left) {
if (n_px_need <= n_px_have) {
for (i = 0; i < n_px_need; i++)
assign_bpp(l, i, &dst[i * stride], -1);
} else {
for (i = 0; i < n_px_have; i++)
assign_bpp(l, i, &dst[i * stride], -1);
memset_bpp(l, n_px_have, l, n_px_have - 1, n_px_need - n_px_have);
}
} else {
if (n_px_need <= n_px_have) {
for (i = 0; i < n_px_need; i++)
assign_bpp(l, n_px_need - 1 - i, &dst[i * stride], -1);
} else {
for (i = 0; i < n_px_have; i++)
assign_bpp(l, n_px_need - 1 - i, &dst[i * stride], -1);
memset_bpp(l, 0, l, n_px_need - n_px_have, n_px_need - n_px_have);
}
}
} else {
memset_val(l, (128 << (bpp - 8)) + 1, 4 << tx);
}
}
return mode;
}
static av_always_inline void intra_recon(AVCodecContext *ctx, ptrdiff_t y_off,
ptrdiff_t uv_off, int bytesperpixel)
{
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
int row = s->row, col = s->col;
int w4 = bwh_tab[1][b->bs][0] << 1, step1d = 1 << b->tx, n;
int h4 = bwh_tab[1][b->bs][1] << 1, x, y, step = 1 << (b->tx * 2);
int end_x = FFMIN(2 * (s->cols - col), w4);
int end_y = FFMIN(2 * (s->rows - row), h4);
int tx = 4 * s->s.h.lossless + b->tx, uvtx = b->uvtx + 4 * s->s.h.lossless;
int uvstep1d = 1 << b->uvtx, p;
uint8_t *dst = s->dst[0], *dst_r = s->s.frames[CUR_FRAME].tf.f->data[0] + y_off;
LOCAL_ALIGNED_32(uint8_t, a_buf, [96]);
LOCAL_ALIGNED_32(uint8_t, l, [64]);
for (n = 0, y = 0; y < end_y; y += step1d) {
uint8_t *ptr = dst, *ptr_r = dst_r;
for (x = 0; x < end_x; x += step1d, ptr += 4 * step1d * bytesperpixel,
ptr_r += 4 * step1d * bytesperpixel, n += step) {
int mode = b->mode[b->bs > BS_8x8 && b->tx == TX_4X4 ?
y * 2 + x : 0];
uint8_t *a = &a_buf[32];
enum TxfmType txtp = vp9_intra_txfm_type[mode];
int eob = b->skip ? 0 : b->tx > TX_8X8 ? AV_RN16A(&s->eob[n]) : s->eob[n];
mode = check_intra_mode(s, mode, &a, ptr_r,
s->s.frames[CUR_FRAME].tf.f->linesize[0],
ptr, s->y_stride, l,
col, x, w4, row, y, b->tx, 0, 0, 0, bytesperpixel);
s->dsp.intra_pred[b->tx][mode](ptr, s->y_stride, l, a);
if (eob)
s->dsp.itxfm_add[tx][txtp](ptr, s->y_stride,
s->block + 16 * n * bytesperpixel, eob);
}
dst_r += 4 * step1d * s->s.frames[CUR_FRAME].tf.f->linesize[0];
dst += 4 * step1d * s->y_stride;
}
// U/V
w4 >>= s->ss_h;
end_x >>= s->ss_h;
end_y >>= s->ss_v;
step = 1 << (b->uvtx * 2);
for (p = 0; p < 2; p++) {
dst = s->dst[1 + p];
dst_r = s->s.frames[CUR_FRAME].tf.f->data[1 + p] + uv_off;
for (n = 0, y = 0; y < end_y; y += uvstep1d) {
uint8_t *ptr = dst, *ptr_r = dst_r;
for (x = 0; x < end_x; x += uvstep1d, ptr += 4 * uvstep1d * bytesperpixel,
ptr_r += 4 * uvstep1d * bytesperpixel, n += step) {
int mode = b->uvmode;
uint8_t *a = &a_buf[32];
int eob = b->skip ? 0 : b->uvtx > TX_8X8 ? AV_RN16A(&s->uveob[p][n]) : s->uveob[p][n];
mode = check_intra_mode(s, mode, &a, ptr_r,
s->s.frames[CUR_FRAME].tf.f->linesize[1],
ptr, s->uv_stride, l, col, x, w4, row, y,
b->uvtx, p + 1, s->ss_h, s->ss_v, bytesperpixel);
s->dsp.intra_pred[b->uvtx][mode](ptr, s->uv_stride, l, a);
if (eob)
s->dsp.itxfm_add[uvtx][DCT_DCT](ptr, s->uv_stride,
s->uvblock[p] + 16 * n * bytesperpixel, eob);
}
dst_r += 4 * uvstep1d * s->s.frames[CUR_FRAME].tf.f->linesize[1];
dst += 4 * uvstep1d * s->uv_stride;
}
}
}
static void intra_recon_8bpp(AVCodecContext *ctx, ptrdiff_t y_off, ptrdiff_t uv_off)
{
intra_recon(ctx, y_off, uv_off, 1);
}
static void intra_recon_16bpp(AVCodecContext *ctx, ptrdiff_t y_off, ptrdiff_t uv_off)
{
intra_recon(ctx, y_off, uv_off, 2);
}
static av_always_inline void mc_luma_unscaled(VP9Context *s, vp9_mc_func (*mc)[2],
uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *ref, ptrdiff_t ref_stride,
ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x, const VP56mv *mv,
int bw, int bh, int w, int h, int bytesperpixel)
{
int mx = mv->x, my = mv->y, th;
y += my >> 3;
x += mx >> 3;
ref += y * ref_stride + x * bytesperpixel;
mx &= 7;
my &= 7;
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
// we use +7 because the last 7 pixels of each sbrow can be changed in
// the longest loopfilter of the next sbrow
th = (y + bh + 4 * !!my + 7) >> 6;
ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
aarch64: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. These are ported from the ARM version; it is essentially a 1:1 port with no extra added features, but with some hand tuning (especially for the plain copy/avg functions). The ARM version isn't very register starved to begin with, so there's not much to be gained from having more spare registers here - we only avoid having to clobber callee-saved registers. Examples of runtimes vs the 32 bit version, on a Cortex A53: ARM AArch64 vp9_avg4_neon: 27.2 23.7 vp9_avg8_neon: 56.5 54.7 vp9_avg16_neon: 169.9 167.4 vp9_avg32_neon: 585.8 585.2 vp9_avg64_neon: 2460.3 2294.7 vp9_avg_8tap_smooth_4h_neon: 132.7 125.2 vp9_avg_8tap_smooth_4hv_neon: 478.8 442.0 vp9_avg_8tap_smooth_4v_neon: 126.0 93.7 vp9_avg_8tap_smooth_8h_neon: 241.7 234.2 vp9_avg_8tap_smooth_8hv_neon: 690.9 646.5 vp9_avg_8tap_smooth_8v_neon: 245.0 205.5 vp9_avg_8tap_smooth_64h_neon: 11273.2 11280.1 vp9_avg_8tap_smooth_64hv_neon: 22980.6 22184.1 vp9_avg_8tap_smooth_64v_neon: 11549.7 10781.1 vp9_put4_neon: 18.0 17.2 vp9_put8_neon: 40.2 37.7 vp9_put16_neon: 97.4 99.5 vp9_put32_neon/armv8: 346.0 307.4 vp9_put64_neon/armv8: 1319.0 1107.5 vp9_put_8tap_smooth_4h_neon: 126.7 118.2 vp9_put_8tap_smooth_4hv_neon: 465.7 434.0 vp9_put_8tap_smooth_4v_neon: 113.0 86.5 vp9_put_8tap_smooth_8h_neon: 229.7 221.6 vp9_put_8tap_smooth_8hv_neon: 658.9 621.3 vp9_put_8tap_smooth_8v_neon: 215.0 187.5 vp9_put_8tap_smooth_64h_neon: 10636.7 10627.8 vp9_put_8tap_smooth_64hv_neon: 21076.8 21026.9 vp9_put_8tap_smooth_64v_neon: 9635.0 9632.4 These are generally about as fast as the corresponding ARM routines on the same CPU (at least on the A53), in most cases marginally faster. The speedup vs C code is pretty much the same as for the 32 bit case; on the A53 it's around 6-13x for ther larger 8tap filters. The exact speedup varies a little, since the C versions generally don't end up exactly as slow/fast as on 32 bit. This is an adapted cherry-pick from libav commit 383d96aa2229f644d9bd77b821ed3a309da5e9fc. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// The arm/aarch64 _hv filters read one more row than what actually is
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// needed, so switch to emulated edge one pixel sooner vertically
// (!!my * 5) than horizontally (!!mx * 4).
if (x < !!mx * 3 || y < !!my * 3 ||
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref - !!my * 3 * ref_stride - !!mx * 3 * bytesperpixel,
160, ref_stride,
bw + !!mx * 7, bh + !!my * 7,
x - !!mx * 3, y - !!my * 3, w, h);
ref = s->edge_emu_buffer + !!my * 3 * 160 + !!mx * 3 * bytesperpixel;
ref_stride = 160;
}
mc[!!mx][!!my](dst, dst_stride, ref, ref_stride, bh, mx << 1, my << 1);
}
static av_always_inline void mc_chroma_unscaled(VP9Context *s, vp9_mc_func (*mc)[2],
uint8_t *dst_u, uint8_t *dst_v,
ptrdiff_t dst_stride,
const uint8_t *ref_u, ptrdiff_t src_stride_u,
const uint8_t *ref_v, ptrdiff_t src_stride_v,
ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x, const VP56mv *mv,
int bw, int bh, int w, int h, int bytesperpixel)
{
int mx = mv->x * (1 << !s->ss_h), my = mv->y * (1 << !s->ss_v), th;
y += my >> 4;
x += mx >> 4;
ref_u += y * src_stride_u + x * bytesperpixel;
ref_v += y * src_stride_v + x * bytesperpixel;
mx &= 15;
my &= 15;
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
// we use +7 because the last 7 pixels of each sbrow can be changed in
// the longest loopfilter of the next sbrow
th = (y + bh + 4 * !!my + 7) >> (6 - s->ss_v);
ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
aarch64: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. These are ported from the ARM version; it is essentially a 1:1 port with no extra added features, but with some hand tuning (especially for the plain copy/avg functions). The ARM version isn't very register starved to begin with, so there's not much to be gained from having more spare registers here - we only avoid having to clobber callee-saved registers. Examples of runtimes vs the 32 bit version, on a Cortex A53: ARM AArch64 vp9_avg4_neon: 27.2 23.7 vp9_avg8_neon: 56.5 54.7 vp9_avg16_neon: 169.9 167.4 vp9_avg32_neon: 585.8 585.2 vp9_avg64_neon: 2460.3 2294.7 vp9_avg_8tap_smooth_4h_neon: 132.7 125.2 vp9_avg_8tap_smooth_4hv_neon: 478.8 442.0 vp9_avg_8tap_smooth_4v_neon: 126.0 93.7 vp9_avg_8tap_smooth_8h_neon: 241.7 234.2 vp9_avg_8tap_smooth_8hv_neon: 690.9 646.5 vp9_avg_8tap_smooth_8v_neon: 245.0 205.5 vp9_avg_8tap_smooth_64h_neon: 11273.2 11280.1 vp9_avg_8tap_smooth_64hv_neon: 22980.6 22184.1 vp9_avg_8tap_smooth_64v_neon: 11549.7 10781.1 vp9_put4_neon: 18.0 17.2 vp9_put8_neon: 40.2 37.7 vp9_put16_neon: 97.4 99.5 vp9_put32_neon/armv8: 346.0 307.4 vp9_put64_neon/armv8: 1319.0 1107.5 vp9_put_8tap_smooth_4h_neon: 126.7 118.2 vp9_put_8tap_smooth_4hv_neon: 465.7 434.0 vp9_put_8tap_smooth_4v_neon: 113.0 86.5 vp9_put_8tap_smooth_8h_neon: 229.7 221.6 vp9_put_8tap_smooth_8hv_neon: 658.9 621.3 vp9_put_8tap_smooth_8v_neon: 215.0 187.5 vp9_put_8tap_smooth_64h_neon: 10636.7 10627.8 vp9_put_8tap_smooth_64hv_neon: 21076.8 21026.9 vp9_put_8tap_smooth_64v_neon: 9635.0 9632.4 These are generally about as fast as the corresponding ARM routines on the same CPU (at least on the A53), in most cases marginally faster. The speedup vs C code is pretty much the same as for the 32 bit case; on the A53 it's around 6-13x for ther larger 8tap filters. The exact speedup varies a little, since the C versions generally don't end up exactly as slow/fast as on 32 bit. This is an adapted cherry-pick from libav commit 383d96aa2229f644d9bd77b821ed3a309da5e9fc. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// The arm/aarch64 _hv filters read one more row than what actually is
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// needed, so switch to emulated edge one pixel sooner vertically
// (!!my * 5) than horizontally (!!mx * 4).
if (x < !!mx * 3 || y < !!my * 3 ||
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_u - !!my * 3 * src_stride_u - !!mx * 3 * bytesperpixel,
160, src_stride_u,
bw + !!mx * 7, bh + !!my * 7,
x - !!mx * 3, y - !!my * 3, w, h);
ref_u = s->edge_emu_buffer + !!my * 3 * 160 + !!mx * 3 * bytesperpixel;
mc[!!mx][!!my](dst_u, dst_stride, ref_u, 160, bh, mx, my);
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_v - !!my * 3 * src_stride_v - !!mx * 3 * bytesperpixel,
160, src_stride_v,
bw + !!mx * 7, bh + !!my * 7,
x - !!mx * 3, y - !!my * 3, w, h);
ref_v = s->edge_emu_buffer + !!my * 3 * 160 + !!mx * 3 * bytesperpixel;
mc[!!mx][!!my](dst_v, dst_stride, ref_v, 160, bh, mx, my);
} else {
mc[!!mx][!!my](dst_u, dst_stride, ref_u, src_stride_u, bh, mx, my);
mc[!!mx][!!my](dst_v, dst_stride, ref_v, src_stride_v, bh, mx, my);
}
}
#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, \
px, py, pw, ph, bw, bh, w, h, i) \
mc_luma_unscaled(s, s->dsp.mc, dst, dst_ls, src, src_ls, tref, row, col, \
mv, bw, bh, w, h, bytesperpixel)
#define mc_chroma_dir(s, mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
row, col, mv, px, py, pw, ph, bw, bh, w, h, i) \
mc_chroma_unscaled(s, s->dsp.mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
row, col, mv, bw, bh, w, h, bytesperpixel)
#define SCALED 0
#define FN(x) x##_8bpp
#define BYTES_PER_PIXEL 1
#include "vp9_mc_template.c"
#undef FN
#undef BYTES_PER_PIXEL
#define FN(x) x##_16bpp
#define BYTES_PER_PIXEL 2
#include "vp9_mc_template.c"
#undef mc_luma_dir
#undef mc_chroma_dir
#undef FN
#undef BYTES_PER_PIXEL
#undef SCALED
static av_always_inline void mc_luma_scaled(VP9Context *s, vp9_scaled_mc_func smc,
vp9_mc_func (*mc)[2],
uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *ref, ptrdiff_t ref_stride,
ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x, const VP56mv *in_mv,
int px, int py, int pw, int ph,
int bw, int bh, int w, int h, int bytesperpixel,
const uint16_t *scale, const uint8_t *step)
{
if (s->s.frames[CUR_FRAME].tf.f->width == ref_frame->f->width &&
s->s.frames[CUR_FRAME].tf.f->height == ref_frame->f->height) {
mc_luma_unscaled(s, mc, dst, dst_stride, ref, ref_stride, ref_frame,
y, x, in_mv, bw, bh, w, h, bytesperpixel);
} else {
#define scale_mv(n, dim) (((int64_t)(n) * scale[dim]) >> 14)
int mx, my;
int refbw_m1, refbh_m1;
int th;
VP56mv mv;
mv.x = av_clip(in_mv->x, -(x + pw - px + 4) * 8, (s->cols * 8 - x + px + 3) * 8);
mv.y = av_clip(in_mv->y, -(y + ph - py + 4) * 8, (s->rows * 8 - y + py + 3) * 8);
// BUG libvpx seems to scale the two components separately. This introduces
// rounding errors but we have to reproduce them to be exactly compatible
// with the output from libvpx...
mx = scale_mv(mv.x * 2, 0) + scale_mv(x * 16, 0);
my = scale_mv(mv.y * 2, 1) + scale_mv(y * 16, 1);
y = my >> 4;
x = mx >> 4;
ref += y * ref_stride + x * bytesperpixel;
mx &= 15;
my &= 15;
refbw_m1 = ((bw - 1) * step[0] + mx) >> 4;
refbh_m1 = ((bh - 1) * step[1] + my) >> 4;
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
// we use +7 because the last 7 pixels of each sbrow can be changed in
// the longest loopfilter of the next sbrow
th = (y + refbh_m1 + 4 + 7) >> 6;
ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
aarch64: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. These are ported from the ARM version; it is essentially a 1:1 port with no extra added features, but with some hand tuning (especially for the plain copy/avg functions). The ARM version isn't very register starved to begin with, so there's not much to be gained from having more spare registers here - we only avoid having to clobber callee-saved registers. Examples of runtimes vs the 32 bit version, on a Cortex A53: ARM AArch64 vp9_avg4_neon: 27.2 23.7 vp9_avg8_neon: 56.5 54.7 vp9_avg16_neon: 169.9 167.4 vp9_avg32_neon: 585.8 585.2 vp9_avg64_neon: 2460.3 2294.7 vp9_avg_8tap_smooth_4h_neon: 132.7 125.2 vp9_avg_8tap_smooth_4hv_neon: 478.8 442.0 vp9_avg_8tap_smooth_4v_neon: 126.0 93.7 vp9_avg_8tap_smooth_8h_neon: 241.7 234.2 vp9_avg_8tap_smooth_8hv_neon: 690.9 646.5 vp9_avg_8tap_smooth_8v_neon: 245.0 205.5 vp9_avg_8tap_smooth_64h_neon: 11273.2 11280.1 vp9_avg_8tap_smooth_64hv_neon: 22980.6 22184.1 vp9_avg_8tap_smooth_64v_neon: 11549.7 10781.1 vp9_put4_neon: 18.0 17.2 vp9_put8_neon: 40.2 37.7 vp9_put16_neon: 97.4 99.5 vp9_put32_neon/armv8: 346.0 307.4 vp9_put64_neon/armv8: 1319.0 1107.5 vp9_put_8tap_smooth_4h_neon: 126.7 118.2 vp9_put_8tap_smooth_4hv_neon: 465.7 434.0 vp9_put_8tap_smooth_4v_neon: 113.0 86.5 vp9_put_8tap_smooth_8h_neon: 229.7 221.6 vp9_put_8tap_smooth_8hv_neon: 658.9 621.3 vp9_put_8tap_smooth_8v_neon: 215.0 187.5 vp9_put_8tap_smooth_64h_neon: 10636.7 10627.8 vp9_put_8tap_smooth_64hv_neon: 21076.8 21026.9 vp9_put_8tap_smooth_64v_neon: 9635.0 9632.4 These are generally about as fast as the corresponding ARM routines on the same CPU (at least on the A53), in most cases marginally faster. The speedup vs C code is pretty much the same as for the 32 bit case; on the A53 it's around 6-13x for ther larger 8tap filters. The exact speedup varies a little, since the C versions generally don't end up exactly as slow/fast as on 32 bit. This is an adapted cherry-pick from libav commit 383d96aa2229f644d9bd77b821ed3a309da5e9fc. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// The arm/aarch64 _hv filters read one more row than what actually is
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// needed, so switch to emulated edge one pixel sooner vertically
// (y + 5 >= h - refbh_m1) than horizontally (x + 4 >= w - refbw_m1).
if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 5 >= h - refbh_m1) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref - 3 * ref_stride - 3 * bytesperpixel,
288, ref_stride,
refbw_m1 + 8, refbh_m1 + 8,
x - 3, y - 3, w, h);
ref = s->edge_emu_buffer + 3 * 288 + 3 * bytesperpixel;
ref_stride = 288;
}
smc(dst, dst_stride, ref, ref_stride, bh, mx, my, step[0], step[1]);
}
}
static av_always_inline void mc_chroma_scaled(VP9Context *s, vp9_scaled_mc_func smc,
vp9_mc_func (*mc)[2],
uint8_t *dst_u, uint8_t *dst_v,
ptrdiff_t dst_stride,
const uint8_t *ref_u, ptrdiff_t src_stride_u,
const uint8_t *ref_v, ptrdiff_t src_stride_v,
ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x, const VP56mv *in_mv,
int px, int py, int pw, int ph,
int bw, int bh, int w, int h, int bytesperpixel,
const uint16_t *scale, const uint8_t *step)
{
if (s->s.frames[CUR_FRAME].tf.f->width == ref_frame->f->width &&
s->s.frames[CUR_FRAME].tf.f->height == ref_frame->f->height) {
mc_chroma_unscaled(s, mc, dst_u, dst_v, dst_stride, ref_u, src_stride_u,
ref_v, src_stride_v, ref_frame,
y, x, in_mv, bw, bh, w, h, bytesperpixel);
} else {
int mx, my;
int refbw_m1, refbh_m1;
int th;
VP56mv mv;
if (s->ss_h) {
// BUG https://code.google.com/p/webm/issues/detail?id=820
mv.x = av_clip(in_mv->x, -(x + pw - px + 4) * 16, (s->cols * 4 - x + px + 3) * 16);
mx = scale_mv(mv.x, 0) + (scale_mv(x * 16, 0) & ~15) + (scale_mv(x * 32, 0) & 15);
} else {
mv.x = av_clip(in_mv->x, -(x + pw - px + 4) * 8, (s->cols * 8 - x + px + 3) * 8);
mx = scale_mv(mv.x * 2, 0) + scale_mv(x * 16, 0);
}
if (s->ss_v) {
// BUG https://code.google.com/p/webm/issues/detail?id=820
mv.y = av_clip(in_mv->y, -(y + ph - py + 4) * 16, (s->rows * 4 - y + py + 3) * 16);
my = scale_mv(mv.y, 1) + (scale_mv(y * 16, 1) & ~15) + (scale_mv(y * 32, 1) & 15);
} else {
mv.y = av_clip(in_mv->y, -(y + ph - py + 4) * 8, (s->rows * 8 - y + py + 3) * 8);
my = scale_mv(mv.y * 2, 1) + scale_mv(y * 16, 1);
}
#undef scale_mv
y = my >> 4;
x = mx >> 4;
ref_u += y * src_stride_u + x * bytesperpixel;
ref_v += y * src_stride_v + x * bytesperpixel;
mx &= 15;
my &= 15;
refbw_m1 = ((bw - 1) * step[0] + mx) >> 4;
refbh_m1 = ((bh - 1) * step[1] + my) >> 4;
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
// we use +7 because the last 7 pixels of each sbrow can be changed in
// the longest loopfilter of the next sbrow
th = (y + refbh_m1 + 4 + 7) >> (6 - s->ss_v);
ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
aarch64: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. These are ported from the ARM version; it is essentially a 1:1 port with no extra added features, but with some hand tuning (especially for the plain copy/avg functions). The ARM version isn't very register starved to begin with, so there's not much to be gained from having more spare registers here - we only avoid having to clobber callee-saved registers. Examples of runtimes vs the 32 bit version, on a Cortex A53: ARM AArch64 vp9_avg4_neon: 27.2 23.7 vp9_avg8_neon: 56.5 54.7 vp9_avg16_neon: 169.9 167.4 vp9_avg32_neon: 585.8 585.2 vp9_avg64_neon: 2460.3 2294.7 vp9_avg_8tap_smooth_4h_neon: 132.7 125.2 vp9_avg_8tap_smooth_4hv_neon: 478.8 442.0 vp9_avg_8tap_smooth_4v_neon: 126.0 93.7 vp9_avg_8tap_smooth_8h_neon: 241.7 234.2 vp9_avg_8tap_smooth_8hv_neon: 690.9 646.5 vp9_avg_8tap_smooth_8v_neon: 245.0 205.5 vp9_avg_8tap_smooth_64h_neon: 11273.2 11280.1 vp9_avg_8tap_smooth_64hv_neon: 22980.6 22184.1 vp9_avg_8tap_smooth_64v_neon: 11549.7 10781.1 vp9_put4_neon: 18.0 17.2 vp9_put8_neon: 40.2 37.7 vp9_put16_neon: 97.4 99.5 vp9_put32_neon/armv8: 346.0 307.4 vp9_put64_neon/armv8: 1319.0 1107.5 vp9_put_8tap_smooth_4h_neon: 126.7 118.2 vp9_put_8tap_smooth_4hv_neon: 465.7 434.0 vp9_put_8tap_smooth_4v_neon: 113.0 86.5 vp9_put_8tap_smooth_8h_neon: 229.7 221.6 vp9_put_8tap_smooth_8hv_neon: 658.9 621.3 vp9_put_8tap_smooth_8v_neon: 215.0 187.5 vp9_put_8tap_smooth_64h_neon: 10636.7 10627.8 vp9_put_8tap_smooth_64hv_neon: 21076.8 21026.9 vp9_put_8tap_smooth_64v_neon: 9635.0 9632.4 These are generally about as fast as the corresponding ARM routines on the same CPU (at least on the A53), in most cases marginally faster. The speedup vs C code is pretty much the same as for the 32 bit case; on the A53 it's around 6-13x for ther larger 8tap filters. The exact speedup varies a little, since the C versions generally don't end up exactly as slow/fast as on 32 bit. This is an adapted cherry-pick from libav commit 383d96aa2229f644d9bd77b821ed3a309da5e9fc. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// The arm/aarch64 _hv filters read one more row than what actually is
arm: vp9: Add NEON optimizations of VP9 MC functions This work is sponsored by, and copyright, Google. The filter coefficients are signed values, where the product of the multiplication with one individual filter coefficient doesn't overflow a 16 bit signed value (the largest filter coefficient is 127). But when the products are accumulated, the resulting sum can overflow the 16 bit signed range. Instead of accumulating in 32 bit, we accumulate the largest product (either index 3 or 4) last with a saturated addition. (The VP8 MC asm does something similar, but slightly simpler, by accumulating each half of the filter separately. In the VP9 MC filters, each half of the filter can also overflow though, so the largest component has to be handled individually.) Examples of relative speedup compared to the C version, from checkasm: Cortex A7 A8 A9 A53 vp9_avg4_neon: 1.71 1.15 1.42 1.49 vp9_avg8_neon: 2.51 3.63 3.14 2.58 vp9_avg16_neon: 2.95 6.76 3.01 2.84 vp9_avg32_neon: 3.29 6.64 2.85 3.00 vp9_avg64_neon: 3.47 6.67 3.14 2.80 vp9_avg_8tap_smooth_4h_neon: 3.22 4.73 2.76 4.67 vp9_avg_8tap_smooth_4hv_neon: 3.67 4.76 3.28 4.71 vp9_avg_8tap_smooth_4v_neon: 5.52 7.60 4.60 6.31 vp9_avg_8tap_smooth_8h_neon: 6.22 9.04 5.12 9.32 vp9_avg_8tap_smooth_8hv_neon: 6.38 8.21 5.72 8.17 vp9_avg_8tap_smooth_8v_neon: 9.22 12.66 8.15 11.10 vp9_avg_8tap_smooth_64h_neon: 7.02 10.23 5.54 11.58 vp9_avg_8tap_smooth_64hv_neon: 6.76 9.46 5.93 9.40 vp9_avg_8tap_smooth_64v_neon: 10.76 14.13 9.46 13.37 vp9_put4_neon: 1.11 1.47 1.00 1.21 vp9_put8_neon: 1.23 2.17 1.94 1.48 vp9_put16_neon: 1.63 4.02 1.73 1.97 vp9_put32_neon: 1.56 4.92 2.00 1.96 vp9_put64_neon: 2.10 5.28 2.03 2.35 vp9_put_8tap_smooth_4h_neon: 3.11 4.35 2.63 4.35 vp9_put_8tap_smooth_4hv_neon: 3.67 4.69 3.25 4.71 vp9_put_8tap_smooth_4v_neon: 5.45 7.27 4.49 6.52 vp9_put_8tap_smooth_8h_neon: 5.97 8.18 4.81 8.56 vp9_put_8tap_smooth_8hv_neon: 6.39 7.90 5.64 8.15 vp9_put_8tap_smooth_8v_neon: 9.03 11.84 8.07 11.51 vp9_put_8tap_smooth_64h_neon: 6.78 9.48 4.88 10.89 vp9_put_8tap_smooth_64hv_neon: 6.99 8.87 5.94 9.56 vp9_put_8tap_smooth_64v_neon: 10.69 13.30 9.43 14.34 For the larger 8tap filters, the speedup vs C code is around 5-14x. This is significantly faster than libvpx's implementation of the same functions, at least when comparing the put_8tap_smooth_64 functions (compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from libvpx). Absolute runtimes from checkasm: Cortex A7 A8 A9 A53 vp9_put_8tap_smooth_64h_neon: 20150.3 14489.4 19733.6 10863.7 libvpx vpx_convolve8_horiz_neon: 52623.3 19736.4 21907.7 25027.7 vp9_put_8tap_smooth_64v_neon: 14455.0 12303.9 13746.4 9628.9 libvpx vpx_convolve8_vert_neon: 42090.0 17706.2 17659.9 16941.2 Thus, on the A9, the horizontal filter is only marginally faster than libvpx, while our version is significantly faster on the other cores, and the vertical filter is significantly faster on all cores. The difference is especially large on the A7. The libvpx implementation does the accumulation in 32 bit, which probably explains most of the differences. This is an adapted cherry-pick from libav commits ffbd1d2b0002576ef0d976a41ff959c635373fdc, 392caa65df3efa8b2d48a80f08a6af4892c61c08, 557c1675cf0e803b2fee43b4c8b58433842c84d0 and 11623217e3c9b859daee544e31acdd0821b61039. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
8 years ago
// needed, so switch to emulated edge one pixel sooner vertically
// (y + 5 >= h - refbh_m1) than horizontally (x + 4 >= w - refbw_m1).
if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 5 >= h - refbh_m1) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_u - 3 * src_stride_u - 3 * bytesperpixel,
288, src_stride_u,
refbw_m1 + 8, refbh_m1 + 8,
x - 3, y - 3, w, h);
ref_u = s->edge_emu_buffer + 3 * 288 + 3 * bytesperpixel;
smc(dst_u, dst_stride, ref_u, 288, bh, mx, my, step[0], step[1]);
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_v - 3 * src_stride_v - 3 * bytesperpixel,
288, src_stride_v,
refbw_m1 + 8, refbh_m1 + 8,
x - 3, y - 3, w, h);
ref_v = s->edge_emu_buffer + 3 * 288 + 3 * bytesperpixel;
smc(dst_v, dst_stride, ref_v, 288, bh, mx, my, step[0], step[1]);
} else {
smc(dst_u, dst_stride, ref_u, src_stride_u, bh, mx, my, step[0], step[1]);
smc(dst_v, dst_stride, ref_v, src_stride_v, bh, mx, my, step[0], step[1]);
}
}
}
#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, \
px, py, pw, ph, bw, bh, w, h, i) \
mc_luma_scaled(s, s->dsp.s##mc, s->dsp.mc, dst, dst_ls, src, src_ls, tref, row, col, \
mv, px, py, pw, ph, bw, bh, w, h, bytesperpixel, \
s->mvscale[b->ref[i]], s->mvstep[b->ref[i]])
#define mc_chroma_dir(s, mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
row, col, mv, px, py, pw, ph, bw, bh, w, h, i) \
mc_chroma_scaled(s, s->dsp.s##mc, s->dsp.mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
row, col, mv, px, py, pw, ph, bw, bh, w, h, bytesperpixel, \
s->mvscale[b->ref[i]], s->mvstep[b->ref[i]])
#define SCALED 1
#define FN(x) x##_scaled_8bpp
#define BYTES_PER_PIXEL 1
#include "vp9_mc_template.c"
#undef FN
#undef BYTES_PER_PIXEL
#define FN(x) x##_scaled_16bpp
#define BYTES_PER_PIXEL 2
#include "vp9_mc_template.c"
#undef mc_luma_dir
#undef mc_chroma_dir
#undef FN
#undef BYTES_PER_PIXEL
#undef SCALED
static av_always_inline void inter_recon(AVCodecContext *ctx, int bytesperpixel)
{
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
int row = s->row, col = s->col;
if (s->mvscale[b->ref[0]][0] || (b->comp && s->mvscale[b->ref[1]][0])) {
if (bytesperpixel == 1) {
inter_pred_scaled_8bpp(ctx);
} else {
inter_pred_scaled_16bpp(ctx);
}
} else {
if (bytesperpixel == 1) {
inter_pred_8bpp(ctx);
} else {
inter_pred_16bpp(ctx);
}
}
if (!b->skip) {
/* mostly copied intra_recon() */
int w4 = bwh_tab[1][b->bs][0] << 1, step1d = 1 << b->tx, n;
int h4 = bwh_tab[1][b->bs][1] << 1, x, y, step = 1 << (b->tx * 2);
int end_x = FFMIN(2 * (s->cols - col), w4);
int end_y = FFMIN(2 * (s->rows - row), h4);
int tx = 4 * s->s.h.lossless + b->tx, uvtx = b->uvtx + 4 * s->s.h.lossless;
int uvstep1d = 1 << b->uvtx, p;
uint8_t *dst = s->dst[0];
// y itxfm add
for (n = 0, y = 0; y < end_y; y += step1d) {
uint8_t *ptr = dst;
for (x = 0; x < end_x; x += step1d,
ptr += 4 * step1d * bytesperpixel, n += step) {
int eob = b->tx > TX_8X8 ? AV_RN16A(&s->eob[n]) : s->eob[n];
if (eob)
s->dsp.itxfm_add[tx][DCT_DCT](ptr, s->y_stride,
s->block + 16 * n * bytesperpixel, eob);
}
dst += 4 * s->y_stride * step1d;
}
// uv itxfm add
end_x >>= s->ss_h;
end_y >>= s->ss_v;
step = 1 << (b->uvtx * 2);
for (p = 0; p < 2; p++) {
dst = s->dst[p + 1];
for (n = 0, y = 0; y < end_y; y += uvstep1d) {
uint8_t *ptr = dst;
for (x = 0; x < end_x; x += uvstep1d,
ptr += 4 * uvstep1d * bytesperpixel, n += step) {
int eob = b->uvtx > TX_8X8 ? AV_RN16A(&s->uveob[p][n]) : s->uveob[p][n];
if (eob)
s->dsp.itxfm_add[uvtx][DCT_DCT](ptr, s->uv_stride,
s->uvblock[p] + 16 * n * bytesperpixel, eob);
}
dst += 4 * uvstep1d * s->uv_stride;
}
}
}
}
static void inter_recon_8bpp(AVCodecContext *ctx)
{
inter_recon(ctx, 1);
}
static void inter_recon_16bpp(AVCodecContext *ctx)
{
inter_recon(ctx, 2);
}
static av_always_inline void mask_edges(uint8_t (*mask)[8][4], int ss_h, int ss_v,
int row_and_7, int col_and_7,
int w, int h, int col_end, int row_end,
enum TxfmMode tx, int skip_inter)
{
static const unsigned wide_filter_col_mask[2] = { 0x11, 0x01 };
static const unsigned wide_filter_row_mask[2] = { 0x03, 0x07 };
// FIXME I'm pretty sure all loops can be replaced by a single LUT if
// we make VP9Filter.mask uint64_t (i.e. row/col all single variable)
// and make the LUT 5-indexed (bl, bp, is_uv, tx and row/col), and then
// use row_and_7/col_and_7 as shifts (1*col_and_7+8*row_and_7)
// the intended behaviour of the vp9 loopfilter is to work on 8-pixel
// edges. This means that for UV, we work on two subsampled blocks at
// a time, and we only use the topleft block's mode information to set
// things like block strength. Thus, for any block size smaller than
// 16x16, ignore the odd portion of the block.
if (tx == TX_4X4 && (ss_v | ss_h)) {
if (h == ss_v) {
if (row_and_7 & 1)
return;
if (!row_end)
h += 1;
}
if (w == ss_h) {
if (col_and_7 & 1)
return;
if (!col_end)
w += 1;
}
}
if (tx == TX_4X4 && !skip_inter) {
int t = 1 << col_and_7, m_col = (t << w) - t, y;
// on 32-px edges, use the 8-px wide loopfilter; else, use 4-px wide
int m_row_8 = m_col & wide_filter_col_mask[ss_h], m_row_4 = m_col - m_row_8;
for (y = row_and_7; y < h + row_and_7; y++) {
int col_mask_id = 2 - !(y & wide_filter_row_mask[ss_v]);
mask[0][y][1] |= m_row_8;
mask[0][y][2] |= m_row_4;
// for odd lines, if the odd col is not being filtered,
// skip odd row also:
// .---. <-- a
// | |
// |___| <-- b
// ^ ^
// c d
//
// if a/c are even row/col and b/d are odd, and d is skipped,
// e.g. right edge of size-66x66.webm, then skip b also (bug)
if ((ss_h & ss_v) && (col_end & 1) && (y & 1)) {
mask[1][y][col_mask_id] |= (t << (w - 1)) - t;
} else {
mask[1][y][col_mask_id] |= m_col;
}
if (!ss_h)
mask[0][y][3] |= m_col;
if (!ss_v) {
if (ss_h && (col_end & 1))
mask[1][y][3] |= (t << (w - 1)) - t;
else
mask[1][y][3] |= m_col;
}
}
} else {
int y, t = 1 << col_and_7, m_col = (t << w) - t;
if (!skip_inter) {
int mask_id = (tx == TX_8X8);
static const unsigned masks[4] = { 0xff, 0x55, 0x11, 0x01 };
int l2 = tx + ss_h - 1, step1d;
int m_row = m_col & masks[l2];
// at odd UV col/row edges tx16/tx32 loopfilter edges, force
// 8wd loopfilter to prevent going off the visible edge.
if (ss_h && tx > TX_8X8 && (w ^ (w - 1)) == 1) {
int m_row_16 = ((t << (w - 1)) - t) & masks[l2];
int m_row_8 = m_row - m_row_16;
for (y = row_and_7; y < h + row_and_7; y++) {
mask[0][y][0] |= m_row_16;
mask[0][y][1] |= m_row_8;
}
} else {
for (y = row_and_7; y < h + row_and_7; y++)
mask[0][y][mask_id] |= m_row;
}
l2 = tx + ss_v - 1;
step1d = 1 << l2;
if (ss_v && tx > TX_8X8 && (h ^ (h - 1)) == 1) {
for (y = row_and_7; y < h + row_and_7 - 1; y += step1d)
mask[1][y][0] |= m_col;
if (y - row_and_7 == h - 1)
mask[1][y][1] |= m_col;
} else {
for (y = row_and_7; y < h + row_and_7; y += step1d)
mask[1][y][mask_id] |= m_col;
}
} else if (tx != TX_4X4) {
int mask_id;
mask_id = (tx == TX_8X8) || (h == ss_v);
mask[1][row_and_7][mask_id] |= m_col;
mask_id = (tx == TX_8X8) || (w == ss_h);
for (y = row_and_7; y < h + row_and_7; y++)
mask[0][y][mask_id] |= t;
} else {
int t8 = t & wide_filter_col_mask[ss_h], t4 = t - t8;
for (y = row_and_7; y < h + row_and_7; y++) {
mask[0][y][2] |= t4;
mask[0][y][1] |= t8;
}
mask[1][row_and_7][2 - !(row_and_7 & wide_filter_row_mask[ss_v])] |= m_col;
}
}
}
static void decode_b(AVCodecContext *ctx, int row, int col,
struct VP9Filter *lflvl, ptrdiff_t yoff, ptrdiff_t uvoff,
enum BlockLevel bl, enum BlockPartition bp)
{
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
enum BlockSize bs = bl * 3 + bp;
int bytesperpixel = s->bytesperpixel;
int w4 = bwh_tab[1][bs][0], h4 = bwh_tab[1][bs][1], lvl;
int emu[2];
AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
s->row = row;
s->row7 = row & 7;
s->col = col;
s->col7 = col & 7;
s->min_mv.x = -(128 + col * 64);
s->min_mv.y = -(128 + row * 64);
s->max_mv.x = 128 + (s->cols - col - w4) * 64;
s->max_mv.y = 128 + (s->rows - row - h4) * 64;
if (s->pass < 2) {
b->bs = bs;
b->bl = bl;
b->bp = bp;
decode_mode(ctx);
b->uvtx = b->tx - ((s->ss_h && w4 * 2 == (1 << b->tx)) ||
(s->ss_v && h4 * 2 == (1 << b->tx)));
if (!b->skip) {
int has_coeffs;
if (bytesperpixel == 1) {
has_coeffs = decode_coeffs_8bpp(ctx);
} else {
has_coeffs = decode_coeffs_16bpp(ctx);
}
if (!has_coeffs && b->bs <= BS_8x8 && !b->intra) {
b->skip = 1;
memset(&s->above_skip_ctx[col], 1, w4);
memset(&s->left_skip_ctx[s->row7], 1, h4);
}
} else {
int row7 = s->row7;
#define SPLAT_ZERO_CTX(v, n) \
switch (n) { \
case 1: v = 0; break; \
case 2: AV_ZERO16(&v); break; \
case 4: AV_ZERO32(&v); break; \
case 8: AV_ZERO64(&v); break; \
case 16: AV_ZERO128(&v); break; \
}
#define SPLAT_ZERO_YUV(dir, var, off, n, dir2) \
do { \
SPLAT_ZERO_CTX(s->dir##_y_##var[off * 2], n * 2); \
if (s->ss_##dir2) { \
SPLAT_ZERO_CTX(s->dir##_uv_##var[0][off], n); \
SPLAT_ZERO_CTX(s->dir##_uv_##var[1][off], n); \
} else { \
SPLAT_ZERO_CTX(s->dir##_uv_##var[0][off * 2], n * 2); \
SPLAT_ZERO_CTX(s->dir##_uv_##var[1][off * 2], n * 2); \
} \
} while (0)
switch (w4) {
case 1: SPLAT_ZERO_YUV(above, nnz_ctx, col, 1, h); break;
case 2: SPLAT_ZERO_YUV(above, nnz_ctx, col, 2, h); break;
case 4: SPLAT_ZERO_YUV(above, nnz_ctx, col, 4, h); break;
case 8: SPLAT_ZERO_YUV(above, nnz_ctx, col, 8, h); break;
}
switch (h4) {
case 1: SPLAT_ZERO_YUV(left, nnz_ctx, row7, 1, v); break;
case 2: SPLAT_ZERO_YUV(left, nnz_ctx, row7, 2, v); break;
case 4: SPLAT_ZERO_YUV(left, nnz_ctx, row7, 4, v); break;
case 8: SPLAT_ZERO_YUV(left, nnz_ctx, row7, 8, v); break;
}
}
if (s->pass == 1) {
s->b++;
s->block += w4 * h4 * 64 * bytesperpixel;
s->uvblock[0] += w4 * h4 * 64 * bytesperpixel >> (s->ss_h + s->ss_v);
s->uvblock[1] += w4 * h4 * 64 * bytesperpixel >> (s->ss_h + s->ss_v);
s->eob += 4 * w4 * h4;
s->uveob[0] += 4 * w4 * h4 >> (s->ss_h + s->ss_v);
s->uveob[1] += 4 * w4 * h4 >> (s->ss_h + s->ss_v);
return;
}
}
// emulated overhangs if the stride of the target buffer can't hold. This
// makes it possible to support emu-edge and so on even if we have large block
// overhangs
emu[0] = (col + w4) * 8 * bytesperpixel > f->linesize[0] ||
(row + h4) > s->rows;
emu[1] = ((col + w4) * 8 >> s->ss_h) * bytesperpixel > f->linesize[1] ||
(row + h4) > s->rows;
if (emu[0]) {
s->dst[0] = s->tmp_y;
s->y_stride = 128;
} else {
s->dst[0] = f->data[0] + yoff;
s->y_stride = f->linesize[0];
}
if (emu[1]) {
s->dst[1] = s->tmp_uv[0];
s->dst[2] = s->tmp_uv[1];
s->uv_stride = 128;
} else {
s->dst[1] = f->data[1] + uvoff;
s->dst[2] = f->data[2] + uvoff;
s->uv_stride = f->linesize[1];
}
if (b->intra) {
if (s->s.h.bpp > 8) {
intra_recon_16bpp(ctx, yoff, uvoff);
} else {
intra_recon_8bpp(ctx, yoff, uvoff);
}
} else {
if (s->s.h.bpp > 8) {
inter_recon_16bpp(ctx);
} else {
inter_recon_8bpp(ctx);
}
}
if (emu[0]) {
int w = FFMIN(s->cols - col, w4) * 8, h = FFMIN(s->rows - row, h4) * 8, n, o = 0;
for (n = 0; o < w; n++) {
int bw = 64 >> n;
av_assert2(n <= 4);
if (w & bw) {
s->dsp.mc[n][0][0][0][0](f->data[0] + yoff + o * bytesperpixel, f->linesize[0],
s->tmp_y + o * bytesperpixel, 128, h, 0, 0);
o += bw;
}
}
}
if (emu[1]) {
int w = FFMIN(s->cols - col, w4) * 8 >> s->ss_h;
int h = FFMIN(s->rows - row, h4) * 8 >> s->ss_v, n, o = 0;
for (n = s->ss_h; o < w; n++) {
int bw = 64 >> n;
av_assert2(n <= 4);
if (w & bw) {
s->dsp.mc[n][0][0][0][0](f->data[1] + uvoff + o * bytesperpixel, f->linesize[1],
s->tmp_uv[0] + o * bytesperpixel, 128, h, 0, 0);
s->dsp.mc[n][0][0][0][0](f->data[2] + uvoff + o * bytesperpixel, f->linesize[2],
s->tmp_uv[1] + o * bytesperpixel, 128, h, 0, 0);
o += bw;
}
}
}
// pick filter level and find edges to apply filter to
if (s->s.h.filter.level &&
(lvl = s->s.h.segmentation.feat[b->seg_id].lflvl[b->intra ? 0 : b->ref[0] + 1]
[b->mode[3] != ZEROMV]) > 0) {
int x_end = FFMIN(s->cols - col, w4), y_end = FFMIN(s->rows - row, h4);
int skip_inter = !b->intra && b->skip, col7 = s->col7, row7 = s->row7;
setctx_2d(&lflvl->level[row7 * 8 + col7], w4, h4, 8, lvl);
mask_edges(lflvl->mask[0], 0, 0, row7, col7, x_end, y_end, 0, 0, b->tx, skip_inter);
if (s->ss_h || s->ss_v)
mask_edges(lflvl->mask[1], s->ss_h, s->ss_v, row7, col7, x_end, y_end,
s->cols & 1 && col + w4 >= s->cols ? s->cols & 7 : 0,
s->rows & 1 && row + h4 >= s->rows ? s->rows & 7 : 0,
b->uvtx, skip_inter);
if (!s->filter_lut.lim_lut[lvl]) {
int sharp = s->s.h.filter.sharpness;
int limit = lvl;
if (sharp > 0) {
limit >>= (sharp + 3) >> 2;
limit = FFMIN(limit, 9 - sharp);
}
limit = FFMAX(limit, 1);
s->filter_lut.lim_lut[lvl] = limit;
s->filter_lut.mblim_lut[lvl] = 2 * (lvl + 2) + limit;
}
}
if (s->pass == 2) {
s->b++;
s->block += w4 * h4 * 64 * bytesperpixel;
s->uvblock[0] += w4 * h4 * 64 * bytesperpixel >> (s->ss_v + s->ss_h);
s->uvblock[1] += w4 * h4 * 64 * bytesperpixel >> (s->ss_v + s->ss_h);
s->eob += 4 * w4 * h4;
s->uveob[0] += 4 * w4 * h4 >> (s->ss_v + s->ss_h);
s->uveob[1] += 4 * w4 * h4 >> (s->ss_v + s->ss_h);
}
}
static void decode_sb(AVCodecContext *ctx, int row, int col, struct VP9Filter *lflvl,
ptrdiff_t yoff, ptrdiff_t uvoff, enum BlockLevel bl)
{
VP9Context *s = ctx->priv_data;
int c = ((s->above_partition_ctx[col] >> (3 - bl)) & 1) |
(((s->left_partition_ctx[row & 0x7] >> (3 - bl)) & 1) << 1);
const uint8_t *p = s->s.h.keyframe || s->s.h.intraonly ? vp9_default_kf_partition_probs[bl][c] :
s->prob.p.partition[bl][c];
enum BlockPartition bp;
ptrdiff_t hbs = 4 >> bl;
AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
ptrdiff_t y_stride = f->linesize[0], uv_stride = f->linesize[1];
int bytesperpixel = s->bytesperpixel;
if (bl == BL_8X8) {
bp = vp8_rac_get_tree(&s->c, vp9_partition_tree, p);
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
} else if (col + hbs < s->cols) { // FIXME why not <=?
if (row + hbs < s->rows) { // FIXME why not <=?
bp = vp8_rac_get_tree(&s->c, vp9_partition_tree, p);
switch (bp) {
case PARTITION_NONE:
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
break;
case PARTITION_H:
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_b(ctx, row + hbs, col, lflvl, yoff, uvoff, bl, bp);
break;
case PARTITION_V:
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
yoff += hbs * 8 * bytesperpixel;
uvoff += hbs * 8 * bytesperpixel >> s->ss_h;
decode_b(ctx, row, col + hbs, lflvl, yoff, uvoff, bl, bp);
break;
case PARTITION_SPLIT:
decode_sb(ctx, row, col, lflvl, yoff, uvoff, bl + 1);
decode_sb(ctx, row, col + hbs, lflvl,
yoff + 8 * hbs * bytesperpixel,
uvoff + (8 * hbs * bytesperpixel >> s->ss_h), bl + 1);
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_sb(ctx, row + hbs, col, lflvl, yoff, uvoff, bl + 1);
decode_sb(ctx, row + hbs, col + hbs, lflvl,
yoff + 8 * hbs * bytesperpixel,
uvoff + (8 * hbs * bytesperpixel >> s->ss_h), bl + 1);
break;
default:
av_assert0(0);
}
} else if (vp56_rac_get_prob_branchy(&s->c, p[1])) {
bp = PARTITION_SPLIT;
decode_sb(ctx, row, col, lflvl, yoff, uvoff, bl + 1);
decode_sb(ctx, row, col + hbs, lflvl,
yoff + 8 * hbs * bytesperpixel,
uvoff + (8 * hbs * bytesperpixel >> s->ss_h), bl + 1);
} else {
bp = PARTITION_H;
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
}
} else if (row + hbs < s->rows) { // FIXME why not <=?
if (vp56_rac_get_prob_branchy(&s->c, p[2])) {
bp = PARTITION_SPLIT;
decode_sb(ctx, row, col, lflvl, yoff, uvoff, bl + 1);
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_sb(ctx, row + hbs, col, lflvl, yoff, uvoff, bl + 1);
} else {
bp = PARTITION_V;
decode_b(ctx, row, col, lflvl, yoff, uvoff, bl, bp);
}
} else {
bp = PARTITION_SPLIT;
decode_sb(ctx, row, col, lflvl, yoff, uvoff, bl + 1);
}
s->counts.partition[bl][c][bp]++;
}
static void decode_sb_mem(AVCodecContext *ctx, int row, int col, struct VP9Filter *lflvl,
ptrdiff_t yoff, ptrdiff_t uvoff, enum BlockLevel bl)
{
VP9Context *s = ctx->priv_data;
VP9Block *b = s->b;
ptrdiff_t hbs = 4 >> bl;
AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
ptrdiff_t y_stride = f->linesize[0], uv_stride = f->linesize[1];
int bytesperpixel = s->bytesperpixel;
if (bl == BL_8X8) {
av_assert2(b->bl == BL_8X8);
decode_b(ctx, row, col, lflvl, yoff, uvoff, b->bl, b->bp);
} else if (s->b->bl == bl) {
decode_b(ctx, row, col, lflvl, yoff, uvoff, b->bl, b->bp);
if (b->bp == PARTITION_H && row + hbs < s->rows) {
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_b(ctx, row + hbs, col, lflvl, yoff, uvoff, b->bl, b->bp);
} else if (b->bp == PARTITION_V && col + hbs < s->cols) {
yoff += hbs * 8 * bytesperpixel;
uvoff += hbs * 8 * bytesperpixel >> s->ss_h;
decode_b(ctx, row, col + hbs, lflvl, yoff, uvoff, b->bl, b->bp);
}
} else {
decode_sb_mem(ctx, row, col, lflvl, yoff, uvoff, bl + 1);
if (col + hbs < s->cols) { // FIXME why not <=?
if (row + hbs < s->rows) {
decode_sb_mem(ctx, row, col + hbs, lflvl, yoff + 8 * hbs * bytesperpixel,
uvoff + (8 * hbs * bytesperpixel >> s->ss_h), bl + 1);
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_sb_mem(ctx, row + hbs, col, lflvl, yoff, uvoff, bl + 1);
decode_sb_mem(ctx, row + hbs, col + hbs, lflvl,
yoff + 8 * hbs * bytesperpixel,
uvoff + (8 * hbs * bytesperpixel >> s->ss_h), bl + 1);
} else {
yoff += hbs * 8 * bytesperpixel;
uvoff += hbs * 8 * bytesperpixel >> s->ss_h;
decode_sb_mem(ctx, row, col + hbs, lflvl, yoff, uvoff, bl + 1);
}
} else if (row + hbs < s->rows) {
yoff += hbs * 8 * y_stride;
uvoff += hbs * 8 * uv_stride >> s->ss_v;
decode_sb_mem(ctx, row + hbs, col, lflvl, yoff, uvoff, bl + 1);
}
}
}
static av_always_inline void filter_plane_cols(VP9Context *s, int col, int ss_h, int ss_v,
uint8_t *lvl, uint8_t (*mask)[4],
uint8_t *dst, ptrdiff_t ls)
{
int y, x, bytesperpixel = s->bytesperpixel;
// filter edges between columns (e.g. block1 | block2)
for (y = 0; y < 8; y += 2 << ss_v, dst += 16 * ls, lvl += 16 << ss_v) {
uint8_t *ptr = dst, *l = lvl, *hmask1 = mask[y], *hmask2 = mask[y + 1 + ss_v];
unsigned hm1 = hmask1[0] | hmask1[1] | hmask1[2], hm13 = hmask1[3];
unsigned hm2 = hmask2[1] | hmask2[2], hm23 = hmask2[3];
unsigned hm = hm1 | hm2 | hm13 | hm23;
for (x = 1; hm & ~(x - 1); x <<= 1, ptr += 8 * bytesperpixel >> ss_h) {
if (col || x > 1) {
if (hm1 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (hmask1[0] & x) {
if (hmask2[0] & x) {
av_assert2(l[8 << ss_v] == L);
s->dsp.loop_filter_16[0](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[2][0](ptr, ls, E, I, H);
}
} else if (hm2 & x) {
L = l[8 << ss_v];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[!!(hmask1[1] & x)]
[!!(hmask2[1] & x)]
[0](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[!!(hmask1[1] & x)]
[0](ptr, ls, E, I, H);
}
} else if (hm2 & x) {
int L = l[8 << ss_v], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[!!(hmask2[1] & x)]
[0](ptr + 8 * ls, ls, E, I, H);
}
}
if (ss_h) {
if (x & 0xAA)
l += 2;
} else {
if (hm13 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (hm23 & x) {
L = l[8 << ss_v];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[0][0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
} else {
s->dsp.loop_filter_8[0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
}
} else if (hm23 & x) {
int L = l[8 << ss_v], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[0][0](ptr + 8 * ls + 4 * bytesperpixel, ls, E, I, H);
}
l++;
}
}
}
}
static av_always_inline void filter_plane_rows(VP9Context *s, int row, int ss_h, int ss_v,
uint8_t *lvl, uint8_t (*mask)[4],
uint8_t *dst, ptrdiff_t ls)
{
int y, x, bytesperpixel = s->bytesperpixel;
// block1
// filter edges between rows (e.g. ------)
// block2
for (y = 0; y < 8; y++, dst += 8 * ls >> ss_v) {
uint8_t *ptr = dst, *l = lvl, *vmask = mask[y];
unsigned vm = vmask[0] | vmask[1] | vmask[2], vm3 = vmask[3];
for (x = 1; vm & ~(x - 1); x <<= (2 << ss_h), ptr += 16 * bytesperpixel, l += 2 << ss_h) {
if (row || y) {
if (vm & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (vmask[0] & x) {
if (vmask[0] & (x << (1 + ss_h))) {
av_assert2(l[1 + ss_h] == L);
s->dsp.loop_filter_16[1](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[2][1](ptr, ls, E, I, H);
}
} else if (vm & (x << (1 + ss_h))) {
L = l[1 + ss_h];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[!!(vmask[1] & x)]
[!!(vmask[1] & (x << (1 + ss_h)))]
[1](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[!!(vmask[1] & x)]
[1](ptr, ls, E, I, H);
}
} else if (vm & (x << (1 + ss_h))) {
int L = l[1 + ss_h], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[!!(vmask[1] & (x << (1 + ss_h)))]
[1](ptr + 8 * bytesperpixel, ls, E, I, H);
}
}
if (!ss_v) {
if (vm3 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (vm3 & (x << (1 + ss_h))) {
L = l[1 + ss_h];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[0][0][1](ptr + ls * 4, ls, E, I, H);
} else {
s->dsp.loop_filter_8[0][1](ptr + ls * 4, ls, E, I, H);
}
} else if (vm3 & (x << (1 + ss_h))) {
int L = l[1 + ss_h], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[0][1](ptr + ls * 4 + 8 * bytesperpixel, ls, E, I, H);
}
}
}
if (ss_v) {
if (y & 1)
lvl += 16;
} else {
lvl += 8;
}
}
}
static void loopfilter_sb(AVCodecContext *ctx, struct VP9Filter *lflvl,
int row, int col, ptrdiff_t yoff, ptrdiff_t uvoff)
{
VP9Context *s = ctx->priv_data;
AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
uint8_t *dst = f->data[0] + yoff;
ptrdiff_t ls_y = f->linesize[0], ls_uv = f->linesize[1];
uint8_t (*uv_masks)[8][4] = lflvl->mask[s->ss_h | s->ss_v];
int p;
// FIXME in how far can we interleave the v/h loopfilter calls? E.g.
// if you think of them as acting on a 8x8 block max, we can interleave
// each v/h within the single x loop, but that only works if we work on
// 8 pixel blocks, and we won't always do that (we want at least 16px
// to use SSE2 optimizations, perhaps 32 for AVX2)
filter_plane_cols(s, col, 0, 0, lflvl->level, lflvl->mask[0][0], dst, ls_y);
filter_plane_rows(s, row, 0, 0, lflvl->level, lflvl->mask[0][1], dst, ls_y);
for (p = 0; p < 2; p++) {
dst = f->data[1 + p] + uvoff;
filter_plane_cols(s, col, s->ss_h, s->ss_v, lflvl->level, uv_masks[0], dst, ls_uv);
filter_plane_rows(s, row, s->ss_h, s->ss_v, lflvl->level, uv_masks[1], dst, ls_uv);
}
}
static void set_tile_offset(int *start, int *end, int idx, int log2_n, int n)
{
int sb_start = ( idx * n) >> log2_n;
int sb_end = ((idx + 1) * n) >> log2_n;
*start = FFMIN(sb_start, n) << 3;
*end = FFMIN(sb_end, n) << 3;
}
static av_always_inline void adapt_prob(uint8_t *p, unsigned ct0, unsigned ct1,
int max_count, int update_factor)
{
unsigned ct = ct0 + ct1, p2, p1;
if (!ct)
return;
update_factor = FASTDIV(update_factor * FFMIN(ct, max_count), max_count);
p1 = *p;
p2 = ((((int64_t) ct0) << 8) + (ct >> 1)) / ct;
p2 = av_clip(p2, 1, 255);
// (p1 * (256 - update_factor) + p2 * update_factor + 128) >> 8
*p = p1 + (((p2 - p1) * update_factor + 128) >> 8);
}
static void adapt_probs(VP9Context *s)
{
int i, j, k, l, m;
prob_context *p = &s->prob_ctx[s->s.h.framectxid].p;
int uf = (s->s.h.keyframe || s->s.h.intraonly || !s->last_keyframe) ? 112 : 128;
// coefficients
for (i = 0; i < 4; i++)
for (j = 0; j < 2; j++)
for (k = 0; k < 2; k++)
for (l = 0; l < 6; l++)
for (m = 0; m < 6; m++) {
uint8_t *pp = s->prob_ctx[s->s.h.framectxid].coef[i][j][k][l][m];
unsigned *e = s->counts.eob[i][j][k][l][m];
unsigned *c = s->counts.coef[i][j][k][l][m];
if (l == 0 && m >= 3) // dc only has 3 pt
break;
adapt_prob(&pp[0], e[0], e[1], 24, uf);
adapt_prob(&pp[1], c[0], c[1] + c[2], 24, uf);
adapt_prob(&pp[2], c[1], c[2], 24, uf);
}
if (s->s.h.keyframe || s->s.h.intraonly) {
memcpy(p->skip, s->prob.p.skip, sizeof(p->skip));
memcpy(p->tx32p, s->prob.p.tx32p, sizeof(p->tx32p));
memcpy(p->tx16p, s->prob.p.tx16p, sizeof(p->tx16p));
memcpy(p->tx8p, s->prob.p.tx8p, sizeof(p->tx8p));
return;
}
// skip flag
for (i = 0; i < 3; i++)
adapt_prob(&p->skip[i], s->counts.skip[i][0], s->counts.skip[i][1], 20, 128);
// intra/inter flag
for (i = 0; i < 4; i++)
adapt_prob(&p->intra[i], s->counts.intra[i][0], s->counts.intra[i][1], 20, 128);
// comppred flag
if (s->s.h.comppredmode == PRED_SWITCHABLE) {
for (i = 0; i < 5; i++)
adapt_prob(&p->comp[i], s->counts.comp[i][0], s->counts.comp[i][1], 20, 128);
}
// reference frames
if (s->s.h.comppredmode != PRED_SINGLEREF) {
for (i = 0; i < 5; i++)
adapt_prob(&p->comp_ref[i], s->counts.comp_ref[i][0],
s->counts.comp_ref[i][1], 20, 128);
}
if (s->s.h.comppredmode != PRED_COMPREF) {
for (i = 0; i < 5; i++) {
uint8_t *pp = p->single_ref[i];
unsigned (*c)[2] = s->counts.single_ref[i];
adapt_prob(&pp[0], c[0][0], c[0][1], 20, 128);
adapt_prob(&pp[1], c[1][0], c[1][1], 20, 128);
}
}
// block partitioning
for (i = 0; i < 4; i++)
for (j = 0; j < 4; j++) {
uint8_t *pp = p->partition[i][j];
unsigned *c = s->counts.partition[i][j];
adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
adapt_prob(&pp[2], c[2], c[3], 20, 128);
}
// tx size
if (s->s.h.txfmmode == TX_SWITCHABLE) {
for (i = 0; i < 2; i++) {
unsigned *c16 = s->counts.tx16p[i], *c32 = s->counts.tx32p[i];
adapt_prob(&p->tx8p[i], s->counts.tx8p[i][0], s->counts.tx8p[i][1], 20, 128);
adapt_prob(&p->tx16p[i][0], c16[0], c16[1] + c16[2], 20, 128);
adapt_prob(&p->tx16p[i][1], c16[1], c16[2], 20, 128);
adapt_prob(&p->tx32p[i][0], c32[0], c32[1] + c32[2] + c32[3], 20, 128);
adapt_prob(&p->tx32p[i][1], c32[1], c32[2] + c32[3], 20, 128);
adapt_prob(&p->tx32p[i][2], c32[2], c32[3], 20, 128);
}
}
// interpolation filter
if (s->s.h.filtermode == FILTER_SWITCHABLE) {
for (i = 0; i < 4; i++) {
uint8_t *pp = p->filter[i];
unsigned *c = s->counts.filter[i];
adapt_prob(&pp[0], c[0], c[1] + c[2], 20, 128);
adapt_prob(&pp[1], c[1], c[2], 20, 128);
}
}
// inter modes
for (i = 0; i < 7; i++) {
uint8_t *pp = p->mv_mode[i];
unsigned *c = s->counts.mv_mode[i];
adapt_prob(&pp[0], c[2], c[1] + c[0] + c[3], 20, 128);
adapt_prob(&pp[1], c[0], c[1] + c[3], 20, 128);
adapt_prob(&pp[2], c[1], c[3], 20, 128);
}
// mv joints
{
uint8_t *pp = p->mv_joint;
unsigned *c = s->counts.mv_joint;
adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
adapt_prob(&pp[2], c[2], c[3], 20, 128);
}
// mv components
for (i = 0; i < 2; i++) {
uint8_t *pp;
unsigned *c, (*c2)[2], sum;
adapt_prob(&p->mv_comp[i].sign, s->counts.mv_comp[i].sign[0],
s->counts.mv_comp[i].sign[1], 20, 128);
pp = p->mv_comp[i].classes;
c = s->counts.mv_comp[i].classes;
sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9] + c[10];
adapt_prob(&pp[0], c[0], sum, 20, 128);
sum -= c[1];
adapt_prob(&pp[1], c[1], sum, 20, 128);
sum -= c[2] + c[3];
adapt_prob(&pp[2], c[2] + c[3], sum, 20, 128);
adapt_prob(&pp[3], c[2], c[3], 20, 128);
sum -= c[4] + c[5];
adapt_prob(&pp[4], c[4] + c[5], sum, 20, 128);
adapt_prob(&pp[5], c[4], c[5], 20, 128);
sum -= c[6];
adapt_prob(&pp[6], c[6], sum, 20, 128);
adapt_prob(&pp[7], c[7] + c[8], c[9] + c[10], 20, 128);
adapt_prob(&pp[8], c[7], c[8], 20, 128);
adapt_prob(&pp[9], c[9], c[10], 20, 128);
adapt_prob(&p->mv_comp[i].class0, s->counts.mv_comp[i].class0[0],
s->counts.mv_comp[i].class0[1], 20, 128);
pp = p->mv_comp[i].bits;
c2 = s->counts.mv_comp[i].bits;
for (j = 0; j < 10; j++)
adapt_prob(&pp[j], c2[j][0], c2[j][1], 20, 128);
for (j = 0; j < 2; j++) {
pp = p->mv_comp[i].class0_fp[j];
c = s->counts.mv_comp[i].class0_fp[j];
adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
adapt_prob(&pp[2], c[2], c[3], 20, 128);
}
pp = p->mv_comp[i].fp;
c = s->counts.mv_comp[i].fp;
adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
adapt_prob(&pp[2], c[2], c[3], 20, 128);
if (s->s.h.highprecisionmvs) {
adapt_prob(&p->mv_comp[i].class0_hp, s->counts.mv_comp[i].class0_hp[0],
s->counts.mv_comp[i].class0_hp[1], 20, 128);
adapt_prob(&p->mv_comp[i].hp, s->counts.mv_comp[i].hp[0],
s->counts.mv_comp[i].hp[1], 20, 128);
}
}
// y intra modes
for (i = 0; i < 4; i++) {
uint8_t *pp = p->y_mode[i];
unsigned *c = s->counts.y_mode[i], sum, s2;
sum = c[0] + c[1] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9];
adapt_prob(&pp[0], c[DC_PRED], sum, 20, 128);
sum -= c[TM_VP8_PRED];
adapt_prob(&pp[1], c[TM_VP8_PRED], sum, 20, 128);
sum -= c[VERT_PRED];
adapt_prob(&pp[2], c[VERT_PRED], sum, 20, 128);
s2 = c[HOR_PRED] + c[DIAG_DOWN_RIGHT_PRED] + c[VERT_RIGHT_PRED];
sum -= s2;
adapt_prob(&pp[3], s2, sum, 20, 128);
s2 -= c[HOR_PRED];
adapt_prob(&pp[4], c[HOR_PRED], s2, 20, 128);
adapt_prob(&pp[5], c[DIAG_DOWN_RIGHT_PRED], c[VERT_RIGHT_PRED], 20, 128);
sum -= c[DIAG_DOWN_LEFT_PRED];
adapt_prob(&pp[6], c[DIAG_DOWN_LEFT_PRED], sum, 20, 128);
sum -= c[VERT_LEFT_PRED];
adapt_prob(&pp[7], c[VERT_LEFT_PRED], sum, 20, 128);
adapt_prob(&pp[8], c[HOR_DOWN_PRED], c[HOR_UP_PRED], 20, 128);
}
// uv intra modes
for (i = 0; i < 10; i++) {
uint8_t *pp = p->uv_mode[i];
unsigned *c = s->counts.uv_mode[i], sum, s2;
sum = c[0] + c[1] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9];
adapt_prob(&pp[0], c[DC_PRED], sum, 20, 128);
sum -= c[TM_VP8_PRED];
adapt_prob(&pp[1], c[TM_VP8_PRED], sum, 20, 128);
sum -= c[VERT_PRED];
adapt_prob(&pp[2], c[VERT_PRED], sum, 20, 128);
s2 = c[HOR_PRED] + c[DIAG_DOWN_RIGHT_PRED] + c[VERT_RIGHT_PRED];
sum -= s2;
adapt_prob(&pp[3], s2, sum, 20, 128);
s2 -= c[HOR_PRED];
adapt_prob(&pp[4], c[HOR_PRED], s2, 20, 128);
adapt_prob(&pp[5], c[DIAG_DOWN_RIGHT_PRED], c[VERT_RIGHT_PRED], 20, 128);
sum -= c[DIAG_DOWN_LEFT_PRED];
adapt_prob(&pp[6], c[DIAG_DOWN_LEFT_PRED], sum, 20, 128);
sum -= c[VERT_LEFT_PRED];
adapt_prob(&pp[7], c[VERT_LEFT_PRED], sum, 20, 128);
adapt_prob(&pp[8], c[HOR_DOWN_PRED], c[HOR_UP_PRED], 20, 128);
}
}
static void free_buffers(VP9Context *s)
{
av_freep(&s->intra_pred_data[0]);
av_freep(&s->b_base);
av_freep(&s->block_base);
}
static av_cold int vp9_decode_free(AVCodecContext *ctx)
{
VP9Context *s = ctx->priv_data;
int i;
for (i = 0; i < 3; i++) {
if (s->s.frames[i].tf.f->buf[0])
vp9_unref_frame(ctx, &s->s.frames[i]);
av_frame_free(&s->s.frames[i].tf.f);
}
for (i = 0; i < 8; i++) {
if (s->s.refs[i].f->buf[0])
ff_thread_release_buffer(ctx, &s->s.refs[i]);
av_frame_free(&s->s.refs[i].f);
if (s->next_refs[i].f->buf[0])
ff_thread_release_buffer(ctx, &s->next_refs[i]);
av_frame_free(&s->next_refs[i].f);
}
free_buffers(s);
av_freep(&s->c_b);
s->c_b_size = 0;
return 0;
}
static int vp9_decode_frame(AVCodecContext *ctx, void *frame,
int *got_frame, AVPacket *pkt)
{
const uint8_t *data = pkt->data;
int size = pkt->size;
VP9Context *s = ctx->priv_data;
int res, tile_row, tile_col, i, ref, row, col;
int retain_segmap_ref = s->s.frames[REF_FRAME_SEGMAP].segmentation_map &&
(!s->s.h.segmentation.enabled || !s->s.h.segmentation.update_map);
ptrdiff_t yoff, uvoff, ls_y, ls_uv;
AVFrame *f;
int bytesperpixel;
if ((res = decode_frame_header(ctx, data, size, &ref)) < 0) {
return res;
} else if (res == 0) {
if (!s->s.refs[ref].f->buf[0]) {
av_log(ctx, AV_LOG_ERROR, "Requested reference %d not available\n", ref);
return AVERROR_INVALIDDATA;
}
if ((res = av_frame_ref(frame, s->s.refs[ref].f)) < 0)
return res;
((AVFrame *)frame)->pts = pkt->pts;
#if FF_API_PKT_PTS
FF_DISABLE_DEPRECATION_WARNINGS
((AVFrame *)frame)->pkt_pts = pkt->pts;
FF_ENABLE_DEPRECATION_WARNINGS
#endif
((AVFrame *)frame)->pkt_dts = pkt->dts;
for (i = 0; i < 8; i++) {
if (s->next_refs[i].f->buf[0])
ff_thread_release_buffer(ctx, &s->next_refs[i]);
if (s->s.refs[i].f->buf[0] &&
(res = ff_thread_ref_frame(&s->next_refs[i], &s->s.refs[i])) < 0)
return res;
}
*got_frame = 1;
return pkt->size;
}
data += res;
size -= res;
if (!retain_segmap_ref || s->s.h.keyframe || s->s.h.intraonly) {
if (s->s.frames[REF_FRAME_SEGMAP].tf.f->buf[0])
vp9_unref_frame(ctx, &s->s.frames[REF_FRAME_SEGMAP]);
if (!s->s.h.keyframe && !s->s.h.intraonly && !s->s.h.errorres && s->s.frames[CUR_FRAME].tf.f->buf[0] &&
(res = vp9_ref_frame(ctx, &s->s.frames[REF_FRAME_SEGMAP], &s->s.frames[CUR_FRAME])) < 0)
return res;
}
if (s->s.frames[REF_FRAME_MVPAIR].tf.f->buf[0])
vp9_unref_frame(ctx, &s->s.frames[REF_FRAME_MVPAIR]);
if (!s->s.h.intraonly && !s->s.h.keyframe && !s->s.h.errorres && s->s.frames[CUR_FRAME].tf.f->buf[0] &&
(res = vp9_ref_frame(ctx, &s->s.frames[REF_FRAME_MVPAIR], &s->s.frames[CUR_FRAME])) < 0)
return res;
if (s->s.frames[CUR_FRAME].tf.f->buf[0])
vp9_unref_frame(ctx, &s->s.frames[CUR_FRAME]);
if ((res = vp9_alloc_frame(ctx, &s->s.frames[CUR_FRAME])) < 0)
return res;
f = s->s.frames[CUR_FRAME].tf.f;
f->key_frame = s->s.h.keyframe;
f->pict_type = (s->s.h.keyframe || s->s.h.intraonly) ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
ls_y = f->linesize[0];
ls_uv =f->linesize[1];
if (s->s.frames[REF_FRAME_SEGMAP].tf.f->buf[0] &&
(s->s.frames[REF_FRAME_MVPAIR].tf.f->width != s->s.frames[CUR_FRAME].tf.f->width ||
s->s.frames[REF_FRAME_MVPAIR].tf.f->height != s->s.frames[CUR_FRAME].tf.f->height)) {
vp9_unref_frame(ctx, &s->s.frames[REF_FRAME_SEGMAP]);
}
// ref frame setup
for (i = 0; i < 8; i++) {
if (s->next_refs[i].f->buf[0])
ff_thread_release_buffer(ctx, &s->next_refs[i]);
if (s->s.h.refreshrefmask & (1 << i)) {
res = ff_thread_ref_frame(&s->next_refs[i], &s->s.frames[CUR_FRAME].tf);
} else if (s->s.refs[i].f->buf[0]) {
res = ff_thread_ref_frame(&s->next_refs[i], &s->s.refs[i]);
}
if (res < 0)
return res;
}
if (ctx->hwaccel) {
res = ctx->hwaccel->start_frame(ctx, NULL, 0);
if (res < 0)
return res;
res = ctx->hwaccel->decode_slice(ctx, pkt->data, pkt->size);
if (res < 0)
return res;
res = ctx->hwaccel->end_frame(ctx);
if (res < 0)
return res;
goto finish;
}
// main tile decode loop
bytesperpixel = s->bytesperpixel;
memset(s->above_partition_ctx, 0, s->cols);
memset(s->above_skip_ctx, 0, s->cols);
if (s->s.h.keyframe || s->s.h.intraonly) {
memset(s->above_mode_ctx, DC_PRED, s->cols * 2);
} else {
memset(s->above_mode_ctx, NEARESTMV, s->cols);
}
memset(s->above_y_nnz_ctx, 0, s->sb_cols * 16);
memset(s->above_uv_nnz_ctx[0], 0, s->sb_cols * 16 >> s->ss_h);
memset(s->above_uv_nnz_ctx[1], 0, s->sb_cols * 16 >> s->ss_h);
memset(s->above_segpred_ctx, 0, s->cols);
s->pass = s->s.frames[CUR_FRAME].uses_2pass =
ctx->active_thread_type == FF_THREAD_FRAME && s->s.h.refreshctx && !s->s.h.parallelmode;
if ((res = update_block_buffers(ctx)) < 0) {
av_log(ctx, AV_LOG_ERROR,
"Failed to allocate block buffers\n");
return res;
}
if (s->s.h.refreshctx && s->s.h.parallelmode) {
int j, k, l, m;
for (i = 0; i < 4; i++) {
for (j = 0; j < 2; j++)
for (k = 0; k < 2; k++)
for (l = 0; l < 6; l++)
for (m = 0; m < 6; m++)
memcpy(s->prob_ctx[s->s.h.framectxid].coef[i][j][k][l][m],
s->prob.coef[i][j][k][l][m], 3);
if (s->s.h.txfmmode == i)
break;
}
s->prob_ctx[s->s.h.framectxid].p = s->prob.p;
ff_thread_finish_setup(ctx);
} else if (!s->s.h.refreshctx) {
ff_thread_finish_setup(ctx);
}
do {
yoff = uvoff = 0;
s->b = s->b_base;
s->block = s->block_base;
s->uvblock[0] = s->uvblock_base[0];
s->uvblock[1] = s->uvblock_base[1];
s->eob = s->eob_base;
s->uveob[0] = s->uveob_base[0];
s->uveob[1] = s->uveob_base[1];
for (tile_row = 0; tile_row < s->s.h.tiling.tile_rows; tile_row++) {
set_tile_offset(&s->tile_row_start, &s->tile_row_end,
tile_row, s->s.h.tiling.log2_tile_rows, s->sb_rows);
if (s->pass != 2) {
for (tile_col = 0; tile_col < s->s.h.tiling.tile_cols; tile_col++) {
int64_t tile_size;
if (tile_col == s->s.h.tiling.tile_cols - 1 &&
tile_row == s->s.h.tiling.tile_rows - 1) {
11 years ago
tile_size = size;
} else {
tile_size = AV_RB32(data);
data += 4;
size -= 4;
}
if (tile_size > size) {
ff_thread_report_progress(&s->s.frames[CUR_FRAME].tf, INT_MAX, 0);
11 years ago
return AVERROR_INVALIDDATA;
}
11 years ago
ff_vp56_init_range_decoder(&s->c_b[tile_col], data, tile_size);
if (vp56_rac_get_prob_branchy(&s->c_b[tile_col], 128)) { // marker bit
ff_thread_report_progress(&s->s.frames[CUR_FRAME].tf, INT_MAX, 0);
11 years ago
return AVERROR_INVALIDDATA;
}
11 years ago
data += tile_size;
size -= tile_size;
}
}
for (row = s->tile_row_start; row < s->tile_row_end;
row += 8, yoff += ls_y * 64, uvoff += ls_uv * 64 >> s->ss_v) {
11 years ago
struct VP9Filter *lflvl_ptr = s->lflvl;
ptrdiff_t yoff2 = yoff, uvoff2 = uvoff;
for (tile_col = 0; tile_col < s->s.h.tiling.tile_cols; tile_col++) {
set_tile_offset(&s->tile_col_start, &s->tile_col_end,
tile_col, s->s.h.tiling.log2_tile_cols, s->sb_cols);
if (s->pass != 2) {
11 years ago
memset(s->left_partition_ctx, 0, 8);
memset(s->left_skip_ctx, 0, 8);
if (s->s.h.keyframe || s->s.h.intraonly) {
11 years ago
memset(s->left_mode_ctx, DC_PRED, 16);
} else {
memset(s->left_mode_ctx, NEARESTMV, 8);
}
memset(s->left_y_nnz_ctx, 0, 16);
memset(s->left_uv_nnz_ctx, 0, 32);
11 years ago
memset(s->left_segpred_ctx, 0, 8);
11 years ago
memcpy(&s->c, &s->c_b[tile_col], sizeof(s->c));
}
for (col = s->tile_col_start;
col < s->tile_col_end;
col += 8, yoff2 += 64 * bytesperpixel,
uvoff2 += 64 * bytesperpixel >> s->ss_h, lflvl_ptr++) {
11 years ago
// FIXME integrate with lf code (i.e. zero after each
// use, similar to invtxfm coefficients, or similar)
if (s->pass != 1) {
11 years ago
memset(lflvl_ptr->mask, 0, sizeof(lflvl_ptr->mask));
}
if (s->pass == 2) {
decode_sb_mem(ctx, row, col, lflvl_ptr,
yoff2, uvoff2, BL_64X64);
} else {
decode_sb(ctx, row, col, lflvl_ptr,
yoff2, uvoff2, BL_64X64);
}
11 years ago
}
if (s->pass != 2) {
11 years ago
memcpy(&s->c_b[tile_col], &s->c, sizeof(s->c));
}
11 years ago
}
if (s->pass == 1) {
continue;
}
11 years ago
// backup pre-loopfilter reconstruction data for intra
// prediction of next row of sb64s
if (row + 8 < s->rows) {
memcpy(s->intra_pred_data[0],
f->data[0] + yoff + 63 * ls_y,
8 * s->cols * bytesperpixel);
11 years ago
memcpy(s->intra_pred_data[1],
f->data[1] + uvoff + ((64 >> s->ss_v) - 1) * ls_uv,
8 * s->cols * bytesperpixel >> s->ss_h);
11 years ago
memcpy(s->intra_pred_data[2],
f->data[2] + uvoff + ((64 >> s->ss_v) - 1) * ls_uv,
8 * s->cols * bytesperpixel >> s->ss_h);
11 years ago
}
// loopfilter one row
if (s->s.h.filter.level) {
11 years ago
yoff2 = yoff;
uvoff2 = uvoff;
lflvl_ptr = s->lflvl;
for (col = 0; col < s->cols;
col += 8, yoff2 += 64 * bytesperpixel,
uvoff2 += 64 * bytesperpixel >> s->ss_h, lflvl_ptr++) {
11 years ago
loopfilter_sb(ctx, lflvl_ptr, row, col, yoff2, uvoff2);
}
}
// FIXME maybe we can make this more finegrained by running the
// loopfilter per-block instead of after each sbrow
// In fact that would also make intra pred left preparation easier?
ff_thread_report_progress(&s->s.frames[CUR_FRAME].tf, row >> 3, 0);
11 years ago
}
}
if (s->pass < 2 && s->s.h.refreshctx && !s->s.h.parallelmode) {
adapt_probs(s);
ff_thread_finish_setup(ctx);
}
} while (s->pass++ == 1);
ff_thread_report_progress(&s->s.frames[CUR_FRAME].tf, INT_MAX, 0);
finish:
// ref frame setup
for (i = 0; i < 8; i++) {
if (s->s.refs[i].f->buf[0])
ff_thread_release_buffer(ctx, &s->s.refs[i]);
if (s->next_refs[i].f->buf[0] &&
(res = ff_thread_ref_frame(&s->s.refs[i], &s->next_refs[i])) < 0)
return res;
}
if (!s->s.h.invisible) {
if ((res = av_frame_ref(frame, s->s.frames[CUR_FRAME].tf.f)) < 0)
return res;
*got_frame = 1;
}
return pkt->size;
}
static void vp9_decode_flush(AVCodecContext *ctx)
{
VP9Context *s = ctx->priv_data;
int i;
for (i = 0; i < 3; i++)
vp9_unref_frame(ctx, &s->s.frames[i]);
for (i = 0; i < 8; i++)
ff_thread_release_buffer(ctx, &s->s.refs[i]);
}
static int init_frames(AVCodecContext *ctx)
{
VP9Context *s = ctx->priv_data;
int i;
for (i = 0; i < 3; i++) {
s->s.frames[i].tf.f = av_frame_alloc();
if (!s->s.frames[i].tf.f) {
vp9_decode_free(ctx);
av_log(ctx, AV_LOG_ERROR, "Failed to allocate frame buffer %d\n", i);
return AVERROR(ENOMEM);
}
}
for (i = 0; i < 8; i++) {
s->s.refs[i].f = av_frame_alloc();
s->next_refs[i].f = av_frame_alloc();
if (!s->s.refs[i].f || !s->next_refs[i].f) {
vp9_decode_free(ctx);
av_log(ctx, AV_LOG_ERROR, "Failed to allocate frame buffer %d\n", i);
return AVERROR(ENOMEM);
}
}
return 0;
}
static av_cold int vp9_decode_init(AVCodecContext *ctx)
{
VP9Context *s = ctx->priv_data;
ctx->internal->allocate_progress = 1;
s->last_bpp = 0;
s->s.h.filter.sharpness = -1;
return init_frames(ctx);
}
#if HAVE_THREADS
static av_cold int vp9_decode_init_thread_copy(AVCodecContext *avctx)
{
return init_frames(avctx);
}
static int vp9_decode_update_thread_context(AVCodecContext *dst, const AVCodecContext *src)
{
int i, res;
VP9Context *s = dst->priv_data, *ssrc = src->priv_data;
for (i = 0; i < 3; i++) {
if (s->s.frames[i].tf.f->buf[0])
vp9_unref_frame(dst, &s->s.frames[i]);
if (ssrc->s.frames[i].tf.f->buf[0]) {
if ((res = vp9_ref_frame(dst, &s->s.frames[i], &ssrc->s.frames[i])) < 0)
return res;
}
}
for (i = 0; i < 8; i++) {
if (s->s.refs[i].f->buf[0])
ff_thread_release_buffer(dst, &s->s.refs[i]);
if (ssrc->next_refs[i].f->buf[0]) {
if ((res = ff_thread_ref_frame(&s->s.refs[i], &ssrc->next_refs[i])) < 0)
return res;
}
}
s->s.h.invisible = ssrc->s.h.invisible;
s->s.h.keyframe = ssrc->s.h.keyframe;
s->s.h.intraonly = ssrc->s.h.intraonly;
s->ss_v = ssrc->ss_v;
s->ss_h = ssrc->ss_h;
s->s.h.segmentation.enabled = ssrc->s.h.segmentation.enabled;
s->s.h.segmentation.update_map = ssrc->s.h.segmentation.update_map;
s->s.h.segmentation.absolute_vals = ssrc->s.h.segmentation.absolute_vals;
s->bytesperpixel = ssrc->bytesperpixel;
s->gf_fmt = ssrc->gf_fmt;
s->w = ssrc->w;
s->h = ssrc->h;
s->s.h.bpp = ssrc->s.h.bpp;
s->bpp_index = ssrc->bpp_index;
s->pix_fmt = ssrc->pix_fmt;
memcpy(&s->prob_ctx, &ssrc->prob_ctx, sizeof(s->prob_ctx));
memcpy(&s->s.h.lf_delta, &ssrc->s.h.lf_delta, sizeof(s->s.h.lf_delta));
memcpy(&s->s.h.segmentation.feat, &ssrc->s.h.segmentation.feat,
sizeof(s->s.h.segmentation.feat));
return 0;
}
#endif
AVCodec ff_vp9_decoder = {
.name = "vp9",
.long_name = NULL_IF_CONFIG_SMALL("Google VP9"),
.type = AVMEDIA_TYPE_VIDEO,
.id = AV_CODEC_ID_VP9,
.priv_data_size = sizeof(VP9Context),
.init = vp9_decode_init,
.close = vp9_decode_free,
.decode = vp9_decode_frame,
.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
.flush = vp9_decode_flush,
.init_thread_copy = ONLY_IF_THREADS_ENABLED(vp9_decode_init_thread_copy),
.update_thread_context = ONLY_IF_THREADS_ENABLED(vp9_decode_update_thread_context),
.profiles = NULL_IF_CONFIG_SMALL(ff_vp9_profiles),
};