From 72ca830f511fcdc01253689615faed25da7c57bf Mon Sep 17 00:00:00 2001 From: "Ronald S. Bultje" Date: Mon, 30 Sep 2013 23:03:30 -0400 Subject: lavc: VP9 decoder MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Originally written by Ronald S. Bultje and Clément Bœsch Further contributions by: Anton Khirnov Diego Biurrun Luca Barbato Martin Storsjö Signed-off-by: Luca Barbato Signed-off-by: Anton Khirnov --- libavcodec/Makefile | 2 + libavcodec/allcodecs.c | 1 + libavcodec/version.h | 2 +- libavcodec/vp9.c | 1270 ++++++++++++++++++++++++ libavcodec/vp9.h | 419 ++++++++ libavcodec/vp9block.c | 1684 ++++++++++++++++++++++++++++++++ libavcodec/vp9data.c | 2133 +++++++++++++++++++++++++++++++++++++++++ libavcodec/vp9data.h | 70 ++ libavcodec/vp9dsp.c | 2174 ++++++++++++++++++++++++++++++++++++++++++ libavcodec/vp9mvs.c | 344 +++++++ libavcodec/vp9prob.c | 274 ++++++ libavcodec/x86/Makefile | 2 + libavcodec/x86/vp9dsp.asm | 277 ++++++ libavcodec/x86/vp9dsp_init.c | 245 +++++ 14 files changed, 8896 insertions(+), 1 deletion(-) create mode 100644 libavcodec/vp9.c create mode 100644 libavcodec/vp9.h create mode 100644 libavcodec/vp9block.c create mode 100644 libavcodec/vp9data.c create mode 100644 libavcodec/vp9data.h create mode 100644 libavcodec/vp9dsp.c create mode 100644 libavcodec/vp9mvs.c create mode 100644 libavcodec/vp9prob.c create mode 100644 libavcodec/x86/vp9dsp.asm create mode 100644 libavcodec/x86/vp9dsp_init.c (limited to 'libavcodec') diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 03d7459bb6..1674d4740f 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -392,6 +392,8 @@ OBJS-$(CONFIG_VP5_DECODER) += vp5.o vp56.o vp56data.o vp56dsp.o \ OBJS-$(CONFIG_VP6_DECODER) += vp6.o vp56.o vp56data.o vp56dsp.o \ vp6dsp.o vp56rac.o OBJS-$(CONFIG_VP8_DECODER) += vp8.o vp8dsp.o vp56rac.o +OBJS-$(CONFIG_VP9_DECODER) += vp9.o vp9data.o vp9dsp.o \ + vp9block.o vp9prob.o vp9mvs.o vp56rac.o OBJS-$(CONFIG_VQA_DECODER) += vqavideo.o OBJS-$(CONFIG_WAVPACK_DECODER) += wavpack.o OBJS-$(CONFIG_WEBP_DECODER) += webp.o diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index faa94b1ecb..6453e300f5 100644 --- a/libavcodec/allcodecs.c +++ b/libavcodec/allcodecs.c @@ -257,6 +257,7 @@ void avcodec_register_all(void) REGISTER_DECODER(VP6A, vp6a); REGISTER_DECODER(VP6F, vp6f); REGISTER_DECODER(VP8, vp8); + REGISTER_DECODER(VP9, vp9); REGISTER_DECODER(VQA, vqa); REGISTER_DECODER(WEBP, webp); REGISTER_ENCDEC (WMV1, wmv1); diff --git a/libavcodec/version.h b/libavcodec/version.h index 38a606422d..e9d09ac98e 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -27,7 +27,7 @@ */ #define LIBAVCODEC_VERSION_MAJOR 55 -#define LIBAVCODEC_VERSION_MINOR 27 +#define LIBAVCODEC_VERSION_MINOR 28 #define LIBAVCODEC_VERSION_MICRO 0 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ diff --git a/libavcodec/vp9.c b/libavcodec/vp9.c new file mode 100644 index 0000000000..9048700fb7 --- /dev/null +++ b/libavcodec/vp9.c @@ -0,0 +1,1270 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/avassert.h" + +#include "avcodec.h" +#include "get_bits.h" +#include "internal.h" +#include "videodsp.h" +#include "vp56.h" +#include "vp9.h" +#include "vp9data.h" + +#define VP9_SYNCCODE 0x498342 +#define MAX_PROB 255 + +static void vp9_decode_flush(AVCodecContext *avctx) +{ + VP9Context *s = avctx->priv_data; + int i; + + for (i = 0; i < FF_ARRAY_ELEMS(s->refs); i++) + av_frame_unref(s->refs[i]); +} + +static int update_size(AVCodecContext *avctx, int w, int h) +{ + VP9Context *s = avctx->priv_data; + uint8_t *p; + + if (s->above_partition_ctx && w == avctx->width && h == avctx->height) + return 0; + + vp9_decode_flush(avctx); + + if (w <= 0 || h <= 0) + return AVERROR_INVALIDDATA; + + avctx->width = w; + avctx->height = h; + 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_free(s->above_partition_ctx); + p = av_malloc(s->sb_cols * + (240 + sizeof(*s->lflvl) + 16 * sizeof(*s->above_mv_ctx) + + 64 * s->sb_rows * (1 + sizeof(*s->mv[0]) * 2))); + if (!p) + return AVERROR(ENOMEM); + 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_mode_ctx, uint8_t *, 16); + assign(s->above_y_nnz_ctx, uint8_t *, 16); + assign(s->above_uv_nnz_ctx[0], uint8_t *, 8); + assign(s->above_uv_nnz_ctx[1], uint8_t *, 8); + assign(s->intra_pred_data[0], uint8_t *, 64); + assign(s->intra_pred_data[1], uint8_t *, 32); + assign(s->intra_pred_data[2], uint8_t *, 32); + 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, VP9Filter *, 1); + assign(s->above_mv_ctx, VP56mv(*)[2], 16); + assign(s->segmentation_map, uint8_t *, 64 * s->sb_rows); + assign(s->mv[0], VP9MVRefPair *, 64 * s->sb_rows); + assign(s->mv[1], VP9MVRefPair *, 64 * s->sb_rows); +#undef assign + + return 0; +} + +// The sign bit is at the end, not the start, of a bit sequence +static av_always_inline int get_bits_with_sign(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) +{ + if (v > 2 * m) + return v; + if (v & 1) + return m - ((v + 1) >> 1); + return m + (v >> 1); +} + +// differential forward probability updates +static int update_prob(VP56RangeCoder *c, int p) +{ + static const int inv_map_table[MAX_PROB - 1] = { + 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, + }; + 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 = av_clip(d, 0, MAX_PROB - 65 - 1); + } + d += 64; + } + + 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 decode_frame_header(AVCodecContext *avctx, + const uint8_t *data, int size, int *ref) +{ + VP9Context *s = avctx->priv_data; + int c, i, j, k, l, m, n, w, h, max, size2, ret, sharp; + int last_invisible; + const uint8_t *data2; + + /* general header */ + if ((ret = init_get_bits8(&s->gb, data, size)) < 0) { + av_log(avctx, AV_LOG_ERROR, "Failed to initialize bitstream reader\n"); + return ret; + } + if (get_bits(&s->gb, 2) != 0x2) { // frame marker + av_log(avctx, AV_LOG_ERROR, "Invalid frame marker\n"); + return AVERROR_INVALIDDATA; + } + s->profile = get_bits1(&s->gb); + if (get_bits1(&s->gb)) { // reserved bit + av_log(avctx, AV_LOG_ERROR, "Reserved bit should be zero\n"); + return AVERROR_INVALIDDATA; + } + if (get_bits1(&s->gb)) { + *ref = get_bits(&s->gb, 3); + return 0; + } + + s->last_keyframe = s->keyframe; + s->keyframe = !get_bits1(&s->gb); + + last_invisible = s->invisible; + s->invisible = !get_bits1(&s->gb); + s->errorres = get_bits1(&s->gb); + // FIXME disable this upon resolution change + s->use_last_frame_mvs = !s->errorres && !last_invisible; + + if (s->keyframe) { + if (get_bits_long(&s->gb, 24) != VP9_SYNCCODE) { // synccode + av_log(avctx, AV_LOG_ERROR, "Invalid sync code\n"); + return AVERROR_INVALIDDATA; + } + s->colorspace = get_bits(&s->gb, 3); + if (s->colorspace == 7) { // RGB = profile 1 + av_log(avctx, AV_LOG_ERROR, "RGB not supported in profile 0\n"); + return AVERROR_INVALIDDATA; + } + s->fullrange = get_bits1(&s->gb); + // for profile 1, here follows the subsampling bits + s->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->intraonly = s->invisible ? get_bits1(&s->gb) : 0; + s->resetctx = s->errorres ? 0 : get_bits(&s->gb, 2); + if (s->intraonly) { + if (get_bits_long(&s->gb, 24) != VP9_SYNCCODE) { // synccode + av_log(avctx, AV_LOG_ERROR, "Invalid sync code\n"); + return AVERROR_INVALIDDATA; + } + s->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->refreshrefmask = get_bits(&s->gb, 8); + s->refidx[0] = get_bits(&s->gb, 3); + s->signbias[0] = get_bits1(&s->gb); + s->refidx[1] = get_bits(&s->gb, 3); + s->signbias[1] = get_bits1(&s->gb); + s->refidx[2] = get_bits(&s->gb, 3); + s->signbias[2] = get_bits1(&s->gb); + if (!s->refs[s->refidx[0]]->buf[0] || + !s->refs[s->refidx[1]]->buf[0] || + !s->refs[s->refidx[2]]->buf[0]) { + av_log(avctx, AV_LOG_ERROR, + "Not all references are available\n"); + return AVERROR_INVALIDDATA; + } + if (get_bits1(&s->gb)) { + w = s->refs[s->refidx[0]]->width; + h = s->refs[s->refidx[0]]->height; + } else if (get_bits1(&s->gb)) { + w = s->refs[s->refidx[1]]->width; + h = s->refs[s->refidx[1]]->height; + } else if (get_bits1(&s->gb)) { + w = s->refs[s->refidx[2]]->width; + h = s->refs[s->refidx[2]]->height; + } else { + 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); + s->highprecisionmvs = get_bits1(&s->gb); + s->filtermode = get_bits1(&s->gb) ? FILTER_SWITCHABLE : + get_bits(&s->gb, 2); + s->allowcompinter = s->signbias[0] != s->signbias[1] || + s->signbias[0] != s->signbias[2]; + if (s->allowcompinter) { + if (s->signbias[0] == s->signbias[1]) { + s->fixcompref = 2; + s->varcompref[0] = 0; + s->varcompref[1] = 1; + } else if (s->signbias[0] == s->signbias[2]) { + s->fixcompref = 1; + s->varcompref[0] = 0; + s->varcompref[1] = 2; + } else { + s->fixcompref = 0; + s->varcompref[0] = 1; + s->varcompref[1] = 2; + } + } + } + } + + s->refreshctx = s->errorres ? 0 : get_bits1(&s->gb); + s->parallelmode = s->errorres ? 1 : get_bits1(&s->gb); + s->framectxid = c = get_bits(&s->gb, 2); + + /* loopfilter header data */ + s->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->filter.sharpness != sharp) + memset(s->filter.lim_lut, 0, sizeof(s->filter.lim_lut)); + s->filter.sharpness = sharp; + if ((s->lf_delta.enabled = get_bits1(&s->gb))) { + if (get_bits1(&s->gb)) { + for (i = 0; i < 4; i++) + if (get_bits1(&s->gb)) + s->lf_delta.ref[i] = get_bits_with_sign(&s->gb, 6); + for (i = 0; i < 2; i++) + if (get_bits1(&s->gb)) + s->lf_delta.mode[i] = get_bits_with_sign(&s->gb, 6); + } + } else { + memset(&s->lf_delta, 0, sizeof(s->lf_delta)); + } + + /* quantization header data */ + s->yac_qi = get_bits(&s->gb, 8); + s->ydc_qdelta = get_bits1(&s->gb) ? get_bits_with_sign(&s->gb, 4) : 0; + s->uvdc_qdelta = get_bits1(&s->gb) ? get_bits_with_sign(&s->gb, 4) : 0; + s->uvac_qdelta = get_bits1(&s->gb) ? get_bits_with_sign(&s->gb, 4) : 0; + s->lossless = s->yac_qi == 0 && s->ydc_qdelta == 0 && + s->uvdc_qdelta == 0 && s->uvac_qdelta == 0; + + /* segmentation header info */ + if ((s->segmentation.enabled = get_bits1(&s->gb))) { + if ((s->segmentation.update_map = get_bits1(&s->gb))) { + for (i = 0; i < 7; i++) + s->prob.seg[i] = get_bits1(&s->gb) ? + get_bits(&s->gb, 8) : 255; + if ((s->segmentation.temporal = get_bits1(&s->gb))) + for (i = 0; i < 3; i++) + s->prob.segpred[i] = get_bits1(&s->gb) ? + get_bits(&s->gb, 8) : 255; + } + + if (get_bits1(&s->gb)) { + s->segmentation.absolute_vals = get_bits1(&s->gb); + for (i = 0; i < 8; i++) { + if ((s->segmentation.feat[i].q_enabled = get_bits1(&s->gb))) + s->segmentation.feat[i].q_val = get_bits_with_sign(&s->gb, 8); + if ((s->segmentation.feat[i].lf_enabled = get_bits1(&s->gb))) + s->segmentation.feat[i].lf_val = get_bits_with_sign(&s->gb, 6); + if ((s->segmentation.feat[i].ref_enabled = get_bits1(&s->gb))) + s->segmentation.feat[i].ref_val = get_bits(&s->gb, 2); + s->segmentation.feat[i].skip_enabled = get_bits1(&s->gb); + } + } + } else { + s->segmentation.feat[0].q_enabled = 0; + s->segmentation.feat[0].lf_enabled = 0; + s->segmentation.feat[0].skip_enabled = 0; + s->segmentation.feat[0].ref_enabled = 0; + } + + // set qmul[] based on Y/UV, AC/DC and segmentation Q idx deltas + for (i = 0; i < (s->segmentation.enabled ? 8 : 1); i++) { + int qyac, qydc, quvac, quvdc, lflvl, sh; + + if (s->segmentation.feat[i].q_enabled) { + if (s->segmentation.absolute_vals) + qyac = s->segmentation.feat[i].q_val; + else + qyac = s->yac_qi + s->segmentation.feat[i].q_val; + } else { + qyac = s->yac_qi; + } + qydc = av_clip_uintp2(qyac + s->ydc_qdelta, 8); + quvdc = av_clip_uintp2(qyac + s->uvdc_qdelta, 8); + quvac = av_clip_uintp2(qyac + s->uvac_qdelta, 8); + qyac = av_clip_uintp2(qyac, 8); + + s->segmentation.feat[i].qmul[0][0] = ff_vp9_dc_qlookup[qydc]; + s->segmentation.feat[i].qmul[0][1] = ff_vp9_ac_qlookup[qyac]; + s->segmentation.feat[i].qmul[1][0] = ff_vp9_dc_qlookup[quvdc]; + s->segmentation.feat[i].qmul[1][1] = ff_vp9_ac_qlookup[quvac]; + + sh = s->filter.level >= 32; + if (s->segmentation.feat[i].lf_enabled) { + if (s->segmentation.absolute_vals) + lflvl = s->segmentation.feat[i].lf_val; + else + lflvl = s->filter.level + s->segmentation.feat[i].lf_val; + } else { + lflvl = s->filter.level; + } + s->segmentation.feat[i].lflvl[0][0] = + s->segmentation.feat[i].lflvl[0][1] = + av_clip_uintp2(lflvl + (s->lf_delta.ref[0] << sh), 6); + for (j = 1; j < 4; j++) { + s->segmentation.feat[i].lflvl[j][0] = + av_clip_uintp2(lflvl + ((s->lf_delta.ref[j] + + s->lf_delta.mode[0]) << sh), 6); + s->segmentation.feat[i].lflvl[j][1] = + av_clip_uintp2(lflvl + ((s->lf_delta.ref[j] + + s->lf_delta.mode[1]) << sh), 6); + } + } + + /* tiling info */ + if ((ret = update_size(avctx, w, h)) < 0) { + av_log(avctx, AV_LOG_ERROR, + "Failed to initialize decoder for %dx%d\n", w, h); + return ret; + } + for (s->tiling.log2_tile_cols = 0; + (s->sb_cols >> s->tiling.log2_tile_cols) > 64; + s->tiling.log2_tile_cols++) ; + for (max = 0; (s->sb_cols >> max) >= 4; max++) ; + max = FFMAX(0, max - 1); + while (max > s->tiling.log2_tile_cols) { + if (get_bits1(&s->gb)) + s->tiling.log2_tile_cols++; + else + break; + } + s->tiling.log2_tile_rows = decode012(&s->gb); + s->tiling.tile_rows = 1 << s->tiling.log2_tile_rows; + if (s->tiling.tile_cols != (1 << s->tiling.log2_tile_cols)) { + s->tiling.tile_cols = 1 << s->tiling.log2_tile_cols; + s->c_b = av_fast_realloc(s->c_b, &s->c_b_size, + sizeof(VP56RangeCoder) * + s->tiling.tile_cols); + if (!s->c_b) { + av_log(avctx, AV_LOG_ERROR, + "Ran out of memory during range coder init\n"); + return AVERROR(ENOMEM); + } + } + + if (s->keyframe || s->errorres || s->intraonly) { + s->prob_ctx[0].p = + s->prob_ctx[1].p = + s->prob_ctx[2].p = + s->prob_ctx[3].p = ff_vp9_default_probs; + memcpy(s->prob_ctx[0].coef, ff_vp9_default_coef_probs, + sizeof(ff_vp9_default_coef_probs)); + memcpy(s->prob_ctx[1].coef, ff_vp9_default_coef_probs, + sizeof(ff_vp9_default_coef_probs)); + memcpy(s->prob_ctx[2].coef, ff_vp9_default_coef_probs, + sizeof(ff_vp9_default_coef_probs)); + memcpy(s->prob_ctx[3].coef, ff_vp9_default_coef_probs, + sizeof(ff_vp9_default_coef_probs)); + } + + // next 16 bits is size of the rest of the header (arith-coded) + size2 = get_bits(&s->gb, 16); + data2 = align_get_bits(&s->gb); + if (size2 > size - (data2 - data)) { + av_log(avctx, 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(avctx, AV_LOG_ERROR, "Marker bit was set\n"); + return AVERROR_INVALIDDATA; + } + + if (s->keyframe || s->intraonly) + memset(s->counts.coef, 0, + sizeof(s->counts.coef) + 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->lossless) { + s->txfmmode = TX_4X4; + } else { + s->txfmmode = vp8_rac_get_uint(&s->c, 2); + if (s->txfmmode == 3) + s->txfmmode += vp8_rac_get(&s->c); + + if (s->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->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->keyframe && !s->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->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->allowcompinter) { + s->comppredmode = vp8_rac_get(&s->c); + if (s->comppredmode) + s->comppredmode += vp8_rac_get(&s->c); + if (s->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->comppredmode = PRED_SINGLEREF; + } + + if (s->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->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->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 int decode_subblock(AVCodecContext *avctx, int row, int col, + VP9Filter *lflvl, + ptrdiff_t yoff, ptrdiff_t uvoff, enum BlockLevel bl) +{ + VP9Context *s = avctx->priv_data; + int c = ((s->above_partition_ctx[col] >> (3 - bl)) & 1) | + (((s->left_partition_ctx[row & 0x7] >> (3 - bl)) & 1) << 1); + int ret; + const uint8_t *p = s->keyframe ? ff_vp9_default_kf_partition_probs[bl][c] + : s->prob.p.partition[bl][c]; + enum BlockPartition bp; + ptrdiff_t hbs = 4 >> bl; + + if (bl == BL_8X8) { + bp = vp8_rac_get_tree(&s->c, ff_vp9_partition_tree, p); + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, bl, bp); + } else if (col + hbs < s->cols) { + if (row + hbs < s->rows) { + bp = vp8_rac_get_tree(&s->c, ff_vp9_partition_tree, p); + switch (bp) { + case PARTITION_NONE: + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, + bl, bp); + break; + case PARTITION_H: + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, + bl, bp); + if (!ret) { + yoff += hbs * 8 * s->cur_frame->linesize[0]; + uvoff += hbs * 4 * s->cur_frame->linesize[1]; + ret = ff_vp9_decode_block(avctx, row + hbs, col, lflvl, + yoff, uvoff, bl, bp); + } + break; + case PARTITION_V: + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, + bl, bp); + if (!ret) { + yoff += hbs * 8; + uvoff += hbs * 4; + ret = ff_vp9_decode_block(avctx, row, col + hbs, lflvl, + yoff, uvoff, bl, bp); + } + break; + case PARTITION_SPLIT: + ret = decode_subblock(avctx, row, col, lflvl, + yoff, uvoff, bl + 1); + if (!ret) { + ret = decode_subblock(avctx, row, col + hbs, lflvl, + yoff + 8 * hbs, uvoff + 4 * hbs, + bl + 1); + if (!ret) { + yoff += hbs * 8 * s->cur_frame->linesize[0]; + uvoff += hbs * 4 * s->cur_frame->linesize[1]; + ret = decode_subblock(avctx, row + hbs, col, lflvl, + yoff, uvoff, bl + 1); + if (!ret) { + ret = decode_subblock(avctx, row + hbs, col + hbs, + lflvl, yoff + 8 * hbs, + uvoff + 4 * hbs, bl + 1); + } + } + } + break; + default: + av_log(avctx, AV_LOG_ERROR, "Unexpected partition %d.", bp); + return AVERROR_INVALIDDATA; + } + } else if (vp56_rac_get_prob_branchy(&s->c, p[1])) { + bp = PARTITION_SPLIT; + ret = decode_subblock(avctx, row, col, lflvl, yoff, uvoff, bl + 1); + if (!ret) + ret = decode_subblock(avctx, row, col + hbs, lflvl, + yoff + 8 * hbs, uvoff + 4 * hbs, bl + 1); + } else { + bp = PARTITION_H; + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, + bl, bp); + } + } else if (row + hbs < s->rows) { + if (vp56_rac_get_prob_branchy(&s->c, p[2])) { + bp = PARTITION_SPLIT; + ret = decode_subblock(avctx, row, col, lflvl, yoff, uvoff, bl + 1); + if (!ret) { + yoff += hbs * 8 * s->cur_frame->linesize[0]; + uvoff += hbs * 4 * s->cur_frame->linesize[1]; + ret = decode_subblock(avctx, row + hbs, col, lflvl, + yoff, uvoff, bl + 1); + } + } else { + bp = PARTITION_V; + ret = ff_vp9_decode_block(avctx, row, col, lflvl, yoff, uvoff, + bl, bp); + } + } else { + bp = PARTITION_SPLIT; + ret = decode_subblock(avctx, row, col, lflvl, yoff, uvoff, bl + 1); + } + s->counts.partition[bl][c][bp]++; + + return ret; +} + +static void loopfilter_subblock(AVCodecContext *avctx, VP9Filter *lflvl, + int row, int col, + ptrdiff_t yoff, ptrdiff_t uvoff) +{ + VP9Context *s = avctx->priv_data; + uint8_t *dst = s->cur_frame->data[0] + yoff, *lvl = lflvl->level; + ptrdiff_t ls_y = s->cur_frame->linesize[0], ls_uv = s->cur_frame->linesize[1]; + int y, x, 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 edges between columns, Y plane (e.g. block1 | block2) + for (y = 0; y < 8; y += 2, dst += 16 * ls_y, lvl += 16) { + uint8_t *ptr = dst, *l = lvl, *hmask1 = lflvl->mask[0][0][y]; + uint8_t *hmask2 = lflvl->mask[0][0][y + 1]; + 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, l++) { + if (hm1 & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + if (col || x > 1) { + if (hmask1[0] & x) { + if (hmask2[0] & x) { + av_assert2(l[8] == L); + s->dsp.loop_filter_16[0](ptr, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[2][0](ptr, ls_y, E, I, H); + } + } else if (hm2 & x) { + L = l[8]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[!!(hmask1[1] & x)] + [!!(hmask2[1] & x)] + [0](ptr, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[!!(hmask1[1] & x)] + [0](ptr, ls_y, E, I, H); + } + } + } else if (hm2 & x) { + int L = l[8], H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + if (col || x > 1) { + s->dsp.loop_filter_8[!!(hmask2[1] & x)] + [0](ptr + 8 * ls_y, ls_y, E, I, H); + } + } + if (hm13 & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + if (hm23 & x) { + L = l[8]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[0][0][0](ptr + 4, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[0][0](ptr + 4, ls_y, E, I, H); + } + } else if (hm23 & x) { + int L = l[8], H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + s->dsp.loop_filter_8[0][0](ptr + 8 * ls_y + 4, ls_y, E, I, H); + } + } + } + + // block1 + // filter edges between rows, Y plane (e.g. ------) + // block2 + dst = s->cur_frame->data[0] + yoff; + lvl = lflvl->level; + for (y = 0; y < 8; y++, dst += 8 * ls_y, lvl += 8) { + uint8_t *ptr = dst, *l = lvl, *vmask = lflvl->mask[0][1][y]; + unsigned vm = vmask[0] | vmask[1] | vmask[2], vm3 = vmask[3]; + + for (x = 1; vm & ~(x - 1); x <<= 2, ptr += 16, l += 2) { + if (row || y) { + if (vm & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + if (vmask[0] & x) { + if (vmask[0] & (x << 1)) { + av_assert2(l[1] == L); + s->dsp.loop_filter_16[1](ptr, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[2][1](ptr, ls_y, E, I, H); + } + } else if (vm & (x << 1)) { + L = l[1]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[!!(vmask[1] & x)] + [!!(vmask[1] & (x << 1))] + [1](ptr, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[!!(vmask[1] & x)] + [1](ptr, ls_y, E, I, H); + } + } else if (vm & (x << 1)) { + int L = l[1], H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + s->dsp.loop_filter_8[!!(vmask[1] & (x << 1))] + [1](ptr + 8, ls_y, E, I, H); + } + } + if (vm3 & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + if (vm3 & (x << 1)) { + L = l[1]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[0][0][1](ptr + ls_y * 4, ls_y, E, I, H); + } else { + s->dsp.loop_filter_8[0][1](ptr + ls_y * 4, ls_y, E, I, H); + } + } else if (vm3 & (x << 1)) { + int L = l[1], H = L >> 4; + int E = s->filter.mblim_lut[L], I = s->filter.lim_lut[L]; + + s->dsp.loop_filter_8[0][1](ptr + ls_y * 4 + 8, ls_y, E, I, H); + } + } + } + + // same principle but for U/V planes + for (p = 0; p < 2; p++) { + lvl = lflvl->level; + dst = s->cur_frame->data[1 + p] + uvoff; + for (y = 0; y < 8; y += 4, dst += 16 * ls_uv, lvl += 32) { + uint8_t *ptr = dst, *l = lvl, *hmask1 = lflvl->mask[1][0][y]; + uint8_t *hmask2 = lflvl->mask[1][0][y + 2]; + unsigned hm1 = hmask1[0] | hmask1[1] | hmask1[2]; + unsigned hm2 = hmask2[1] | hmask2[2], hm = hm1 | hm2; + + for (x = 1; hm & ~(x - 1); x <<= 1, ptr += 4) { + if (col || x > 1) { + if (hm1 & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L]; + int I = s->filter.lim_lut[L]; + + if (hmask1[0] & x) { + if (hmask2[0] & x) { + av_assert2(l[16] == L); + s->dsp.loop_filter_16[0](ptr, ls_uv, E, I, H); + } else { + s->dsp.loop_filter_8[2][0](ptr, ls_uv, E, I, H); + } + } else if (hm2 & x) { + L = l[16]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[!!(hmask1[1] & x)] + [!!(hmask2[1] & x)] + [0](ptr, ls_uv, E, I, H); + } else { + s->dsp.loop_filter_8[!!(hmask1[1] & x)] + [0](ptr, ls_uv, E, I, H); + } + } else if (hm2 & x) { + int L = l[16], H = L >> 4; + int E = s->filter.mblim_lut[L]; + int I = s->filter.lim_lut[L]; + + s->dsp.loop_filter_8[!!(hmask2[1] & x)] + [0](ptr + 8 * ls_uv, ls_uv, E, I, H); + } + } + if (x & 0xAA) + l += 2; + } + } + lvl = lflvl->level; + dst = s->cur_frame->data[1 + p] + uvoff; + for (y = 0; y < 8; y++, dst += 4 * ls_uv) { + uint8_t *ptr = dst, *l = lvl, *vmask = lflvl->mask[1][1][y]; + unsigned vm = vmask[0] | vmask[1] | vmask[2]; + + for (x = 1; vm & ~(x - 1); x <<= 4, ptr += 16, l += 4) { + if (row || y) { + if (vm & x) { + int L = *l, H = L >> 4; + int E = s->filter.mblim_lut[L]; + int I = s->filter.lim_lut[L]; + + if (vmask[0] & x) { + if (vmask[0] & (x << 2)) { + av_assert2(l[2] == L); + s->dsp.loop_filter_16[1](ptr, ls_uv, E, I, H); + } else { + s->dsp.loop_filter_8[2][1](ptr, ls_uv, E, I, H); + } + } else if (vm & (x << 2)) { + L = l[2]; + H |= (L >> 4) << 8; + E |= s->filter.mblim_lut[L] << 8; + I |= s->filter.lim_lut[L] << 8; + s->dsp.loop_filter_mix2[!!(vmask[1] & x)] + [!!(vmask[1] & (x << 2))] + [1](ptr, ls_uv, E, I, H); + } else { + s->dsp.loop_filter_8[!!(vmask[1] & x)] + [1](ptr, ls_uv, E, I, H); + } + } else if (vm & (x << 2)) { + int L = l[2], H = L >> 4; + int E = s->filter.mblim_lut[L]; + int I = s->filter.lim_lut[L]; + + s->dsp.loop_filter_8[!!(vmask[1] & (x << 2))] + [1](ptr + 8, ls_uv, E, I, H); + } + } + } + if (y & 1) + lvl += 16; + } + } +} + +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 int vp9_decode_frame(AVCodecContext *avctx, AVFrame *frame, + int *got_frame, const uint8_t *data, int size) +{ + VP9Context *s = avctx->priv_data; + int ret, tile_row, tile_col, i, ref = -1, row, col; + ptrdiff_t yoff = 0, uvoff = 0; + + ret = decode_frame_header(avctx, data, size, &ref); + if (ret < 0) { + return ret; + } else if (!ret) { + if (!s->refs[ref]->buf[0]) { + av_log(avctx, AV_LOG_ERROR, + "Requested reference %d not available\n", ref); + return AVERROR_INVALIDDATA; + } + + ret = av_frame_ref(frame, s->refs[ref]); + if (ret < 0) + return ret; + *got_frame = 1; + return 0; + } + data += ret; + size -= ret; + + s->cur_frame = frame; + + av_frame_unref(s->cur_frame); + if ((ret = ff_get_buffer(avctx, s->cur_frame, + s->refreshrefmask ? AV_GET_BUFFER_FLAG_REF : 0)) < 0) + return ret; + s->cur_frame->key_frame = s->keyframe; + s->cur_frame->pict_type = s->keyframe ? AV_PICTURE_TYPE_I + : AV_PICTURE_TYPE_P; + + // main tile decode loop + memset(s->above_partition_ctx, 0, s->cols); + memset(s->above_skip_ctx, 0, s->cols); + if (s->keyframe || s->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 * 8); + memset(s->above_uv_nnz_ctx[1], 0, s->sb_cols * 8); + memset(s->above_segpred_ctx, 0, s->cols); + for (tile_row = 0; tile_row < s->tiling.tile_rows; tile_row++) { + set_tile_offset(&s->tiling.tile_row_start, &s->tiling.tile_row_end, + tile_row, s->tiling.log2_tile_rows, s->sb_rows); + for (tile_col = 0; tile_col < s->tiling.tile_cols; tile_col++) { + int64_t tile_size; + + if (tile_col == s->tiling.tile_cols - 1 && + tile_row == s->tiling.tile_rows - 1) { + tile_size = size; + } else { + tile_size = AV_RB32(data); + data += 4; + size -= 4; + } + if (tile_size > size) + return AVERROR_INVALIDDATA; + 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 + return AVERROR_INVALIDDATA; + data += tile_size; + size -= tile_size; + } + + for (row = s->tiling.tile_row_start; + row < s->tiling.tile_row_end; + row += 8, yoff += s->cur_frame->linesize[0] * 64, + uvoff += s->cur_frame->linesize[1] * 32) { + VP9Filter *lflvl = s->lflvl; + ptrdiff_t yoff2 = yoff, uvoff2 = uvoff; + + for (tile_col = 0; tile_col < s->tiling.tile_cols; tile_col++) { + set_tile_offset(&s->tiling.tile_col_start, + &s->tiling.tile_col_end, + tile_col, s->tiling.log2_tile_cols, s->sb_cols); + + memset(s->left_partition_ctx, 0, 8); + memset(s->left_skip_ctx, 0, 8); + if (s->keyframe || s->intraonly) + 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, 16); + memset(s->left_segpred_ctx, 0, 8); + + memcpy(&s->c, &s->c_b[tile_col], sizeof(s->c)); + for (col = s->tiling.tile_col_start; + col < s->tiling.tile_col_end; + col += 8, yoff2 += 64, uvoff2 += 32, lflvl++) { + // FIXME integrate with lf code (i.e. zero after each + // use, similar to invtxfm coefficients, or similar) + memset(lflvl->mask, 0, sizeof(lflvl->mask)); + + if ((ret = decode_subblock(avctx, row, col, lflvl, + yoff2, uvoff2, BL_64X64)) < 0) + return ret; + } + memcpy(&s->c_b[tile_col], &s->c, sizeof(s->c)); + } + + // backup pre-loopfilter reconstruction data for intra + // prediction of next row of sb64s + if (row + 8 < s->rows) { + memcpy(s->intra_pred_data[0], + s->cur_frame->data[0] + yoff + + 63 * s->cur_frame->linesize[0], + 8 * s->cols); + memcpy(s->intra_pred_data[1], + s->cur_frame->data[1] + uvoff + + 31 * s->cur_frame->linesize[1], + 4 * s->cols); + memcpy(s->intra_pred_data[2], + s->cur_frame->data[2] + uvoff + + 31 * s->cur_frame->linesize[2], + 4 * s->cols); + } + + // loopfilter one row + if (s->filter.level) { + yoff2 = yoff; + uvoff2 = uvoff; + lflvl = s->lflvl; + for (col = 0; col < s->cols; + col += 8, yoff2 += 64, uvoff2 += 32, lflvl++) + loopfilter_subblock(avctx, lflvl, row, col, yoff2, uvoff2); + } + } + } + + // bw adaptivity (or in case of parallel decoding mode, fw adaptivity + // probability maintenance between frames) + if (s->refreshctx) { + if (s->parallelmode) { + memcpy(s->prob_ctx[s->framectxid].coef, s->prob.coef, + sizeof(s->prob.coef)); + s->prob_ctx[s->framectxid].p = s->prob.p; + } else { + ff_vp9_adapt_probs(s); + } + } + FFSWAP(VP9MVRefPair *, s->mv[0], s->mv[1]); + + // ref frame setup + for (i = 0; i < 8; i++) + if (s->refreshrefmask & (1 << i)) { + av_frame_unref(s->refs[i]); + ret = av_frame_ref(s->refs[i], s->cur_frame); + if (ret < 0) + return ret; + } + + if (s->invisible) + av_frame_unref(s->cur_frame); + else + *got_frame = 1; + + return 0; +} + +static int vp9_decode_packet(AVCodecContext *avctx, void *frame, + int *got_frame, AVPacket *avpkt) +{ + const uint8_t *data = avpkt->data; + int size = avpkt->size; + int marker, ret; + + /* Read superframe index - this is a collection of individual frames + * that together lead to one visible frame */ + marker = data[size - 1]; + if ((marker & 0xe0) == 0xc0) { + int nbytes = 1 + ((marker >> 3) & 0x3); + int n_frames = 1 + (marker & 0x7); + int idx_sz = 2 + n_frames * nbytes; + + if (size >= idx_sz && data[size - idx_sz] == marker) { + const uint8_t *idx = data + size + 1 - idx_sz; + + while (n_frames--) { + int sz = AV_RL32(idx); + + if (nbytes < 4) + sz &= (1 << (8 * nbytes)) - 1; + idx += nbytes; + + if (sz > size) { + av_log(avctx, AV_LOG_ERROR, + "Superframe packet size too big: %d > %d\n", + sz, size); + return AVERROR_INVALIDDATA; + } + + ret = vp9_decode_frame(avctx, frame, got_frame, data, sz); + if (ret < 0) + return ret; + data += sz; + size -= sz; + } + return size; + } + } + + /* If we get here, there was no valid superframe index, i.e. this is just + * one whole single frame. Decode it as such from the complete input buf. */ + if ((ret = vp9_decode_frame(avctx, frame, got_frame, data, size)) < 0) + return ret; + return size; +} + +static av_cold int vp9_decode_free(AVCodecContext *avctx) +{ + VP9Context *s = avctx->priv_data; + int i; + + for (i = 0; i < FF_ARRAY_ELEMS(s->refs); i++) + av_frame_free(&s->refs[i]); + + av_freep(&s->c_b); + av_freep(&s->above_partition_ctx); + + return 0; +} + +static av_cold int vp9_decode_init(AVCodecContext *avctx) +{ + VP9Context *s = avctx->priv_data; + int i; + + avctx->pix_fmt = AV_PIX_FMT_YUV420P; + + ff_vp9dsp_init(&s->dsp); + ff_videodsp_init(&s->vdsp, 8); + + for (i = 0; i < FF_ARRAY_ELEMS(s->refs); i++) { + s->refs[i] = av_frame_alloc(); + if (!s->refs[i]) { + vp9_decode_free(avctx); + return AVERROR(ENOMEM); + } + } + + s->filter.sharpness = -1; + + return 0; +} + +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, + .decode = vp9_decode_packet, + .flush = vp9_decode_flush, + .close = vp9_decode_free, + .capabilities = CODEC_CAP_DR1, +}; diff --git a/libavcodec/vp9.h b/libavcodec/vp9.h new file mode 100644 index 0000000000..0a6c6eed42 --- /dev/null +++ b/libavcodec/vp9.h @@ -0,0 +1,419 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVCODEC_VP9_H +#define AVCODEC_VP9_H + +#include +#include + +#include "libavutil/internal.h" + +#include "avcodec.h" +#include "vp56.h" + +enum TxfmMode { + TX_4X4, + TX_8X8, + TX_16X16, + TX_32X32, + N_TXFM_SIZES, + TX_SWITCHABLE = N_TXFM_SIZES, + N_TXFM_MODES +}; + +enum TxfmType { + DCT_DCT, + DCT_ADST, + ADST_DCT, + ADST_ADST, + N_TXFM_TYPES +}; + +enum IntraPredMode { + VERT_PRED, + HOR_PRED, + DC_PRED, + DIAG_DOWN_LEFT_PRED, + DIAG_DOWN_RIGHT_PRED, + VERT_RIGHT_PRED, + HOR_DOWN_PRED, + VERT_LEFT_PRED, + HOR_UP_PRED, + TM_VP8_PRED, + LEFT_DC_PRED, + TOP_DC_PRED, + DC_128_PRED, + DC_127_PRED, + DC_129_PRED, + N_INTRA_PRED_MODES +}; + +enum FilterMode { + FILTER_8TAP_SMOOTH, + FILTER_8TAP_REGULAR, + FILTER_8TAP_SHARP, + FILTER_BILINEAR, + FILTER_SWITCHABLE, +}; + +enum BlockPartition { + PARTITION_NONE, // [ ] <-. + PARTITION_H, // [-] | + PARTITION_V, // [|] | + PARTITION_SPLIT, // [+] --' +}; + +enum InterPredMode { + NEARESTMV = 10, + NEARMV = 11, + ZEROMV = 12, + NEWMV = 13, +}; + +enum MVJoint { + MV_JOINT_ZERO, + MV_JOINT_H, + MV_JOINT_V, + MV_JOINT_HV, +}; + +typedef struct ProbContext { + uint8_t y_mode[4][9]; + uint8_t uv_mode[10][9]; + uint8_t filter[4][2]; + uint8_t mv_mode[7][3]; + uint8_t intra[4]; + uint8_t comp[5]; + uint8_t single_ref[5][2]; + uint8_t comp_ref[5]; + uint8_t tx32p[2][3]; + uint8_t tx16p[2][2]; + uint8_t tx8p[2]; + uint8_t skip[3]; + uint8_t mv_joint[3]; + struct { + uint8_t sign; + uint8_t classes[10]; + uint8_t class0; + uint8_t bits[10]; + uint8_t class0_fp[2][3]; + uint8_t fp[3]; + uint8_t class0_hp; + uint8_t hp; + } mv_comp[2]; + uint8_t partition[4][4][3]; +} ProbContext; + +typedef void (*vp9_mc_func)(uint8_t *dst, const uint8_t *ref, + ptrdiff_t dst_stride, + ptrdiff_t ref_stride, + int h, int mx, int my); + +typedef struct VP9DSPContext { + /* + * dimension 1: 0=4x4, 1=8x8, 2=16x16, 3=32x32 + * dimension 2: intra prediction modes + * + * dst/left/top is aligned by transform-size (i.e. 4, 8, 16 or 32 pixels) + * stride is aligned by 16 pixels + * top[-1] is top/left; top[4,7] is top-right for 4x4 + */ + // FIXME(rbultje) maybe replace left/top pointers with HAVE_TOP/ + // HAVE_LEFT/HAVE_TOPRIGHT flags instead, and then handle it in-place? + // also needs to fit in with what h264/vp8/etc do + void (*intra_pred[N_TXFM_SIZES][N_INTRA_PRED_MODES])(uint8_t *dst, + ptrdiff_t stride, + const uint8_t *left, + const uint8_t *top); + + /* + * dimension 1: 0=4x4, 1=8x8, 2=16x16, 3=32x32, 4=lossless (3-4=dct only) + * dimension 2: 0=dct/dct, 1=dct/adst, 2=adst/dct, 3=adst/adst + * + * dst is aligned by transform-size (i.e. 4, 8, 16 or 32 pixels) + * stride is aligned by 16 pixels + * block is 16-byte aligned + * eob indicates the position (+1) of the last non-zero coefficient, + * in scan-order. This can be used to write faster versions, e.g. a + * dc-only 4x4/8x8/16x16/32x32, or a 4x4-only (eob<10) 8x8/16x16/32x32, + * etc. + */ + // FIXME also write idct_add_block() versions for whole (inter) pred + // blocks, so we can do 2 4x4s at once + void (*itxfm_add[N_TXFM_SIZES + 1][N_TXFM_TYPES])(uint8_t *dst, + ptrdiff_t stride, + int16_t *block, int eob); + + /* + * dimension 1: width of filter (0=4, 1=8, 2=16) + * dimension 2: 0=col-edge filter (h), 1=row-edge filter (v) + * + * dst/stride are aligned by 8 + */ + void (*loop_filter_8[3][2])(uint8_t *dst, ptrdiff_t stride, + int mb_lim, int lim, int hev_thr); + + /* + * dimension 1: 0=col-edge filter (h), 1=row-edge filter (v) + * + * The width of filter is assumed to be 16; dst/stride are aligned by 16 + */ + void (*loop_filter_16[2])(uint8_t *dst, ptrdiff_t stride, + int mb_lim, int lim, int hev_thr); + + /* + * dimension 1/2: width of filter (0=4, 1=8) for each filter half + * dimension 3: 0=col-edge filter (h), 1=row-edge filter (v) + * + * dst/stride are aligned by operation size + * this basically calls loop_filter[d1][d3][0](), followed by + * loop_filter[d2][d3][0]() on the next 8 pixels + * mb_lim/lim/hev_thr contain two values in the lowest two bytes of the + * integer. + */ + // FIXME perhaps a mix4 that operates on 32px (for AVX2) + void (*loop_filter_mix2[2][2][2])(uint8_t *dst, ptrdiff_t stride, + int mb_lim, int lim, int hev_thr); + + /* + * dimension 1: hsize (0: 64, 1: 32, 2: 16, 3: 8, 4: 4) + * dimension 2: filter type (0: smooth, 1: regular, 2: sharp, 3: bilin) + * dimension 3: averaging type (0: put, 1: avg) + * dimension 4: x subpel interpolation (0: none, 1: 8tap/bilin) + * dimension 5: y subpel interpolation (1: none, 1: 8tap/bilin) + * + * dst/stride are aligned by hsize + */ + vp9_mc_func mc[5][4][2][2][2]; +} VP9DSPContext; + +enum CompPredMode { + PRED_SINGLEREF, + PRED_COMPREF, + PRED_SWITCHABLE, +}; + +typedef struct VP9MVRefPair { + VP56mv mv[2]; + int8_t ref[2]; +} VP9MVRefPair; + +typedef 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 */]; +} VP9Filter; + +enum BlockLevel { + BL_64X64, + BL_32X32, + BL_16X16, + BL_8X8, +}; + +enum BlockSize { + BS_64x64, + BS_64x32, + BS_32x64, + BS_32x32, + BS_32x16, + BS_16x32, + BS_16x16, + BS_16x8, + BS_8x16, + BS_8x8, + BS_8x4, + BS_4x8, + BS_4x4, + N_BS_SIZES, +}; + +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; + + int row, row7, col, col7; + uint8_t *dst[3]; + ptrdiff_t y_stride, uv_stride; +} VP9Block; + +typedef struct VP9Context { + VP9DSPContext dsp; + VideoDSPContext vdsp; + GetBitContext gb; + VP56RangeCoder c; + VP56RangeCoder *c_b; + unsigned c_b_size; + VP9Block b; + + // bitstream header + uint8_t profile; + uint8_t keyframe, last_keyframe; + uint8_t invisible; + uint8_t use_last_frame_mvs; + uint8_t errorres; + uint8_t colorspace; + uint8_t fullrange; + uint8_t intraonly; + uint8_t resetctx; + uint8_t refreshrefmask; + uint8_t highprecisionmvs; + enum FilterMode filtermode; + uint8_t allowcompinter; + uint8_t fixcompref; + uint8_t refreshctx; + uint8_t parallelmode; + uint8_t framectxid; + uint8_t refidx[3]; + uint8_t signbias[3]; + uint8_t varcompref[2]; + AVFrame *refs[8]; + AVFrame *cur_frame; + + struct { + uint8_t level; + int8_t sharpness; + uint8_t lim_lut[64]; + uint8_t mblim_lut[64]; + } filter; + struct { + uint8_t enabled; + int8_t mode[2]; + int8_t ref[4]; + } lf_delta; + uint8_t yac_qi; + int8_t ydc_qdelta, uvdc_qdelta, uvac_qdelta; + uint8_t lossless; + struct { + uint8_t enabled; + uint8_t temporal; + uint8_t absolute_vals; + uint8_t update_map; + struct { + uint8_t q_enabled; + uint8_t lf_enabled; + uint8_t ref_enabled; + uint8_t skip_enabled; + uint8_t ref_val; + int16_t q_val; + int8_t lf_val; + int16_t qmul[2][2]; + uint8_t lflvl[4][2]; + } feat[8]; + } segmentation; + struct { + unsigned log2_tile_cols, log2_tile_rows; + unsigned tile_cols, tile_rows; + unsigned tile_row_start, tile_row_end, tile_col_start, tile_col_end; + } tiling; + unsigned sb_cols, sb_rows, rows, cols; + struct { + ProbContext p; + uint8_t coef[4][2][2][6][6][3]; + } prob_ctx[4]; + struct { + ProbContext p; + uint8_t coef[4][2][2][6][6][11]; + uint8_t seg[7]; + uint8_t segpred[3]; + } 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; + enum TxfmMode txfmmode; + enum CompPredMode comppredmode; + + // contextual (left/above) cache + uint8_t left_partition_ctx[8], *above_partition_ctx; + uint8_t left_mode_ctx[16], *above_mode_ctx; + // FIXME maybe merge some of the below in a flags field? + uint8_t left_y_nnz_ctx[16], *above_y_nnz_ctx; + uint8_t left_uv_nnz_ctx[2][8], *above_uv_nnz_ctx[2]; + uint8_t left_skip_ctx[8], *above_skip_ctx; // 1bit + uint8_t left_txfm_ctx[8], *above_txfm_ctx; // 2bit + uint8_t left_segpred_ctx[8], *above_segpred_ctx; // 1bit + uint8_t left_intra_ctx[8], *above_intra_ctx; // 1bit + uint8_t left_comp_ctx[8], *above_comp_ctx; // 1bit + uint8_t left_ref_ctx[8], *above_ref_ctx; // 2bit + uint8_t left_filter_ctx[8], *above_filter_ctx; + VP56mv left_mv_ctx[16][2], (*above_mv_ctx)[2]; + + // whole-frame cache + uint8_t *intra_pred_data[3]; + uint8_t *segmentation_map; + VP9MVRefPair *mv[2]; + VP9Filter *lflvl; + DECLARE_ALIGNED(32, uint8_t, edge_emu_buffer)[71 * 80]; + + // block reconstruction intermediates + DECLARE_ALIGNED(32, int16_t, block)[4096]; + DECLARE_ALIGNED(32, int16_t, uvblock)[2][1024]; + uint8_t eob[256]; + uint8_t uveob[2][64]; + VP56mv min_mv, max_mv; + DECLARE_ALIGNED(32, uint8_t, tmp_y)[64 * 64]; + DECLARE_ALIGNED(32, uint8_t, tmp_uv)[2][32 * 32]; +} VP9Context; + +void ff_vp9dsp_init(VP9DSPContext *dsp); + +void ff_vp9dsp_init_x86(VP9DSPContext *dsp); + +void ff_vp9_fill_mv(VP9Context *s, VP56mv *mv, int mode, int sb); + +void ff_vp9_adapt_probs(VP9Context *s); + +int ff_vp9_decode_block(AVCodecContext *avctx, int row, int col, + VP9Filter *lflvl, ptrdiff_t yoff, ptrdiff_t uvoff, + enum BlockLevel bl, enum BlockPartition bp); + +#endif /* AVCODEC_VP9_H */ diff --git a/libavcodec/vp9block.c b/libavcodec/vp9block.c new file mode 100644 index 0000000000..e6865934c4 --- /dev/null +++ b/libavcodec/vp9block.c @@ -0,0 +1,1684 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/avassert.h" + +#include "avcodec.h" +#include "get_bits.h" +#include "internal.h" +#include "videodsp.h" +#include "vp56.h" +#include "vp9.h" +#include "vp9data.h" + +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 }, + } +}; + +// differential forward probability updates +static void decode_mode(VP9Context *s, VP9Block *const b) +{ + 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 + }; + int row = b->row, col = b->col, row7 = b->row7; + enum TxfmMode max_tx = max_tx_for_bl_bp[b->bs]; + int w4 = FFMIN(s->cols - col, bwh_tab[1][b->bs][0]); + int h4 = FFMIN(s->rows - row, bwh_tab[1][b->bs][1]); + int have_a = row > 0, have_l = col > s->tiling.tile_col_start; + int y; + + if (!s->segmentation.enabled) { + b->seg_id = 0; + } else if (s->keyframe || s->intraonly) { + b->seg_id = s->segmentation.update_map ? + vp8_rac_get_tree(&s->c, ff_vp9_segmentation_tree, s->prob.seg) : 0; + } else if (!s->segmentation.update_map || + (s->segmentation.temporal && + vp56_rac_get_prob_branchy(&s->c, + s->prob.segpred[s->above_segpred_ctx[col] + + s->left_segpred_ctx[row7]]))) { + int pred = 8, x; + + for (y = 0; y < h4; y++) + for (x = 0; x < w4; x++) + pred = FFMIN(pred, + s->segmentation_map[(y + row) * 8 * s->sb_cols + x + col]); + b->seg_id = pred; + + 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, ff_vp9_segmentation_tree, + s->prob.seg); + + memset(&s->above_segpred_ctx[col], 0, w4); + memset(&s->left_segpred_ctx[row7], 0, h4); + } + if ((s->segmentation.enabled && s->segmentation.update_map) || s->keyframe) { + for (y = 0; y < h4; y++) + memset(&s->segmentation_map[(y + row) * 8 * s->sb_cols + col], + b->seg_id, w4); + } + + b->skip = s->segmentation.enabled && + s->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->keyframe || s->intraonly) { + b->intra = 1; + } else if (s->segmentation.feat[b->seg_id].ref_enabled) { + b->intra = !s->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->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->txfmmode); + } + + if (s->keyframe || s->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, ff_vp9_intramode_tree, + ff_vp9_default_kf_ymode_probs[a[0]][l[0]]); + if (b->bs != BS_8x4) { + b->mode[1] = vp8_rac_get_tree(&s->c, ff_vp9_intramode_tree, + ff_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, ff_vp9_intramode_tree, + ff_vp9_default_kf_ymode_probs[a[0]][l[1]]); + if (b->bs != BS_8x4) { + b->mode[3] = vp8_rac_get_tree(&s->c, ff_vp9_intramode_tree, + ff_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, ff_vp9_intramode_tree, + ff_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, ff_vp9_intramode_tree, + ff_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, ff_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, ff_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, ff_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, ff_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, ff_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, ff_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->segmentation.feat[b->seg_id].ref_enabled) { + av_assert2(s->segmentation.feat[b->seg_id].ref_val != 0); + b->comp = 0; + b->ref[0] = s->segmentation.feat[b->seg_id].ref_val - 1; + } else { + // read comp_pred flag + if (s->comppredmode != PRED_SWITCHABLE) { + b->comp = s->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->fixcompref); + } else if (s->left_comp_ctx[row7]) { + c = 2 + (s->above_intra_ctx[col] || + s->above_ref_ctx[col] == s->fixcompref); + } else { + c = (!s->above_intra_ctx[col] && + s->above_ref_ctx[col] == s->fixcompref) ^ + (!s->left_intra_ctx[row7] && + s->left_ref_ctx[row & 7] == s->fixcompref); + } + } else { + c = s->above_comp_ctx[col] ? 3 : + (!s->above_intra_ctx[col] && s->above_ref_ctx[col] == s->fixcompref); + } + } else if (have_l) { + c = s->left_comp_ctx[row7] ? 3 : + (!s->left_intra_ctx[row7] && s->left_ref_ctx[row7] == s->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->signbias[s->fixcompref], var_idx = !fix_idx, c, bit; + + b->ref[fix_idx] = s->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->varcompref[1]); + } + } else if (s->left_intra_ctx[row7]) { + c = 1 + 2 * (s->above_ref_ctx[col] != s->varcompref[1]); + } else { + int refl = s->left_ref_ctx[row7], refa = s->above_ref_ctx[col]; + + if (refl == refa && refa == s->varcompref[1]) { + c = 0; + } else if (!s->left_comp_ctx[row7] && !s->above_comp_ctx[col]) { + if ((refa == s->fixcompref && refl == s->varcompref[0]) || + (refl == s->fixcompref && refa == s->varcompref[0])) { + c = 4; + } else { + c = (refa == refl) ? 3 : 1; + } + } else if (!s->left_comp_ctx[row7]) { + if (refa == s->varcompref[1] && refl != s->varcompref[1]) { + c = 1; + } else { + c = (refl == s->varcompref[1] && + refa != s->varcompref[1]) ? 2 : 4; + } + } else if (!s->above_comp_ctx[col]) { + if (refl == s->varcompref[1] && refa != s->varcompref[1]) { + c = 1; + } else { + c = (refa == s->varcompref[1] && + refl != s->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->varcompref[1]); + } else { + c = 3 * (s->above_ref_ctx[col] != s->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->varcompref[1]); + } else { + c = 3 * (s->left_ref_ctx[row7] != s->varcompref[1]); + } + } else { + c = 2; + } + bit = vp56_rac_get_prob(&s->c, s->prob.p.comp_ref[c]); + b->ref[var_idx] = s->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->fixcompref || !s->left_ref_ctx[row7] || + !s->above_ref_ctx[col]); + } else { + c = (3 * !s->above_ref_ctx[col]) + + (!s->fixcompref || !s->left_ref_ctx[row7]); + } + } else if (s->above_comp_ctx[col]) { + c = (3 * !s->left_ref_ctx[row7]) + + (!s->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->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->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->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->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->fixcompref == 1 || + s->left_ref_ctx[row7] == 1); + } else { + c = 2; + } + } else if (!s->left_ref_ctx[row7]) { + c = 1 + 2 * (s->fixcompref == 1 || + s->above_ref_ctx[col] == 1); + } else { + c = 3 * (s->left_ref_ctx[row7] == 1) + + (s->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->fixcompref == 1 || + s->left_ref_ctx[row7] == 1); + } else { + c = 3 * (s->above_ref_ctx[col] == 1) + + (s->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->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->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->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, ff_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->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; + } + + b->filter = vp8_rac_get_tree(&s->c, ff_vp9_filter_tree, + s->prob.p.filter[c]); + s->counts.filter[c][b->filter]++; + } else { + b->filter = s->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, ff_vp9_inter_mode_tree, + s->prob.p.mv_mode[c]); + s->counts.mv_mode[c][b->mode[0] - 10]++; + ff_vp9_fill_mv(s, b->mv[0], b->mode[0], 0); + + if (b->bs != BS_8x4) { + b->mode[1] = vp8_rac_get_tree(&s->c, ff_vp9_inter_mode_tree, + s->prob.p.mv_mode[c]); + s->counts.mv_mode[c][b->mode[1] - 10]++; + ff_vp9_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, ff_vp9_inter_mode_tree, + s->prob.p.mv_mode[c]); + s->counts.mv_mode[c][b->mode[2] - 10]++; + ff_vp9_fill_mv(s, b->mv[2], b->mode[2], 2); + + if (b->bs != BS_8x4) { + b->mode[3] = vp8_rac_get_tree(&s->c, ff_vp9_inter_mode_tree, + s->prob.p.mv_mode[c]); + s->counts.mv_mode[c][b->mode[3] - 10]++; + ff_vp9_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 { + ff_vp9_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]); + } + } + + // FIXME this can probably be optimized + memset(&s->above_skip_ctx[col], b->skip, w4); + memset(&s->left_skip_ctx[row7], b->skip, h4); + memset(&s->above_txfm_ctx[col], b->tx, w4); + memset(&s->left_txfm_ctx[row7], b->tx, h4); + memset(&s->above_partition_ctx[col], above_ctx[b->bs], w4); + memset(&s->left_partition_ctx[row7], left_ctx[b->bs], h4); + if (!s->keyframe && !s->intraonly) { + memset(&s->above_intra_ctx[col], b->intra, w4); + memset(&s->left_intra_ctx[row7], b->intra, h4); + memset(&s->above_comp_ctx[col], b->comp, w4); + memset(&s->left_comp_ctx[row7], b->comp, h4); + memset(&s->above_mode_ctx[col], b->mode[3], w4); + memset(&s->left_mode_ctx[row7], b->mode[3], h4); + if (s->filtermode == FILTER_SWITCHABLE && !b->intra) { + memset(&s->above_filter_ctx[col], b->filter, w4); + memset(&s->left_filter_ctx[row7], b->filter, h4); + b->filter = ff_vp9_filter_lut[b->filter]; + } + 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); + } + } + + if (!b->intra) { // FIXME write 0xff or -1 if intra, so we can use this + // as a direct check in above branches + int vref = b->ref[b->comp ? s->signbias[s->varcompref[0]] : 0]; + + memset(&s->above_ref_ctx[col], vref, w4); + memset(&s->left_ref_ctx[row7], vref, h4); + } + } + + // FIXME kinda ugly + for (y = 0; y < h4; y++) { + int x, o = (row + y) * s->sb_cols * 8 + col; + + if (b->intra) { + for (x = 0; x < w4; x++) { + s->mv[0][o + x].ref[0] = + s->mv[0][o + x].ref[1] = -1; + } + } else if (b->comp) { + for (x = 0; x < w4; x++) { + s->mv[0][o + x].ref[0] = b->ref[0]; + s->mv[0][o + x].ref[1] = b->ref[1]; + AV_COPY32(&s->mv[0][o + x].mv[0], &b->mv[3][0]); + AV_COPY32(&s->mv[0][o + x].mv[1], &b->mv[3][1]); + } + } else { + for (x = 0; x < w4; x++) { + s->mv[0][o + x].ref[0] = b->ref[0]; + s->mv[0][o + x].ref[1] = -1; + AV_COPY32(&s->mv[0][o + x].mv[0], &b->mv[3][0]); + } + } + } +} + +// FIXME remove tx argument, and merge cnt/eob arguments? +static int decode_block_coeffs(VP56RangeCoder *c, int16_t *coef, int n_coeffs, + enum TxfmMode tx, 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], ff_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 = vp56_rac_get_prob(c, 159) + 5; + } else { + val = (vp56_rac_get_prob(c, 165) << 1) + 7; + 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 = (vp56_rac_get_prob(c, 173) << 2) + 11; + val += (vp56_rac_get_prob(c, 148) << 1); + val += vp56_rac_get_prob(c, 140); + } else { + val = (vp56_rac_get_prob(c, 176) << 3) + 19; + 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 = (vp56_rac_get_prob(c, 180) << 4) + 35; + 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 = (vp56_rac_get_prob(c, 254) << 13) + 67; + 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); + } + } + } + if (!--band_left) + band_left = band_counts[++band]; + if (tx == TX_32X32) // FIXME slow + coef[rc] = ((vp8_rac_get(c) ? -val : val) * qmul[!!i]) / 2; + else + 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(AVCodecContext *avctx) +{ + VP9Context *s = avctx->priv_data; + VP9Block *const b = &s->b; + int row = b->row, col = b->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, step1d = 1 << b->tx, step = 1 << (b->tx * 2); + int uvstep1d = 1 << b->uvtx, uvstep = 1 << (b->uvtx * 2), ret; + int16_t (*qmul)[2] = s->segmentation.feat[b->seg_id].qmul; + int tx = 4 * s->lossless + b->tx; + const int16_t **yscans = ff_vp9_scans[tx]; + const int16_t (**ynbs)[2] = ff_vp9_scans_nb[tx]; + const int16_t *uvscan = ff_vp9_scans[b->uvtx][DCT_DCT]; + const int16_t (*uvnb)[2] = ff_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, 0 }, + { 1, 2, 3, 4, 11, 64 - 21, 0 }, + { 1, 2, 3, 4, 11, 256 - 21, 0 }, + { 1, 2, 3, 4, 11, 1024 - 21, 0 }, + }; + const int16_t *y_band_counts = band_counts[b->tx]; + const int16_t *uv_band_counts = band_counts[b->uvtx]; + + /* y tokens */ + if (b->tx > TX_4X4) { // FIXME slow + for (y = 0; y < end_y; y += step1d) + for (x = 1; x < step1d; x++) + l[y] |= l[y + x]; + for (x = 0; x < end_x; x += step1d) + for (y = 1; y < step1d; y++) + a[x] |= a[x + y]; + } + for (n = 0, y = 0; y < end_y; y += step1d) { + for (x = 0; x < end_x; x += step1d, n += step) { + enum TxfmType txtp = ff_vp9_intra_txfm_type[b->mode[b->tx == TX_4X4 && + b->bs > BS_8x8 ? + n : 0]]; + int nnz = a[x] + l[y]; + if ((ret = decode_block_coeffs(&s->c, s->block + 16 * n, 16 * step, + b->tx, c, e, p, nnz, yscans[txtp], + ynbs[txtp], y_band_counts, + qmul[0])) < 0) + return ret; + a[x] = l[y] = !!ret; + if (b->tx > TX_8X8) + AV_WN16A(&s->eob[n], ret); + else + s->eob[n] = ret; + } + } + if (b->tx > TX_4X4) { // FIXME slow + for (y = 0; y < end_y; y += step1d) + memset(&l[y + 1], l[y], FFMIN(end_y - y - 1, step1d - 1)); + for (x = 0; x < end_x; x += step1d) + memset(&a[x + 1], a[x], FFMIN(end_x - x - 1, step1d - 1)); + } + + 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 >>= 1; + h4 >>= 1; + end_x >>= 1; + end_y >>= 1; + for (pl = 0; pl < 2; pl++) { + a = &s->above_uv_nnz_ctx[pl][col]; + l = &s->left_uv_nnz_ctx[pl][row & 7]; + if (b->uvtx > TX_4X4) { // FIXME slow + for (y = 0; y < end_y; y += uvstep1d) + for (x = 1; x < uvstep1d; x++) + l[y] |= l[y + x]; + for (x = 0; x < end_x; x += uvstep1d) + for (y = 1; y < uvstep1d; y++) + a[x] |= a[x + y]; + } + for (n = 0, y = 0; y < end_y; y += uvstep1d) { + for (x = 0; x < end_x; x += uvstep1d, n += uvstep) { + int nnz = a[x] + l[y]; + if ((ret = decode_block_coeffs(&s->c, s->uvblock[pl] + 16 * n, + 16 * uvstep, b->uvtx, c, e, p, + nnz, uvscan, uvnb, + uv_band_counts, qmul[1])) < 0) + return ret; + a[x] = l[y] = !!ret; + if (b->uvtx > TX_8X8) + AV_WN16A(&s->uveob[pl][n], ret); + else + s->uveob[pl][n] = ret; + } + } + if (b->uvtx > TX_4X4) { // FIXME slow + for (y = 0; y < end_y; y += uvstep1d) + memset(&l[y + 1], l[y], FFMIN(end_y - y - 1, uvstep1d - 1)); + for (x = 0; x < end_x; x += uvstep1d) + memset(&a[x + 1], a[x], FFMIN(end_x - x - 1, uvstep1d - 1)); + } + } + + return 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 have_top = row > 0 || y > 0; + int have_left = col > s->tiling.tile_col_start || x > 0; + int have_right = x < w - 1; + 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; + } 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 }, + [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 = NULL, *topleft = NULL; + int n_px_need = 4 << tx, n_px_have = (((s->cols - col) << !p) - 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 >> !!p) + x * 4 : + y == 0 ? &dst_edge[-stride_edge] : &dst_inner[-stride_inner]; + if (have_left) + topleft = !(row & 7) && !y ? + s->intra_pred_data[p] + col * (8 >> !!p) + x * 4 : + 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); + } else { + memcpy(*a, top, n_px_have); + memset(&(*a)[n_px_have], (*a)[n_px_have - 1], + n_px_need - n_px_have); + } + } else { + memset(*a, 127, n_px_need); + } + if (edges[mode].needs_topleft) { + if (have_left && have_top) + (*a)[-1] = topleft[-1]; + else + (*a)[-1] = have_top ? 129 : 127; + } + 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], &top[4], 4); + } else { + memset(&(*a)[4], (*a)[3], 4); + } + } + } + } + if (edges[mode].needs_left) { + if (have_left) { + int i; + int n_px_need = 4 << tx; + int n_px_have = (((s->rows - row) << !p) - y) * 4; + uint8_t *dst = x == 0 ? dst_edge : dst_inner; + ptrdiff_t stride = x == 0 ? stride_edge : stride_inner; + + if (n_px_need <= n_px_have) { + for (i = 0; i < n_px_need; i++) + l[i] = dst[i * stride - 1]; + } else { + for (i = 0; i < n_px_have; i++) + l[i] = dst[i * stride - 1]; + memset(&l[i], l[i - 1], n_px_need - n_px_have); + } + } else { + memset(l, 129, 4 << tx); + } + } + + return mode; +} + +static void intra_recon(AVCodecContext *avctx, ptrdiff_t y_off, ptrdiff_t uv_off) +{ + VP9Context *s = avctx->priv_data; + VP9Block *const b = &s->b; + int row = b->row, col = b->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->lossless + b->tx, uvtx = b->uvtx + 4 * s->lossless; + int uvstep1d = 1 << b->uvtx, p; + uint8_t *dst = b->dst[0], *dst_r = s->cur_frame->data[0] + y_off; + + 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, ptr_r += 4 * step1d, n += step) { + int mode = b->mode[b->bs > BS_8x8 && b->tx == TX_4X4 ? + y * 2 + x : 0]; + LOCAL_ALIGNED_16(uint8_t, a_buf, [48]); + uint8_t *a = &a_buf[16], l[32]; + enum TxfmType txtp = ff_vp9_intra_txfm_type[mode]; + int eob = b->tx > TX_8X8 ? AV_RN16A(&s->eob[n]) : s->eob[n]; + + mode = check_intra_mode(s, mode, &a, ptr_r, + s->cur_frame->linesize[0], + ptr, b->y_stride, l, + col, x, w4, row, y, b->tx, 0); + s->dsp.intra_pred[b->tx][mode](ptr, b->y_stride, l, a); + if (eob) + s->dsp.itxfm_add[tx][txtp](ptr, b->y_stride, + s->block + 16 * n, eob); + } + dst_r += 4 * s->cur_frame->linesize[0] * step1d; + dst += 4 * b->y_stride * step1d; + } + + // U/V + h4 >>= 1; + w4 >>= 1; + end_x >>= 1; + end_y >>= 1; + step = 1 << (b->uvtx * 2); + for (p = 0; p < 2; p++) { + dst = b->dst[1 + p]; + dst_r = s->cur_frame->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, + ptr_r += 4 * uvstep1d, n += step) { + int mode = b->uvmode; + LOCAL_ALIGNED_16(uint8_t, a_buf, [48]); + uint8_t *a = &a_buf[16], l[32]; + int eob = b->uvtx > TX_8X8 ? AV_RN16A(&s->uveob[p][n]) + : s->uveob[p][n]; + + mode = check_intra_mode(s, mode, &a, ptr_r, + s->cur_frame->linesize[1], + ptr, b->uv_stride, l, + col, x, w4, row, y, b->uvtx, p + 1); + s->dsp.intra_pred[b->uvtx][mode](ptr, b->uv_stride, l, a); + if (eob) + s->dsp.itxfm_add[uvtx][DCT_DCT](ptr, b->uv_stride, + s->uvblock[p] + 16 * n, + eob); + } + dst_r += 4 * uvstep1d * s->cur_frame->linesize[1]; + dst += 4 * uvstep1d * b->uv_stride; + } + } +} + +static av_always_inline void mc_luma_dir(VP9Context *s, vp9_mc_func(*mc)[2], + uint8_t *dst, ptrdiff_t dst_stride, + const uint8_t *ref, + ptrdiff_t ref_stride, + ptrdiff_t y, ptrdiff_t x, + const VP56mv *mv, + int bw, int bh, int w, int h) +{ + int mx = mv->x, my = mv->y; + + y += my >> 3; + x += mx >> 3; + ref += y * ref_stride + x; + mx &= 7; + my &= 7; + // FIXME bilinear filter only needs 0/1 pixels, not 3/4 + if (x < !!mx * 3 || y < !!my * 3 || + x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) { + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, + ref - !!my * 3 * ref_stride - !!mx * 3, + 80, + ref_stride, + bw + !!mx * 7, bh + !!my * 7, + x - !!mx * 3, y - !!my * 3, w, h); + ref = s->edge_emu_buffer + !!my * 3 * 80 + !!mx * 3; + ref_stride = 80; + } + mc[!!mx][!!my](dst, ref, dst_stride, ref_stride, bh, mx << 1, my << 1); +} + +static av_always_inline void mc_chroma_dir(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, + ptrdiff_t y, ptrdiff_t x, + const VP56mv *mv, + int bw, int bh, int w, int h) +{ + int mx = mv->x, my = mv->y; + + y += my >> 4; + x += mx >> 4; + ref_u += y * src_stride_u + x; + ref_v += y * src_stride_v + x; + mx &= 15; + my &= 15; + // FIXME bilinear filter only needs 0/1 pixels, not 3/4 + if (x < !!mx * 3 || y < !!my * 3 || + x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) { + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, + ref_u - !!my * 3 * src_stride_u - !!mx * 3, + 80, + 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 * 80 + !!mx * 3; + mc[!!mx][!!my](dst_u, ref_u, dst_stride, 80, bh, mx, my); + + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, + ref_v - !!my * 3 * src_stride_v - !!mx * 3, + 80, + 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 * 80 + !!mx * 3; + mc[!!mx][!!my](dst_v, ref_v, dst_stride, 80, bh, mx, my); + } else { + mc[!!mx][!!my](dst_u, ref_u, dst_stride, src_stride_u, bh, mx, my); + mc[!!mx][!!my](dst_v, ref_v, dst_stride, src_stride_v, bh, mx, my); + } +} + +static int inter_recon(AVCodecContext *avctx) +{ + static const uint8_t bwlog_tab[2][N_BS_SIZES] = { + { 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 }, + { 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 }, + }; + VP9Context *s = avctx->priv_data; + VP9Block *const b = &s->b; + int row = b->row, col = b->col; + AVFrame *ref1 = s->refs[s->refidx[b->ref[0]]]; + AVFrame *ref2 = b->comp ? s->refs[s->refidx[b->ref[1]]] : NULL; + int w = avctx->width, h = avctx->height; + ptrdiff_t ls_y = b->y_stride, ls_uv = b->uv_stride; + + if (!ref1->data[0] || (b->comp && !ref2->data[0])) + return AVERROR_INVALIDDATA; + + // y inter pred + if (b->bs > BS_8x8) { + if (b->bs == BS_8x4) { + mc_luma_dir(s, s->dsp.mc[3][b->filter][0], b->dst[0], ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, col << 3, &b->mv[0][0], 8, 4, w, h); + mc_luma_dir(s, s->dsp.mc[3][b->filter][0], + b->dst[0] + 4 * ls_y, ls_y, + ref1->data[0], ref1->linesize[0], + (row << 3) + 4, col << 3, &b->mv[2][0], 8, 4, w, h); + + if (b->comp) { + mc_luma_dir(s, s->dsp.mc[3][b->filter][1], b->dst[0], ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, col << 3, &b->mv[0][1], 8, 4, w, h); + mc_luma_dir(s, s->dsp.mc[3][b->filter][1], + b->dst[0] + 4 * ls_y, ls_y, + ref2->data[0], ref2->linesize[0], + (row << 3) + 4, col << 3, &b->mv[2][1], 8, 4, w, h); + } + } else if (b->bs == BS_4x8) { + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0], ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, col << 3, &b->mv[0][0], 4, 8, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0] + 4, ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, (col << 3) + 4, &b->mv[1][0], 4, 8, w, h); + + if (b->comp) { + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0], ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, col << 3, &b->mv[0][1], 4, 8, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0] + 4, ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, (col << 3) + 4, &b->mv[1][1], 4, 8, w, h); + } + } else { + av_assert2(b->bs == BS_4x4); + + // FIXME if two horizontally adjacent blocks have the same MV, + // do a w8 instead of a w4 call + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0], ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, col << 3, &b->mv[0][0], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0] + 4, ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, (col << 3) + 4, &b->mv[1][0], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], + b->dst[0] + 4 * ls_y, ls_y, + ref1->data[0], ref1->linesize[0], + (row << 3) + 4, col << 3, &b->mv[2][0], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][0], + b->dst[0] + 4 * ls_y + 4, ls_y, + ref1->data[0], ref1->linesize[0], + (row << 3) + 4, (col << 3) + 4, &b->mv[3][0], 4, 4, w, h); + + if (b->comp) { + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0], ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, col << 3, &b->mv[0][1], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0] + 4, ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, (col << 3) + 4, &b->mv[1][1], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], + b->dst[0] + 4 * ls_y, ls_y, + ref2->data[0], ref2->linesize[0], + (row << 3) + 4, col << 3, &b->mv[2][1], 4, 4, w, h); + mc_luma_dir(s, s->dsp.mc[4][b->filter][1], + b->dst[0] + 4 * ls_y + 4, ls_y, + ref2->data[0], ref2->linesize[0], + (row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, w, h); + } + } + } else { + int bwl = bwlog_tab[0][b->bs]; + int bw = bwh_tab[0][b->bs][0] * 4; + int bh = bwh_tab[0][b->bs][1] * 4; + + mc_luma_dir(s, s->dsp.mc[bwl][b->filter][0], b->dst[0], ls_y, + ref1->data[0], ref1->linesize[0], + row << 3, col << 3, &b->mv[0][0], bw, bh, w, h); + + if (b->comp) + mc_luma_dir(s, s->dsp.mc[bwl][b->filter][1], b->dst[0], ls_y, + ref2->data[0], ref2->linesize[0], + row << 3, col << 3, &b->mv[0][1], bw, bh, w, h); + } + + // uv inter pred + { + int bwl = bwlog_tab[1][b->bs]; + int bw = bwh_tab[1][b->bs][0] * 4, bh = bwh_tab[1][b->bs][1] * 4; + VP56mv mvuv; + + w = (w + 1) >> 1; + h = (h + 1) >> 1; + if (b->bs > BS_8x8) { + mvuv.x = ROUNDED_DIV(b->mv[0][0].x + b->mv[1][0].x + + b->mv[2][0].x + b->mv[3][0].x, 4); + mvuv.y = ROUNDED_DIV(b->mv[0][0].y + b->mv[1][0].y + + b->mv[2][0].y + b->mv[3][0].y, 4); + } else { + mvuv = b->mv[0][0]; + } + + mc_chroma_dir(s, s->dsp.mc[bwl][b->filter][0], + b->dst[1], b->dst[2], ls_uv, + ref1->data[1], ref1->linesize[1], + ref1->data[2], ref1->linesize[2], + row << 2, col << 2, &mvuv, bw, bh, w, h); + + if (b->comp) { + if (b->bs > BS_8x8) { + mvuv.x = ROUNDED_DIV(b->mv[0][1].x + b->mv[1][1].x + + b->mv[2][1].x + b->mv[3][1].x, 4); + mvuv.y = ROUNDED_DIV(b->mv[0][1].y + b->mv[1][1].y + + b->mv[2][1].y + b->mv[3][1].y, 4); + } else { + mvuv = b->mv[0][1]; + } + mc_chroma_dir(s, s->dsp.mc[bwl][b->filter][1], + b->dst[1], b->dst[2], ls_uv, + ref2->data[1], ref2->linesize[1], + ref2->data[2], ref2->linesize[2], + row << 2, col << 2, &mvuv, bw, bh, w, h); + } + } + + if (!b->skip) { + /* mostly copied intra_reconn() */ + + 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->lossless + b->tx, uvtx = b->uvtx + 4 * s->lossless; + int uvstep1d = 1 << b->uvtx, p; + uint8_t *dst = b->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, 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, b->y_stride, + s->block + 16 * n, eob); + } + dst += 4 * b->y_stride * step1d; + } + + // uv itxfm add + h4 >>= 1; + w4 >>= 1; + end_x >>= 1; + end_y >>= 1; + step = 1 << (b->uvtx * 2); + for (p = 0; p < 2; p++) { + dst = b->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, 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, b->uv_stride, + s->uvblock[p] + 16 * n, eob); + } + dst += 4 * uvstep1d * b->uv_stride; + } + } + } + return 0; +} + +static av_always_inline void mask_edges(VP9Filter *lflvl, int is_uv, + int row_and_7, int col_and_7, + int w, int h, int col_end, int row_end, + enum TxfmMode tx, int skip_inter) +{ + // 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 && is_uv) { + if (h == 1) { + if (row_and_7 & 1) + return; + if (!row_end) + h += 1; + } + if (w == 1) { + 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; + int m_col_odd = (t << (w - 1)) - t; + + // on 32-px edges, use the 8-px wide loopfilter; else, use 4-px wide + if (is_uv) { + int m_row_8 = m_col & 0x01, 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 & 7); + + lflvl->mask[is_uv][0][y][1] |= m_row_8; + lflvl->mask[is_uv][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 ((col_end & 1) && (y & 1)) { + lflvl->mask[is_uv][1][y][col_mask_id] |= m_col_odd; + } else { + lflvl->mask[is_uv][1][y][col_mask_id] |= m_col; + } + } + } else { + int m_row_8 = m_col & 0x11, 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 & 3); + + lflvl->mask[is_uv][0][y][1] |= m_row_8; // row edge + lflvl->mask[is_uv][0][y][2] |= m_row_4; + lflvl->mask[is_uv][1][y][col_mask_id] |= m_col; // col edge + lflvl->mask[is_uv][0][y][3] |= m_col; + lflvl->mask[is_uv][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); + int l2 = tx + is_uv - 1, step1d = 1 << l2; + static const unsigned masks[4] = { 0xff, 0x55, 0x11, 0x01 }; + 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 (is_uv && 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++) { + lflvl->mask[is_uv][0][y][0] |= m_row_16; + lflvl->mask[is_uv][0][y][1] |= m_row_8; + } + } else { + for (y = row_and_7; y < h + row_and_7; y++) + lflvl->mask[is_uv][0][y][mask_id] |= m_row; + } + + if (is_uv && tx > TX_8X8 && (h ^ (h - 1)) == 1) { + for (y = row_and_7; y < h + row_and_7 - 1; y += step1d) + lflvl->mask[is_uv][1][y][0] |= m_col; + if (y - row_and_7 == h - 1) + lflvl->mask[is_uv][1][y][1] |= m_col; + } else { + for (y = row_and_7; y < h + row_and_7; y += step1d) + lflvl->mask[is_uv][1][y][mask_id] |= m_col; + } + } else if (tx != TX_4X4) { + int mask_id; + + mask_id = (tx == TX_8X8) || (is_uv && h == 1); + lflvl->mask[is_uv][1][row_and_7][mask_id] |= m_col; + mask_id = (tx == TX_8X8) || (is_uv && w == 1); + for (y = row_and_7; y < h + row_and_7; y++) + lflvl->mask[is_uv][0][y][mask_id] |= t; + } else if (is_uv) { + int t8 = t & 0x01, t4 = t - t8; + + for (y = row_and_7; y < h + row_and_7; y++) { + lflvl->mask[is_uv][0][y][2] |= t4; + lflvl->mask[is_uv][0][y][1] |= t8; + } + lflvl->mask[is_uv][1][row_and_7][2 - !(row_and_7 & 7)] |= m_col; + } else { + int t8 = t & 0x11, t4 = t - t8; + + for (y = row_and_7; y < h + row_and_7; y++) { + lflvl->mask[is_uv][0][y][2] |= t4; + lflvl->mask[is_uv][0][y][1] |= t8; + } + lflvl->mask[is_uv][1][row_and_7][2 - !(row_and_7 & 3)] |= m_col; + } + } +} + +int ff_vp9_decode_block(AVCodecContext *avctx, int row, int col, + VP9Filter *lflvl, ptrdiff_t yoff, ptrdiff_t uvoff, + enum BlockLevel bl, enum BlockPartition bp) +{ + VP9Context *s = avctx->priv_data; + VP9Block *const b = &s->b; + enum BlockSize bs = bl * 3 + bp; + int ret, y, w4 = bwh_tab[1][bs][0], h4 = bwh_tab[1][bs][1], lvl; + int emu[2]; + + b->row = row; + b->row7 = row & 7; + b->col = col; + b->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; + + b->bs = bs; + decode_mode(s, b); + b->uvtx = b->tx - (w4 * 2 == (1 << b->tx) || h4 * 2 == (1 << b->tx)); + + if (!b->skip) { + if ((ret = decode_coeffs(avctx)) < 0) + return ret; + } else { + int pl; + + memset(&s->above_y_nnz_ctx[col * 2], 0, w4 * 2); + memset(&s->left_y_nnz_ctx[(row & 7) << 1], 0, h4 * 2); + for (pl = 0; pl < 2; pl++) { + memset(&s->above_uv_nnz_ctx[pl][col], 0, w4); + memset(&s->left_uv_nnz_ctx[pl][row & 7], 0, h4); + } + } + + /* Emulated overhangs if the stride of the target buffer can't hold. + * This allows to support emu-edge and so on even if we have large + * block overhangs. */ + emu[0] = (col + w4) * 8 > s->cur_frame->linesize[0] || + (row + h4) > s->rows + 2 * !(avctx->flags & CODEC_FLAG_EMU_EDGE); + emu[1] = (col + w4) * 4 > s->cur_frame->linesize[1] || + (row + h4) > s->rows + 2 * !(avctx->flags & CODEC_FLAG_EMU_EDGE); + if (emu[0]) { + b->dst[0] = s->tmp_y; + b->y_stride = 64; + } else { + b->dst[0] = s->cur_frame->data[0] + yoff; + b->y_stride = s->cur_frame->linesize[0]; + } + if (emu[1]) { + b->dst[1] = s->tmp_uv[0]; + b->dst[2] = s->tmp_uv[1]; + b->uv_stride = 32; + } else { + b->dst[1] = s->cur_frame->data[1] + uvoff; + b->dst[2] = s->cur_frame->data[2] + uvoff; + b->uv_stride = s->cur_frame->linesize[1]; + } + if (b->intra) { + intra_recon(avctx, yoff, uvoff); + } else { + if ((ret = inter_recon(avctx)) < 0) + return ret; + } + if (emu[0]) { + int w = FFMIN(s->cols - col, w4) * 8; + int h = FFMIN(s->rows - row, h4) * 8; + int 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](s->cur_frame->data[0] + yoff + o, + s->tmp_y + o, + s->cur_frame->linesize[0], + 64, h, 0, 0); + o += bw; + } + } + } + if (emu[1]) { + int w = FFMIN(s->cols - col, w4) * 4; + int h = FFMIN(s->rows - row, h4) * 4; + int n, o = 0; + + for (n = 1; o < w; n++) { + int bw = 64 >> n; + + av_assert2(n <= 4); + if (w & bw) { + s->dsp.mc[n][0][0][0][0](s->cur_frame->data[1] + uvoff + o, + s->tmp_uv[0] + o, + s->cur_frame->linesize[1], + 32, h, 0, 0); + s->dsp.mc[n][0][0][0][0](s->cur_frame->data[2] + uvoff + o, + s->tmp_uv[1] + o, + s->cur_frame->linesize[2], + 32, h, 0, 0); + o += bw; + } + } + } + + // pick filter level and find edges to apply filter to + if (s->filter.level && + (lvl = s->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); + int y_end = FFMIN(s->rows - row, h4); + int skip_inter = !b->intra && b->skip; + + for (y = 0; y < h4; y++) + memset(&lflvl->level[((row & 7) + y) * 8 + (col & 7)], lvl, w4); + mask_edges(lflvl, 0, row & 7, col & 7, x_end, y_end, 0, 0, b->tx, skip_inter); + mask_edges(lflvl, 1, row & 7, col & 7, 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.lim_lut[lvl]) { + int sharp = s->filter.sharpness; + int limit = lvl; + + if (sharp > 0) { + limit >>= (sharp + 3) >> 2; + limit = FFMIN(limit, 9 - sharp); + } + limit = FFMAX(limit, 1); + + s->filter.lim_lut[lvl] = limit; + s->filter.mblim_lut[lvl] = 2 * (lvl + 2) + limit; + } + } + + return 0; +} diff --git a/libavcodec/vp9data.c b/libavcodec/vp9data.c new file mode 100644 index 0000000000..374fa8bb8c --- /dev/null +++ b/libavcodec/vp9data.c @@ -0,0 +1,2133 @@ +/* + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "vp9.h" +#include "vp9data.h" + +const int8_t ff_vp9_partition_tree[3][2] = { + { -PARTITION_NONE, 1 }, // '0' + { -PARTITION_H, 2 }, // '10' + { -PARTITION_V, -PARTITION_SPLIT }, // '110', '111' +}; + +const uint8_t ff_vp9_default_kf_partition_probs[4][4][3] = { + { /* 64x64 -> 32x32 */ + { 174, 35, 49 } /* a/l both not split */, + { 68, 11, 27 } /* a split, l not split */, + { 57, 15, 9 } /* l split, a not split */, + { 12, 3, 3 } /* a/l both split */ + }, { /* 32x32 -> 16x16 */ + { 150, 40, 39 } /* a/l both not split */, + { 78, 12, 26 } /* a split, l not split */, + { 67, 33, 11 } /* l split, a not split */, + { 24, 7, 5 } /* a/l both split */, + }, { /* 16x16 -> 8x8 */ + { 149, 53, 53 } /* a/l both not split */, + { 94, 20, 48 } /* a split, l not split */, + { 83, 53, 24 } /* l split, a not split */, + { 52, 18, 18 } /* a/l both split */, + }, { /* 8x8 -> 4x4 */ + { 158, 97, 94 } /* a/l both not split */, + { 93, 24, 99 } /* a split, l not split */, + { 85, 119, 44 } /* l split, a not split */, + { 62, 59, 67 } /* a/l both split */, + }, +}; + +const int8_t ff_vp9_segmentation_tree[7][2] = { + { 1, 2 }, + { 3, 4 }, + { 5, 6 }, + { -0, -1 }, // '00x' + { -2, -3 }, // '01x' + { -4, -5 }, // '10x' + { -6, -7 }, // '11x' +}; + +const int8_t ff_vp9_intramode_tree[9][2] = { + { -DC_PRED, 1 }, // '0' + { -TM_VP8_PRED, 2 }, // '10' + { -VERT_PRED, 3 }, // '110' + { 4, 6 }, + { -HOR_PRED, 5 }, // '11100' + { -DIAG_DOWN_RIGHT_PRED, -VERT_RIGHT_PRED }, // '11101x' + { -DIAG_DOWN_LEFT_PRED, 7 }, // '11110' + { -VERT_LEFT_PRED, 8 }, // '111110' + { -HOR_DOWN_PRED, -HOR_UP_PRED }, // '111111x' +}; + +const uint8_t ff_vp9_default_kf_ymode_probs[10][10][9] = { + { /* above = v */ + { 43, 46, 168, 134, 107, 128, 69, 142, 92 } /* left = v */, + { 44, 29, 68, 159, 201, 177, 50, 57, 77 } /* left = h */, + { 63, 36, 126, 146, 123, 158, 60, 90, 96 } /* left = dc */, + { 58, 38, 76, 114, 97, 172, 78, 133, 92 } /* left = d45 */, + { 46, 41, 76, 140, 63, 184, 69, 112, 57 } /* left = d135 */, + { 38, 32, 85, 140, 46, 112, 54, 151, 133 } /* left = d117 */, + { 39, 27, 61, 131, 110, 175, 44, 75, 136 } /* left = d153 */, + { 47, 35, 80, 100, 74, 143, 64, 163, 74 } /* left = d63 */, + { 52, 30, 74, 113, 130, 175, 51, 64, 58 } /* left = d27 */, + { 36, 61, 116, 114, 128, 162, 80, 125, 82 } /* left = tm */ + }, { /* above = h */ + { 55, 44, 68, 166, 179, 192, 57, 57, 108 } /* left = v */, + { 42, 26, 11, 199, 241, 228, 23, 15, 85 } /* left = h */, + { 82, 26, 26, 171, 208, 204, 44, 32, 105 } /* left = dc */, + { 68, 42, 19, 131, 160, 199, 55, 52, 83 } /* left = d45 */, + { 58, 50, 25, 139, 115, 232, 39, 52, 118 } /* left = d135 */, + { 50, 35, 33, 153, 104, 162, 64, 59, 131 } /* left = d117 */, + { 44, 24, 16, 150, 177, 202, 33, 19, 156 } /* left = d153 */, + { 53, 49, 21, 110, 116, 168, 59, 80, 76 } /* left = d63 */, + { 55, 27, 12, 153, 203, 218, 26, 27, 49 } /* left = d27 */, + { 38, 72, 19, 168, 203, 212, 50, 50, 107 } /* left = tm */ + }, { /* above = dc */ + { 92, 45, 102, 136, 116, 180, 74, 90, 100 } /* left = v */, + { 73, 32, 19, 187, 222, 215, 46, 34, 100 } /* left = h */, + { 137, 30, 42, 148, 151, 207, 70, 52, 91 } /* left = dc */, + { 91, 30, 32, 116, 121, 186, 93, 86, 94 } /* left = d45 */, + { 72, 35, 36, 149, 68, 206, 68, 63, 105 } /* left = d135 */, + { 73, 31, 28, 138, 57, 124, 55, 122, 151 } /* left = d117 */, + { 67, 23, 21, 140, 126, 197, 40, 37, 171 } /* left = d153 */, + { 74, 32, 27, 107, 86, 160, 63, 134, 102 } /* left = d63 */, + { 86, 27, 28, 128, 154, 212, 45, 43, 53 } /* left = d27 */, + { 59, 67, 44, 140, 161, 202, 78, 67, 119 } /* left = tm */ + }, { /* above = d45 */ + { 59, 38, 83, 112, 103, 162, 98, 136, 90 } /* left = v */, + { 62, 30, 23, 158, 200, 207, 59, 57, 50 } /* left = h */, + { 103, 26, 36, 129, 132, 201, 83, 80, 93 } /* left = dc */, + { 67, 30, 29, 84, 86, 191, 102, 91, 59 } /* left = d45 */, + { 60, 32, 33, 112, 71, 220, 64, 89, 104 } /* left = d135 */, + { 53, 26, 34, 130, 56, 149, 84, 120, 103 } /* left = d117 */, + { 53, 21, 23, 133, 109, 210, 56, 77, 172 } /* left = d153 */, + { 61, 29, 29, 93, 97, 165, 83, 175, 162 } /* left = d63 */, + { 77, 19, 29, 112, 142, 228, 55, 66, 36 } /* left = d27 */, + { 47, 47, 43, 114, 137, 181, 100, 99, 95 } /* left = tm */ + }, { /* above = d135 */ + { 53, 40, 55, 139, 69, 183, 61, 80, 110 } /* left = v */, + { 40, 29, 19, 161, 180, 207, 43, 24, 91 } /* left = h */, + { 69, 23, 29, 128, 83, 199, 46, 44, 101 } /* left = dc */, + { 60, 34, 19, 105, 61, 198, 53, 64, 89 } /* left = d45 */, + { 52, 31, 22, 158, 40, 209, 58, 62, 89 } /* left = d135 */, + { 44, 31, 29, 147, 46, 158, 56, 102, 198 } /* left = d117 */, + { 35, 19, 12, 135, 87, 209, 41, 45, 167 } /* left = d153 */, + { 51, 38, 25, 113, 58, 164, 70, 93, 97 } /* left = d63 */, + { 55, 25, 21, 118, 95, 215, 38, 39, 66 } /* left = d27 */, + { 47, 54, 34, 146, 108, 203, 72, 103, 151 } /* left = tm */ + }, { /* above = d117 */ + { 46, 27, 80, 150, 55, 124, 55, 121, 135 } /* left = v */, + { 36, 23, 27, 165, 149, 166, 54, 64, 118 } /* left = h */, + { 64, 19, 37, 156, 66, 138, 49, 95, 133 } /* left = dc */, + { 53, 21, 36, 131, 63, 163, 60, 109, 81 } /* left = d45 */, + { 40, 26, 35, 154, 40, 185, 51, 97, 123 } /* left = d135 */, + { 35, 19, 34, 179, 19, 97, 48, 129, 124 } /* left = d117 */, + { 36, 20, 26, 136, 62, 164, 33, 77, 154 } /* left = d153 */, + { 45, 26, 28, 129, 45, 129, 49, 147, 123 } /* left = d63 */, + { 45, 18, 32, 130, 90, 157, 40, 79, 91 } /* left = d27 */, + { 38, 44, 51, 136, 74, 162, 57, 97, 121 } /* left = tm */ + }, { /* above = d153 */ + { 56, 39, 58, 133, 117, 173, 48, 53, 187 } /* left = v */, + { 35, 21, 12, 161, 212, 207, 20, 23, 145 } /* left = h */, + { 75, 17, 22, 136, 138, 185, 32, 34, 166 } /* left = dc */, + { 56, 29, 19, 117, 109, 181, 55, 68, 112 } /* left = d45 */, + { 47, 29, 17, 153, 64, 220, 59, 51, 114 } /* left = d135 */, + { 46, 16, 24, 136, 76, 147, 41, 64, 172 } /* left = d117 */, + { 34, 17, 11, 108, 152, 187, 13, 15, 209 } /* left = d153 */, + { 55, 30, 18, 122, 79, 179, 44, 88, 116 } /* left = d63 */, + { 51, 24, 14, 115, 133, 209, 32, 26, 104 } /* left = d27 */, + { 37, 49, 25, 129, 168, 164, 41, 54, 148 } /* left = tm */ + }, { /* above = d63 */ + { 48, 34, 86, 101, 92, 146, 78, 179, 134 } /* left = v */, + { 47, 22, 24, 138, 187, 178, 68, 69, 59 } /* left = h */, + { 78, 23, 39, 111, 117, 170, 74, 124, 94 } /* left = dc */, + { 56, 25, 33, 105, 112, 187, 95, 177, 129 } /* left = d45 */, + { 48, 31, 27, 114, 63, 183, 82, 116, 56 } /* left = d135 */, + { 43, 28, 37, 121, 63, 123, 61, 192, 169 } /* left = d117 */, + { 42, 17, 24, 109, 97, 177, 56, 76, 122 } /* left = d153 */, + { 46, 23, 32, 74, 86, 150, 67, 183, 88 } /* left = d63 */, + { 58, 18, 28, 105, 139, 182, 70, 92, 63 } /* left = d27 */, + { 36, 38, 48, 92, 122, 165, 88, 137, 91 } /* left = tm */ + }, { /* above = d27 */ + { 62, 44, 61, 123, 105, 189, 48, 57, 64 } /* left = v */, + { 47, 25, 17, 175, 222, 220, 24, 30, 86 } /* left = h */, + { 82, 22, 32, 127, 143, 213, 39, 41, 70 } /* left = dc */, + { 68, 36, 17, 106, 102, 206, 59, 74, 74 } /* left = d45 */, + { 57, 39, 23, 151, 68, 216, 55, 63, 58 } /* left = d135 */, + { 49, 30, 35, 141, 70, 168, 82, 40, 115 } /* left = d117 */, + { 51, 25, 15, 136, 129, 202, 38, 35, 139 } /* left = d153 */, + { 59, 39, 19, 114, 75, 180, 77, 104, 42 } /* left = d63 */, + { 68, 26, 16, 111, 141, 215, 29, 28, 28 } /* left = d27 */, + { 40, 61, 26, 126, 152, 206, 61, 59, 93 } /* left = tm */ + }, { /* above = tm */ + { 44, 78, 115, 132, 119, 173, 71, 112, 93 } /* left = v */, + { 39, 38, 21, 184, 227, 206, 42, 32, 64 } /* left = h */, + { 65, 70, 60, 155, 159, 199, 61, 60, 81 } /* left = dc */, + { 58, 47, 36, 124, 137, 193, 80, 82, 78 } /* left = d45 */, + { 49, 50, 35, 144, 95, 205, 63, 78, 59 } /* left = d135 */, + { 41, 53, 52, 148, 71, 142, 65, 128, 51 } /* left = d117 */, + { 40, 36, 28, 143, 143, 202, 40, 55, 137 } /* left = d153 */, + { 42, 44, 44, 104, 105, 164, 64, 130, 80 } /* left = d63 */, + { 52, 34, 29, 129, 183, 227, 42, 35, 43 } /* left = d27 */, + { 43, 81, 53, 140, 169, 204, 68, 84, 72 } /* left = tm */ + } +}; + +const uint8_t ff_vp9_default_kf_uvmode_probs[10][9] = { + { 118, 15, 123, 148, 131, 101, 44, 93, 131 } /* y = v */, + { 113, 12, 23, 188, 226, 142, 26, 32, 125 } /* y = h */, + { 144, 11, 54, 157, 195, 130, 46, 58, 108 } /* y = dc */, + { 120, 11, 50, 123, 163, 135, 64, 77, 103 } /* y = d45 */, + { 113, 9, 36, 155, 111, 157, 32, 44, 161 } /* y = d135 */, + { 116, 9, 55, 176, 76, 96, 37, 61, 149 } /* y = d117 */, + { 115, 9, 28, 141, 161, 167, 21, 25, 193 } /* y = d153 */, + { 116, 12, 64, 120, 140, 125, 49, 115, 121 } /* y = d63 */, + { 120, 12, 32, 145, 195, 142, 32, 38, 86 } /* y = d27 */, + { 102, 19, 66, 162, 182, 122, 35, 59, 128 } /* y = tm */ +}; + +const int8_t ff_vp9_inter_mode_tree[3][2] = { + { -ZEROMV, 1 }, // '0' + { -NEARESTMV, 2 }, // '10' + { -NEARMV, -NEWMV }, // '11x' +}; + +const int8_t ff_vp9_filter_tree[2][2] = { + { -0, 1 }, // '0' + { -1, -2 }, // '1x' +}; + +const enum FilterMode ff_vp9_filter_lut[3] = { + FILTER_8TAP_REGULAR, + FILTER_8TAP_SMOOTH, + FILTER_8TAP_SHARP, +}; + +const int16_t ff_vp9_dc_qlookup[256] = { + 4, 8, 8, 9, 10, 11, 12, 12, + 13, 14, 15, 16, 17, 18, 19, 19, + 20, 21, 22, 23, 24, 25, 26, 26, + 27, 28, 29, 30, 31, 32, 32, 33, + 34, 35, 36, 37, 38, 38, 39, 40, + 41, 42, 43, 43, 44, 45, 46, 47, + 48, 48, 49, 50, 51, 52, 53, 53, + 54, 55, 56, 57, 57, 58, 59, 60, + 61, 62, 62, 63, 64, 65, 66, 66, + 67, 68, 69, 70, 70, 71, 72, 73, + 74, 74, 75, 76, 77, 78, 78, 79, + 80, 81, 81, 82, 83, 84, 85, 85, + 87, 88, 90, 92, 93, 95, 96, 98, + 99, 101, 102, 104, 105, 107, 108, 110, + 111, 113, 114, 116, 117, 118, 120, 121, + 123, 125, 127, 129, 131, 134, 136, 138, + 140, 142, 144, 146, 148, 150, 152, 154, + 156, 158, 161, 164, 166, 169, 172, 174, + 177, 180, 182, 185, 187, 190, 192, 195, + 199, 202, 205, 208, 211, 214, 217, 220, + 223, 226, 230, 233, 237, 240, 243, 247, + 250, 253, 257, 261, 265, 269, 272, 276, + 280, 284, 288, 292, 296, 300, 304, 309, + 313, 317, 322, 326, 330, 335, 340, 344, + 349, 354, 359, 364, 369, 374, 379, 384, + 389, 395, 400, 406, 411, 417, 423, 429, + 435, 441, 447, 454, 461, 467, 475, 482, + 489, 497, 505, 513, 522, 530, 539, 549, + 559, 569, 579, 590, 602, 614, 626, 640, + 654, 668, 684, 700, 717, 736, 755, 775, + 796, 819, 843, 869, 896, 925, 955, 988, + 1022, 1058, 1098, 1139, 1184, 1232, 1282, 1336, +}; + +const int16_t ff_vp9_ac_qlookup[256] = { + 4, 8, 9, 10, 11, 12, 13, 14, + 15, 16, 17, 18, 19, 20, 21, 22, + 23, 24, 25, 26, 27, 28, 29, 30, + 31, 32, 33, 34, 35, 36, 37, 38, + 39, 40, 41, 42, 43, 44, 45, 46, + 47, 48, 49, 50, 51, 52, 53, 54, + 55, 56, 57, 58, 59, 60, 61, 62, + 63, 64, 65, 66, 67, 68, 69, 70, + 71, 72, 73, 74, 75, 76, 77, 78, + 79, 80, 81, 82, 83, 84, 85, 86, + 87, 88, 89, 90, 91, 92, 93, 94, + 95, 96, 97, 98, 99, 100, 101, 102, + 104, 106, 108, 110, 112, 114, 116, 118, + 120, 122, 124, 126, 128, 130, 132, 134, + 136, 138, 140, 142, 144, 146, 148, 150, + 152, 155, 158, 161, 164, 167, 170, 173, + 176, 179, 182, 185, 188, 191, 194, 197, + 200, 203, 207, 211, 215, 219, 223, 227, + 231, 235, 239, 243, 247, 251, 255, 260, + 265, 270, 275, 280, 285, 290, 295, 300, + 305, 311, 317, 323, 329, 335, 341, 347, + 353, 359, 366, 373, 380, 387, 394, 401, + 408, 416, 424, 432, 440, 448, 456, 465, + 474, 483, 492, 501, 510, 520, 530, 540, + 550, 560, 571, 582, 593, 604, 615, 627, + 639, 651, 663, 676, 689, 702, 715, 729, + 743, 757, 771, 786, 801, 816, 832, 848, + 864, 881, 898, 915, 933, 951, 969, 988, + 1007, 1026, 1046, 1066, 1087, 1108, 1129, 1151, + 1173, 1196, 1219, 1243, 1267, 1292, 1317, 1343, + 1369, 1396, 1423, 1451, 1479, 1508, 1537, 1567, + 1597, 1628, 1660, 1692, 1725, 1759, 1793, 1828, +}; + +const enum TxfmType ff_vp9_intra_txfm_type[14] = { + [VERT_PRED] = ADST_DCT, + [HOR_PRED] = DCT_ADST, + [DC_PRED] = DCT_DCT, + [DIAG_DOWN_LEFT_PRED] = DCT_DCT, + [DIAG_DOWN_RIGHT_PRED] = ADST_ADST, + [VERT_RIGHT_PRED] = ADST_DCT, + [HOR_DOWN_PRED] = DCT_ADST, + [VERT_LEFT_PRED] = ADST_DCT, + [HOR_UP_PRED] = DCT_ADST, + [TM_VP8_PRED] = ADST_ADST, + [NEARESTMV] = DCT_DCT, + [NEARMV] = DCT_DCT, + [ZEROMV] = DCT_DCT, + [NEWMV] = DCT_DCT, +}; + +const int16_t ff_vp9_default_scan_4x4[16] = { + 0, 1, 4, 5, + 2, 8, 3, 6, + 12, 9, 7, 10, + 13, 11, 14, 15, +}; + +const int16_t ff_vp9_col_scan_4x4[16] = { + 0, 1, 2, 4, + 3, 5, 6, 8, + 7, 9, 10, 12, + 13, 11, 14, 15, +}; + +const int16_t ff_vp9_row_scan_4x4[16] = { + 0, 4, 1, 8, + 5, 12, 9, 2, + 6, 13, 3, 10, + 7, 14, 11, 15, +}; + +const int16_t ff_vp9_default_scan_8x8[64] = { + 0, 1, 8, 2, 9, 16, 10, 3, + 17, 24, 18, 11, 4, 25, 32, 19, + 12, 26, 5, 33, 20, 27, 40, 13, + 34, 6, 41, 28, 21, 35, 42, 48, + 14, 7, 36, 29, 43, 56, 49, 22, + 15, 37, 50, 44, 57, 30, 23, 51, + 45, 58, 38, 31, 52, 59, 39, 46, + 53, 60, 47, 54, 61, 55, 62, 63, +}; + +const int16_t ff_vp9_col_scan_8x8[64] = { + 0, 1, 2, 8, 3, 9, 4, 10, + 16, 5, 11, 17, 12, 18, 6, 24, + 19, 13, 25, 7, 26, 20, 32, 14, + 27, 21, 33, 28, 34, 15, 22, 35, + 40, 29, 41, 36, 23, 30, 42, 37, + 48, 43, 31, 44, 49, 38, 50, 56, + 45, 39, 51, 57, 52, 46, 58, 53, + 59, 47, 60, 54, 61, 55, 62, 63, +}; + +const int16_t ff_vp9_row_scan_8x8[64] = { + 0, 8, 16, 1, 9, 24, 2, 17, + 32, 10, 25, 3, 40, 18, 11, 33, + 26, 19, 4, 48, 41, 34, 12, 27, + 56, 20, 5, 42, 35, 13, 49, 28, + 6, 21, 43, 36, 14, 50, 29, 57, + 7, 44, 22, 37, 51, 15, 58, 30, + 23, 45, 52, 38, 59, 31, 46, 53, + 39, 60, 47, 61, 54, 62, 55, 63, +}; + +const int16_t ff_vp9_default_scan_16x16[256] = { + 0, 1, 16, 2, 17, 32, 3, 18, 33, 48, 4, 34, 19, 49, 20, 5, + 35, 64, 50, 36, 65, 21, 6, 51, 80, 66, 37, 22, 52, 7, 81, 67, + 38, 82, 53, 23, 96, 68, 8, 83, 97, 54, 39, 69, 112, 24, 98, 84, + 70, 55, 9, 40, 85, 99, 113, 128, 25, 114, 100, 71, 86, 56, 10, 41, + 115, 101, 129, 116, 72, 87, 26, 130, 144, 102, 57, 11, 42, 117, 131, 145, + 88, 103, 27, 73, 132, 118, 146, 58, 160, 12, 43, 133, 147, 104, 89, 119, + 161, 74, 148, 134, 28, 162, 59, 13, 176, 120, 149, 90, 135, 105, 163, 44, + 75, 177, 164, 29, 150, 121, 136, 178, 165, 14, 106, 60, 91, 151, 45, 179, + 192, 137, 166, 122, 76, 180, 152, 30, 61, 15, 107, 167, 181, 193, 92, 208, + 46, 138, 123, 153, 194, 77, 168, 182, 31, 195, 209, 183, 108, 139, 62, 154, + 47, 196, 93, 169, 210, 197, 224, 124, 184, 211, 78, 109, 170, 155, 63, 198, + 212, 185, 225, 240, 140, 94, 199, 125, 79, 213, 226, 171, 186, 156, 214, 200, + 110, 227, 141, 95, 241, 215, 228, 201, 126, 242, 187, 172, 157, 229, 111, 216, + 243, 142, 202, 230, 127, 217, 244, 173, 188, 231, 158, 203, 143, 245, 218, 232, + 189, 246, 159, 174, 233, 247, 219, 204, 175, 190, 248, 234, 205, 220, 249, 191, + 235, 221, 250, 206, 222, 251, 236, 207, 237, 223, 252, 238, 253, 239, 254, 255, +}; + +const int16_t ff_vp9_col_scan_16x16[256] = { + 0, 1, 2, 3, 16, 4, 17, 5, 18, 6, 19, 32, 20, 7, 33, 21, + 34, 8, 35, 22, 48, 36, 9, 49, 23, 50, 37, 10, 38, 51, 24, 64, + 52, 11, 65, 39, 25, 53, 66, 54, 40, 67, 12, 80, 26, 68, 55, 81, + 41, 69, 13, 27, 82, 56, 70, 83, 42, 14, 84, 96, 71, 28, 57, 85, + 97, 15, 72, 98, 43, 86, 58, 99, 29, 87, 100, 112, 73, 44, 101, 59, + 30, 113, 88, 114, 74, 128, 102, 45, 31, 115, 60, 103, 89, 116, 75, 129, + 117, 46, 104, 90, 61, 130, 118, 131, 132, 105, 76, 47, 119, 144, 91, 62, + 133, 106, 145, 120, 146, 134, 77, 147, 121, 92, 135, 148, 63, 107, 136, 122, + 93, 149, 160, 78, 150, 137, 108, 161, 162, 151, 123, 79, 138, 163, 152, 94, + 164, 109, 165, 153, 124, 139, 176, 166, 95, 177, 167, 110, 154, 178, 125, 179, + 140, 168, 155, 111, 180, 192, 181, 169, 141, 126, 182, 193, 194, 156, 183, 170, + 195, 127, 142, 196, 184, 208, 197, 157, 171, 143, 185, 198, 209, 199, 210, 172, + 158, 186, 211, 224, 212, 200, 240, 159, 213, 225, 187, 201, 173, 226, 214, 215, + 227, 202, 228, 188, 241, 216, 174, 229, 242, 203, 243, 217, 230, 175, 189, 244, + 231, 204, 218, 232, 245, 219, 246, 190, 233, 205, 191, 247, 234, 248, 220, 206, + 249, 235, 221, 207, 250, 236, 222, 251, 223, 237, 238, 252, 239, 253, 254, 255, +}; + +const int16_t ff_vp9_row_scan_16x16[256] = { + 0, 16, 32, 1, 48, 17, 64, 33, 2, 80, 18, 49, 96, 34, 3, 65, + 19, 112, 50, 81, 35, 4, 128, 66, 20, 97, 51, 82, 5, 144, 36, 67, + 113, 98, 21, 52, 160, 83, 129, 37, 68, 6, 114, 176, 99, 53, 22, 84, + 145, 38, 69, 130, 7, 115, 192, 100, 54, 23, 85, 161, 146, 131, 39, 70, + 208, 116, 8, 101, 177, 55, 86, 24, 162, 147, 132, 71, 224, 117, 40, 102, + 9, 148, 56, 87, 193, 163, 240, 133, 178, 25, 118, 72, 41, 103, 164, 10, + 149, 88, 134, 209, 179, 57, 119, 194, 26, 73, 165, 150, 104, 42, 135, 11, + 180, 120, 89, 225, 195, 58, 27, 210, 151, 181, 166, 74, 43, 105, 12, 136, + 90, 59, 241, 121, 28, 196, 167, 211, 152, 44, 182, 137, 75, 13, 226, 106, + 122, 60, 197, 91, 168, 29, 183, 153, 14, 76, 212, 138, 45, 107, 15, 198, + 92, 227, 169, 30, 123, 154, 61, 242, 184, 213, 139, 46, 77, 31, 108, 170, + 199, 185, 124, 228, 93, 155, 214, 62, 140, 243, 78, 47, 200, 109, 186, 171, + 201, 94, 63, 215, 229, 156, 79, 125, 141, 110, 216, 187, 172, 244, 202, 230, + 217, 95, 157, 126, 245, 111, 142, 231, 188, 127, 158, 218, 173, 232, 246, 233, + 203, 143, 247, 174, 189, 159, 219, 204, 248, 234, 249, 175, 190, 220, 205, 250, + 235, 191, 221, 251, 236, 206, 252, 222, 207, 237, 223, 253, 238, 254, 239, 255, +}; + +const int16_t ff_vp9_default_scan_32x32[1024] = { + 0, 1, 32, 2, 33, 64, 3, 34, 65, 4, 96, 35, 66, 5, 36, 97, + 67, 128, 98, 68, 37, 6, 129, 99, 7, 160, 69, 38, 130, 100, 161, 131, + 39, 70, 8, 101, 162, 132, 192, 71, 40, 9, 102, 163, 133, 193, 72, 224, + 103, 41, 164, 10, 194, 134, 165, 73, 104, 135, 225, 42, 195, 11, 256, 166, + 226, 196, 74, 105, 136, 43, 12, 167, 197, 227, 257, 75, 106, 137, 228, 44, + 198, 168, 258, 288, 13, 229, 76, 107, 199, 138, 259, 169, 289, 45, 230, 260, + 200, 108, 14, 170, 139, 320, 290, 77, 231, 261, 46, 201, 140, 291, 109, 232, + 321, 262, 171, 78, 292, 15, 322, 202, 263, 352, 172, 293, 233, 141, 323, 110, + 47, 203, 264, 234, 294, 353, 324, 16, 79, 204, 265, 295, 325, 173, 354, 142, + 235, 384, 48, 296, 111, 266, 355, 326, 80, 17, 205, 236, 174, 356, 385, 327, + 143, 297, 267, 357, 386, 112, 49, 328, 298, 206, 416, 237, 358, 387, 81, 175, + 18, 329, 359, 388, 299, 330, 389, 113, 417, 238, 360, 50, 207, 418, 390, 331, + 19, 448, 361, 82, 419, 391, 239, 51, 362, 420, 114, 449, 480, 421, 83, 363, + 450, 422, 512, 451, 423, 115, 452, 481, 453, 482, 454, 544, 483, 455, 513, 484, + 514, 485, 515, 486, 545, 576, 487, 546, 547, 608, 577, 578, 579, 609, 610, 611, + 20, 144, 268, 392, 516, 640, 21, 52, 145, 176, 269, 300, 393, 424, 517, 548, + 641, 672, 22, 53, 84, 146, 177, 208, 270, 301, 332, 394, 425, 456, 518, 549, + 580, 642, 673, 704, 23, 54, 85, 116, 147, 178, 209, 240, 271, 302, 333, 364, + 395, 426, 457, 488, 519, 550, 581, 612, 643, 674, 705, 736, 55, 86, 117, 179, + 210, 241, 303, 334, 365, 427, 458, 489, 551, 582, 613, 675, 706, 737, 87, 118, + 211, 242, 335, 366, 459, 490, 583, 614, 707, 738, 119, 243, 367, 491, 615, 739, + 24, 148, 272, 396, 520, 644, 768, 25, 56, 149, 180, 273, 304, 397, 428, 521, + 552, 645, 676, 769, 800, 26, 57, 88, 150, 181, 212, 274, 305, 336, 398, 429, + 460, 522, 553, 584, 646, 677, 708, 770, 801, 832, 27, 58, 89, 120, 151, 182, + 213, 244, 275, 306, 337, 368, 399, 430, 461, 492, 523, 554, 585, 616, 647, 678, + 709, 740, 771, 802, 833, 864, 59, 90, 121, 183, 214, 245, 307, 338, 369, 431, + 462, 493, 555, 586, 617, 679, 710, 741, 803, 834, 865, 91, 122, 215, 246, 339, + 370, 463, 494, 587, 618, 711, 742, 835, 866, 123, 247, 371, 495, 619, 743, 867, + 28, 152, 276, 400, 524, 648, 772, 896, 29, 60, 153, 184, 277, 308, 401, 432, + 525, 556, 649, 680, 773, 804, 897, 928, 30, 61, 92, 154, 185, 216, 278, 309, + 340, 402, 433, 464, 526, 557, 588, 650, 681, 712, 774, 805, 836, 898, 929, 960, + 31, 62, 93, 124, 155, 186, 217, 248, 279, 310, 341, 372, 403, 434, 465, 496, + 527, 558, 589, 620, 651, 682, 713, 744, 775, 806, 837, 868, 899, 930, 961, 992, + 63, 94, 125, 187, 218, 249, 311, 342, 373, 435, 466, 497, 559, 590, 621, 683, + 714, 745, 807, 838, 869, 931, 962, 993, 95, 126, 219, 250, 343, 374, 467, 498, + 591, 622, 715, 746, 839, 870, 963, 994, 127, 251, 375, 499, 623, 747, 871, 995, + 156, 280, 404, 528, 652, 776, 900, 157, 188, 281, 312, 405, 436, 529, 560, 653, + 684, 777, 808, 901, 932, 158, 189, 220, 282, 313, 344, 406, 437, 468, 530, 561, + 592, 654, 685, 716, 778, 809, 840, 902, 933, 964, 159, 190, 221, 252, 283, 314, + 345, 376, 407, 438, 469, 500, 531, 562, 593, 624, 655, 686, 717, 748, 779, 810, + 841, 872, 903, 934, 965, 996, 191, 222, 253, 315, 346, 377, 439, 470, 501, 563, + 594, 625, 687, 718, 749, 811, 842, 873, 935, 966, 997, 223, 254, 347, 378, 471, + 502, 595, 626, 719, 750, 843, 874, 967, 998, 255, 379, 503, 627, 751, 875, 999, + 284, 408, 532, 656, 780, 904, 285, 316, 409, 440, 533, 564, 657, 688, 781, 812, + 905, 936, 286, 317, 348, 410, 441, 472, 534, 565, 596, 658, 689, 720, 782, 813, + 844, 906, 937, 968, 287, 318, 349, 380, 411, 442, 473, 504, 535, 566, 597, 628, + 659, 690, 721, 752, 783, 814, 845, 876, 907, 938, 969, 1000, 319, 350, 381, 443, + 474, 505, 567, 598, 629, 691, 722, 753, 815, 846, 877, 939, 970, 1001, 351, 382, + 475, 506, 599, 630, 723, 754, 847, 878, 971, 1002, 383, 507, 631, 755, 879, 1003, + 412, 536, 660, 784, 908, 413, 444, 537, 568, 661, 692, 785, 816, 909, 940, 414, + 445, 476, 538, 569, 600, 662, 693, 724, 786, 817, 848, 910, 941, 972, 415, 446, + 477, 508, 539, 570, 601, 632, 663, 694, 725, 756, 787, 818, 849, 880, 911, 942, + 973, 1004, 447, 478, 509, 571, 602, 633, 695, 726, 757, 819, 850, 881, 943, 974, + 1005, 479, 510, 603, 634, 727, 758, 851, 882, 975, 1006, 511, 635, 759, 883, 1007, + 540, 664, 788, 912, 541, 572, 665, 696, 789, 820, 913, 944, 542, 573, 604, 666, + 697, 728, 790, 821, 852, 914, 945, 976, 543, 574, 605, 636, 667, 698, 729, 760, + 791, 822, 853, 884, 915, 946, 977, 1008, 575, 606, 637, 699, 730, 761, 823, 854, + 885, 947, 978, 1009, 607, 638, 731, 762, 855, 886, 979, 1010, 639, 763, 887, 1011, + 668, 792, 916, 669, 700, 793, 824, 917, 948, 670, 701, 732, 794, 825, 856, 918, + 949, 980, 671, 702, 733, 764, 795, 826, 857, 888, 919, 950, 981, 1012, 703, 734, + 765, 827, 858, 889, 951, 982, 1013, 735, 766, 859, 890, 983, 1014, 767, 891, 1015, + 796, 920, 797, 828, 921, 952, 798, 829, 860, 922, 953, 984, 799, 830, 861, 892, + 923, 954, 985, 1016, 831, 862, 893, 955, 986, 1017, 863, 894, 987, 1018, 895, 1019, + 924, 925, 956, 926, 957, 988, 927, 958, 989, 1020, 959, 990, 1021, 991, 1022, 1023, +}; + +const int16_t *ff_vp9_scans[5][4] = { + { + ff_vp9_default_scan_4x4, ff_vp9_col_scan_4x4, + ff_vp9_row_scan_4x4, ff_vp9_default_scan_4x4 + }, { + ff_vp9_default_scan_8x8, ff_vp9_col_scan_8x8, + ff_vp9_row_scan_8x8, ff_vp9_default_scan_8x8 + }, { + ff_vp9_default_scan_16x16, ff_vp9_col_scan_16x16, + ff_vp9_row_scan_16x16, ff_vp9_default_scan_16x16 + }, { + ff_vp9_default_scan_32x32, ff_vp9_default_scan_32x32, + ff_vp9_default_scan_32x32, ff_vp9_default_scan_32x32 + }, { // lossless + ff_vp9_default_scan_4x4, ff_vp9_default_scan_4x4, + ff_vp9_default_scan_4x4, ff_vp9_default_scan_4x4 + } +}; + +const int16_t ff_vp9_default_scan_4x4_nb[16][2] = { + { 0, 0 }, { 0, 0 }, { 4, 1 }, { 1, 1 }, + { 4, 4 }, { 2, 2 }, { 5, 2 }, { 8, 8 }, + { 8, 5 }, { 6, 3 }, { 9, 6 }, { 12, 9 }, + { 10, 7 }, { 13, 10 }, { 14, 11 }, { 0, 0 }, +}; + +const int16_t ff_vp9_col_scan_4x4_nb[16][2] = { + { 0, 0 }, { 1, 1 }, { 0, 0 }, { 2, 2 }, + { 4, 4 }, { 5, 5 }, { 4, 4 }, { 6, 6 }, + { 8, 8 }, { 9, 9 }, { 8, 8 }, { 12, 12 }, + { 10, 10 }, { 13, 13 }, { 14, 14 }, { 0, 0 }, +}; + +const int16_t ff_vp9_row_scan_4x4_nb[16][2] = { + { 0, 0 }, { 0, 0 }, { 4, 4 }, { 1, 1 }, + { 8, 8 }, { 5, 5 }, { 1, 1 }, { 2, 2 }, + { 9, 9 }, { 2, 2 }, { 6, 6 }, { 3, 3 }, + { 10, 10 }, { 7, 7 }, { 11, 11 }, { 0, 0 }, +}; + +const int16_t ff_vp9_default_scan_8x8_nb[64][2] = { + { 0, 0 }, { 0, 0 }, { 1, 1 }, { 8, 1 }, + { 8, 8 }, { 9, 2 }, { 2, 2 }, { 16, 9 }, + { 16, 16 }, { 17, 10 }, { 10, 3 }, { 3, 3 }, + { 24, 17 }, { 24, 24 }, { 18, 11 }, { 11, 4 }, + { 25, 18 }, { 4, 4 }, { 32, 25 }, { 19, 12 }, + { 26, 19 }, { 32, 32 }, { 12, 5 }, { 33, 26 }, + { 5, 5 }, { 40, 33 }, { 27, 20 }, { 20, 13 }, + { 34, 27 }, { 41, 34 }, { 40, 40 }, { 13, 6 }, + { 6, 6 }, { 35, 28 }, { 28, 21 }, { 42, 35 }, + { 48, 48 }, { 48, 41 }, { 21, 14 }, { 14, 7 }, + { 36, 29 }, { 49, 42 }, { 43, 36 }, { 56, 49 }, + { 29, 22 }, { 22, 15 }, { 50, 43 }, { 44, 37 }, + { 57, 50 }, { 37, 30 }, { 30, 23 }, { 51, 44 }, + { 58, 51 }, { 38, 31 }, { 45, 38 }, { 52, 45 }, + { 59, 52 }, { 46, 39 }, { 53, 46 }, { 60, 53 }, + { 54, 47 }, { 61, 54 }, { 62, 55 }, { 0, 0 }, +}; + +const int16_t ff_vp9_col_scan_8x8_nb[64][2] = { + { 0, 0 }, { 1, 1 }, { 0, 0 }, { 2, 2 }, + { 8, 8 }, { 3, 3 }, { 9, 9 }, { 8, 8 }, + { 4, 4 }, { 10, 10 }, { 16, 16 }, { 11, 11 }, + { 17, 17 }, { 5, 5 }, { 16, 16 }, { 18, 18 }, + { 12, 12 }, { 24, 24 }, { 6, 6 }, { 25, 25 }, + { 19, 19 }, { 24, 24 }, { 13, 13 }, { 26, 26 }, + { 20, 20 }, { 32, 32 }, { 27, 27 }, { 33, 33 }, + { 14, 14 }, { 21, 21 }, { 34, 34 }, { 32, 32 }, + { 28, 28 }, { 40, 40 }, { 35, 35 }, { 22, 22 }, + { 29, 29 }, { 41, 41 }, { 36, 36 }, { 40, 40 }, + { 42, 42 }, { 30, 30 }, { 43, 43 }, { 48, 48 }, + { 37, 37 }, { 49, 49 }, { 48, 48 }, { 44, 44 }, + { 38, 38 }, { 50, 50 }, { 56, 56 }, { 51, 51 }, + { 45, 45 }, { 57, 57 }, { 52, 52 }, { 58, 58 }, + { 46, 46 }, { 59, 59 }, { 53, 53 }, { 60, 60 }, + { 54, 54 }, { 61, 61 }, { 62, 62 }, { 0, 0 }, +}; + +const int16_t ff_vp9_row_scan_8x8_nb[64][2] = { + { 0, 0 }, { 8, 8 }, { 0, 0 }, { 1, 1 }, + { 16, 16 }, { 1, 1 }, { 9, 9 }, { 24, 24 }, + { 2, 2 }, { 17, 17 }, { 2, 2 }, { 32, 32 }, + { 10, 10 }, { 3, 3 }, { 25, 25 }, { 18, 18 }, + { 11, 11 }, { 3, 3 }, { 40, 40 }, { 33, 33 }, + { 26, 26 }, { 4, 4 }, { 19, 19 }, { 48, 48 }, + { 12, 12 }, { 4, 4 }, { 34, 34 }, { 27, 27 }, + { 5, 5 }, { 41, 41 }, { 20, 20 }, { 5, 5 }, + { 13, 13 }, { 35, 35 }, { 28, 28 }, { 6, 6 }, + { 42, 42 }, { 21, 21 }, { 49, 49 }, { 6, 6 }, + { 36, 36 }, { 14, 14 }, { 29, 29 }, { 43, 43 }, + { 7, 7 }, { 50, 50 }, { 22, 22 }, { 15, 15 }, + { 37, 37 }, { 44, 44 }, { 30, 30 }, { 51, 51 }, + { 23, 23 }, { 38, 38 }, { 45, 45 }, { 31, 31 }, + { 52, 52 }, { 39, 39 }, { 53, 53 }, { 46, 46 }, + { 54, 54 }, { 47, 47 }, { 55, 55 }, { 0, 0 }, +}; + +const int16_t ff_vp9_default_scan_16x16_nb[256][2] = { + { 0, 0 }, { 0, 0 }, { 1, 1 }, { 16, 1 }, + { 16, 16 }, { 2, 2 }, { 17, 2 }, { 32, 17 }, + { 32, 32 }, { 3, 3 }, { 33, 18 }, { 18, 3 }, + { 48, 33 }, { 19, 4 }, { 4, 4 }, { 34, 19 }, + { 48, 48 }, { 49, 34 }, { 35, 20 }, { 64, 49 }, + { 20, 5 }, { 5, 5 }, { 50, 35 }, { 64, 64 }, + { 65, 50 }, { 36, 21 }, { 21, 6 }, { 51, 36 }, + { 6, 6 }, { 80, 65 }, { 66, 51 }, { 37, 22 }, + { 81, 66 }, { 52, 37 }, { 22, 7 }, { 80, 80 }, + { 67, 52 }, { 7, 7 }, { 82, 67 }, { 96, 81 }, + { 53, 38 }, { 38, 23 }, { 68, 53 }, { 96, 96 }, + { 23, 8 }, { 97, 82 }, { 83, 68 }, { 69, 54 }, + { 54, 39 }, { 8, 8 }, { 39, 24 }, { 84, 69 }, + { 98, 83 }, { 112, 97 }, { 112, 112 }, { 24, 9 }, + { 113, 98 }, { 99, 84 }, { 70, 55 }, { 85, 70 }, + { 55, 40 }, { 9, 9 }, { 40, 25 }, { 114, 99 }, + { 100, 85 }, { 128, 113 }, { 115, 100 }, { 71, 56 }, + { 86, 71 }, { 25, 10 }, { 129, 114 }, { 128, 128 }, + { 101, 86 }, { 56, 41 }, { 10, 10 }, { 41, 26 }, + { 116, 101 }, { 130, 115 }, { 144, 129 }, { 87, 72 }, + { 102, 87 }, { 26, 11 }, { 72, 57 }, { 131, 116 }, + { 117, 102 }, { 145, 130 }, { 57, 42 }, { 144, 144 }, + { 11, 11 }, { 42, 27 }, { 132, 117 }, { 146, 131 }, + { 103, 88 }, { 88, 73 }, { 118, 103 }, { 160, 145 }, + { 73, 58 }, { 147, 132 }, { 133, 118 }, { 27, 12 }, + { 161, 146 }, { 58, 43 }, { 12, 12 }, { 160, 160 }, + { 119, 104 }, { 148, 133 }, { 89, 74 }, { 134, 119 }, + { 104, 89 }, { 162, 147 }, { 43, 28 }, { 74, 59 }, + { 176, 161 }, { 163, 148 }, { 28, 13 }, { 149, 134 }, + { 120, 105 }, { 135, 120 }, { 177, 162 }, { 164, 149 }, + { 13, 13 }, { 105, 90 }, { 59, 44 }, { 90, 75 }, + { 150, 135 }, { 44, 29 }, { 178, 163 }, { 176, 176 }, + { 136, 121 }, { 165, 150 }, { 121, 106 }, { 75, 60 }, + { 179, 164 }, { 151, 136 }, { 29, 14 }, { 60, 45 }, + { 14, 14 }, { 106, 91 }, { 166, 151 }, { 180, 165 }, + { 192, 177 }, { 91, 76 }, { 192, 192 }, { 45, 30 }, + { 137, 122 }, { 122, 107 }, { 152, 137 }, { 193, 178 }, + { 76, 61 }, { 167, 152 }, { 181, 166 }, { 30, 15 }, + { 194, 179 }, { 208, 193 }, { 182, 167 }, { 107, 92 }, + { 138, 123 }, { 61, 46 }, { 153, 138 }, { 46, 31 }, + { 195, 180 }, { 92, 77 }, { 168, 153 }, { 209, 194 }, + { 196, 181 }, { 208, 208 }, { 123, 108 }, { 183, 168 }, + { 210, 195 }, { 77, 62 }, { 108, 93 }, { 169, 154 }, + { 154, 139 }, { 62, 47 }, { 197, 182 }, { 211, 196 }, + { 184, 169 }, { 224, 209 }, { 224, 224 }, { 139, 124 }, + { 93, 78 }, { 198, 183 }, { 124, 109 }, { 78, 63 }, + { 212, 197 }, { 225, 210 }, { 170, 155 }, { 185, 170 }, + { 155, 140 }, { 213, 198 }, { 199, 184 }, { 109, 94 }, + { 226, 211 }, { 140, 125 }, { 94, 79 }, { 240, 225 }, + { 214, 199 }, { 227, 212 }, { 200, 185 }, { 125, 110 }, + { 241, 226 }, { 186, 171 }, { 171, 156 }, { 156, 141 }, + { 228, 213 }, { 110, 95 }, { 215, 200 }, { 242, 227 }, + { 141, 126 }, { 201, 186 }, { 229, 214 }, { 126, 111 }, + { 216, 201 }, { 243, 228 }, { 172, 157 }, { 187, 172 }, + { 230, 215 }, { 157, 142 }, { 202, 187 }, { 142, 127 }, + { 244, 229 }, { 217, 202 }, { 231, 216 }, { 188, 173 }, + { 245, 230 }, { 158, 143 }, { 173, 158 }, { 232, 217 }, + { 246, 231 }, { 218, 203 }, { 203, 188 }, { 174, 159 }, + { 189, 174 }, { 247, 232 }, { 233, 218 }, { 204, 189 }, + { 219, 204 }, { 248, 233 }, { 190, 175 }, { 234, 219 }, + { 220, 205 }, { 249, 234 }, { 205, 190 }, { 221, 206 }, + { 250, 235 }, { 235, 220 }, { 206, 191 }, { 236, 221 }, + { 222, 207 }, { 251, 236 }, { 237, 222 }, { 252, 237 }, + { 238, 223 }, { 253, 238 }, { 254, 239 }, { 0, 0 }, +}; + +const int16_t ff_vp9_col_scan_16x16_nb[256][2] = { + { 0, 0 }, { 1, 1 }, { 2, 2 }, { 0, 0 }, + { 3, 3 }, { 16, 16 }, { 4, 4 }, { 17, 17 }, + { 5, 5 }, { 18, 18 }, { 16, 16 }, { 19, 19 }, + { 6, 6 }, { 32, 32 }, { 20, 20 }, { 33, 33 }, + { 7, 7 }, { 34, 34 }, { 21, 21 }, { 32, 32 }, + { 35, 35 }, { 8, 8 }, { 48, 48 }, { 22, 22 }, + { 49, 49 }, { 36, 36 }, { 9, 9 }, { 37, 37 }, + { 50, 50 }, { 23, 23 }, { 48, 48 }, { 51, 51 }, + { 10, 10 }, { 64, 64 }, { 38, 38 }, { 24, 24 }, + { 52, 52 }, { 65, 65 }, { 53, 53 }, { 39, 39 }, + { 66, 66 }, { 11, 11 }, { 64, 64 }, { 25, 25 }, + { 67, 67 }, { 54, 54 }, { 80, 80 }, { 40, 40 }, + { 68, 68 }, { 12, 12 }, { 26, 26 }, { 81, 81 }, + { 55, 55 }, { 69, 69 }, { 82, 82 }, { 41, 41 }, + { 13, 13 }, { 83, 83 }, { 80, 80 }, { 70, 70 }, + { 27, 27 }, { 56, 56 }, { 84, 84 }, { 96, 96 }, + { 14, 14 }, { 71, 71 }, { 97, 97 }, { 42, 42 }, + { 85, 85 }, { 57, 57 }, { 98, 98 }, { 28, 28 }, + { 86, 86 }, { 99, 99 }, { 96, 96 }, { 72, 72 }, + { 43, 43 }, { 100, 100 }, { 58, 58 }, { 29, 29 }, + { 112, 112 }, { 87, 87 }, { 113, 113 }, { 73, 73 }, + { 112, 112 }, { 101, 101 }, { 44, 44 }, { 30, 30 }, + { 114, 114 }, { 59, 59 }, { 102, 102 }, { 88, 88 }, + { 115, 115 }, { 74, 74 }, { 128, 128 }, { 116, 116 }, + { 45, 45 }, { 103, 103 }, { 89, 89 }, { 60, 60 }, + { 129, 129 }, { 117, 117 }, { 130, 130 }, { 131, 131 }, + { 104, 104 }, { 75, 75 }, { 46, 46 }, { 118, 118 }, + { 128, 128 }, { 90, 90 }, { 61, 61 }, { 132, 132 }, + { 105, 105 }, { 144, 144 }, { 119, 119 }, { 145, 145 }, + { 133, 133 }, { 76, 76 }, { 146, 146 }, { 120, 120 }, + { 91, 91 }, { 134, 134 }, { 147, 147 }, { 62, 62 }, + { 106, 106 }, { 135, 135 }, { 121, 121 }, { 92, 92 }, + { 148, 148 }, { 144, 144 }, { 77, 77 }, { 149, 149 }, + { 136, 136 }, { 107, 107 }, { 160, 160 }, { 161, 161 }, + { 150, 150 }, { 122, 122 }, { 78, 78 }, { 137, 137 }, + { 162, 162 }, { 151, 151 }, { 93, 93 }, { 163, 163 }, + { 108, 108 }, { 164, 164 }, { 152, 152 }, { 123, 123 }, + { 138, 138 }, { 160, 160 }, { 165, 165 }, { 94, 94 }, + { 176, 176 }, { 166, 166 }, { 109, 109 }, { 153, 153 }, + { 177, 177 }, { 124, 124 }, { 178, 178 }, { 139, 139 }, + { 167, 167 }, { 154, 154 }, { 110, 110 }, { 179, 179 }, + { 176, 176 }, { 180, 180 }, { 168, 168 }, { 140, 140 }, + { 125, 125 }, { 181, 181 }, { 192, 192 }, { 193, 193 }, + { 155, 155 }, { 182, 182 }, { 169, 169 }, { 194, 194 }, + { 126, 126 }, { 141, 141 }, { 195, 195 }, { 183, 183 }, + { 192, 192 }, { 196, 196 }, { 156, 156 }, { 170, 170 }, + { 142, 142 }, { 184, 184 }, { 197, 197 }, { 208, 208 }, + { 198, 198 }, { 209, 209 }, { 171, 171 }, { 157, 157 }, + { 185, 185 }, { 210, 210 }, { 208, 208 }, { 211, 211 }, + { 199, 199 }, { 224, 224 }, { 158, 158 }, { 212, 212 }, + { 224, 224 }, { 186, 186 }, { 200, 200 }, { 172, 172 }, + { 225, 225 }, { 213, 213 }, { 214, 214 }, { 226, 226 }, + { 201, 201 }, { 227, 227 }, { 187, 187 }, { 240, 240 }, + { 215, 215 }, { 173, 173 }, { 228, 228 }, { 241, 241 }, + { 202, 202 }, { 242, 242 }, { 216, 216 }, { 229, 229 }, + { 174, 174 }, { 188, 188 }, { 243, 243 }, { 230, 230 }, + { 203, 203 }, { 217, 217 }, { 231, 231 }, { 244, 244 }, + { 218, 218 }, { 245, 245 }, { 189, 189 }, { 232, 232 }, + { 204, 204 }, { 190, 190 }, { 246, 246 }, { 233, 233 }, + { 247, 247 }, { 219, 219 }, { 205, 205 }, { 248, 248 }, + { 234, 234 }, { 220, 220 }, { 206, 206 }, { 249, 249 }, + { 235, 235 }, { 221, 221 }, { 250, 250 }, { 222, 222 }, + { 236, 236 }, { 237, 237 }, { 251, 251 }, { 238, 238 }, + { 252, 252 }, { 253, 253 }, { 254, 254 }, { 0, 0 }, +}; + +const int16_t ff_vp9_row_scan_16x16_nb[256][2] = { + { 0, 0 }, { 16, 16 }, { 0, 0 }, { 32, 32 }, + { 1, 1 }, { 48, 48 }, { 17, 17 }, { 1, 1 }, + { 64, 64 }, { 2, 2 }, { 33, 33 }, { 80, 80 }, + { 18, 18 }, { 2, 2 }, { 49, 49 }, { 3, 3 }, + { 96, 96 }, { 34, 34 }, { 65, 65 }, { 19, 19 }, + { 3, 3 }, { 112, 112 }, { 50, 50 }, { 4, 4 }, + { 81, 81 }, { 35, 35 }, { 66, 66 }, { 4, 4 }, + { 128, 128 }, { 20, 20 }, { 51, 51 }, { 97, 97 }, + { 82, 82 }, { 5, 5 }, { 36, 36 }, { 144, 144 }, + { 67, 67 }, { 113, 113 }, { 21, 21 }, { 52, 52 }, + { 5, 5 }, { 98, 98 }, { 160, 160 }, { 83, 83 }, + { 37, 37 }, { 6, 6 }, { 68, 68 }, { 129, 129 }, + { 22, 22 }, { 53, 53 }, { 114, 114 }, { 6, 6 }, + { 99, 99 }, { 176, 176 }, { 84, 84 }, { 38, 38 }, + { 7, 7 }, { 69, 69 }, { 145, 145 }, { 130, 130 }, + { 115, 115 }, { 23, 23 }, { 54, 54 }, { 192, 192 }, + { 100, 100 }, { 7, 7 }, { 85, 85 }, { 161, 161 }, + { 39, 39 }, { 70, 70 }, { 8, 8 }, { 146, 146 }, + { 131, 131 }, { 116, 116 }, { 55, 55 }, { 208, 208 }, + { 101, 101 }, { 24, 24 }, { 86, 86 }, { 8, 8 }, + { 132, 132 }, { 40, 40 }, { 71, 71 }, { 177, 177 }, + { 147, 147 }, { 224, 224 }, { 117, 117 }, { 162, 162 }, + { 9, 9 }, { 102, 102 }, { 56, 56 }, { 25, 25 }, + { 87, 87 }, { 148, 148 }, { 9, 9 }, { 133, 133 }, + { 72, 72 }, { 118, 118 }, { 193, 193 }, { 163, 163 }, + { 41, 41 }, { 103, 103 }, { 178, 178 }, { 10, 10 }, + { 57, 57 }, { 149, 149 }, { 134, 134 }, { 88, 88 }, + { 26, 26 }, { 119, 119 }, { 10, 10 }, { 164, 164 }, + { 104, 104 }, { 73, 73 }, { 209, 209 }, { 179, 179 }, + { 42, 42 }, { 11, 11 }, { 194, 194 }, { 135, 135 }, + { 165, 165 }, { 150, 150 }, { 58, 58 }, { 27, 27 }, + { 89, 89 }, { 11, 11 }, { 120, 120 }, { 74, 74 }, + { 43, 43 }, { 225, 225 }, { 105, 105 }, { 12, 12 }, + { 180, 180 }, { 151, 151 }, { 195, 195 }, { 136, 136 }, + { 28, 28 }, { 166, 166 }, { 121, 121 }, { 59, 59 }, + { 12, 12 }, { 210, 210 }, { 90, 90 }, { 106, 106 }, + { 44, 44 }, { 181, 181 }, { 75, 75 }, { 152, 152 }, + { 13, 13 }, { 167, 167 }, { 137, 137 }, { 13, 13 }, + { 60, 60 }, { 196, 196 }, { 122, 122 }, { 29, 29 }, + { 91, 91 }, { 14, 14 }, { 182, 182 }, { 76, 76 }, + { 211, 211 }, { 153, 153 }, { 14, 14 }, { 107, 107 }, + { 138, 138 }, { 45, 45 }, { 226, 226 }, { 168, 168 }, + { 197, 197 }, { 123, 123 }, { 30, 30 }, { 61, 61 }, + { 15, 15 }, { 92, 92 }, { 154, 154 }, { 183, 183 }, + { 169, 169 }, { 108, 108 }, { 212, 212 }, { 77, 77 }, + { 139, 139 }, { 198, 198 }, { 46, 46 }, { 124, 124 }, + { 227, 227 }, { 62, 62 }, { 31, 31 }, { 184, 184 }, + { 93, 93 }, { 170, 170 }, { 155, 155 }, { 185, 185 }, + { 78, 78 }, { 47, 47 }, { 199, 199 }, { 213, 213 }, + { 140, 140 }, { 63, 63 }, { 109, 109 }, { 125, 125 }, + { 94, 94 }, { 200, 200 }, { 171, 171 }, { 156, 156 }, + { 228, 228 }, { 186, 186 }, { 214, 214 }, { 201, 201 }, + { 79, 79 }, { 141, 141 }, { 110, 110 }, { 229, 229 }, + { 95, 95 }, { 126, 126 }, { 215, 215 }, { 172, 172 }, + { 111, 111 }, { 142, 142 }, { 202, 202 }, { 157, 157 }, + { 216, 216 }, { 230, 230 }, { 217, 217 }, { 187, 187 }, + { 127, 127 }, { 231, 231 }, { 158, 158 }, { 173, 173 }, + { 143, 143 }, { 203, 203 }, { 188, 188 }, { 232, 232 }, + { 218, 218 }, { 233, 233 }, { 159, 159 }, { 174, 174 }, + { 204, 204 }, { 189, 189 }, { 234, 234 }, { 219, 219 }, + { 175, 175 }, { 205, 205 }, { 235, 235 }, { 220, 220 }, + { 190, 190 }, { 236, 236 }, { 206, 206 }, { 191, 191 }, + { 221, 221 }, { 207, 207 }, { 237, 237 }, { 222, 222 }, + { 238, 238 }, { 223, 223 }, { 239, 239 }, { 0, 0 }, +}; + +const int16_t ff_vp9_default_scan_32x32_nb[1024][2] = { + { 0, 0 }, { 0, 0 }, { 1, 1 }, { 32, 1 }, + { 32, 32 }, { 2, 2 }, { 33, 2 }, { 64, 33 }, + { 3, 3 }, { 64, 64 }, { 34, 3 }, { 65, 34 }, + { 4, 4 }, { 35, 4 }, { 96, 65 }, { 66, 35 }, + { 96, 96 }, { 97, 66 }, { 67, 36 }, { 36, 5 }, + { 5, 5 }, { 128, 97 }, { 98, 67 }, { 6, 6 }, + { 128, 128 }, { 68, 37 }, { 37, 6 }, { 129, 98 }, + { 99, 68 }, { 160, 129 }, { 130, 99 }, { 38, 7 }, + { 69, 38 }, { 7, 7 }, { 100, 69 }, { 161, 130 }, + { 131, 100 }, { 160, 160 }, { 70, 39 }, { 39, 8 }, + { 8, 8 }, { 101, 70 }, { 162, 131 }, { 132, 101 }, + { 192, 161 }, { 71, 40 }, { 192, 192 }, { 102, 71 }, + { 40, 9 }, { 163, 132 }, { 9, 9 }, { 193, 162 }, + { 133, 102 }, { 164, 133 }, { 72, 41 }, { 103, 72 }, + { 134, 103 }, { 224, 193 }, { 41, 10 }, { 194, 163 }, + { 10, 10 }, { 224, 224 }, { 165, 134 }, { 225, 194 }, + { 195, 164 }, { 73, 42 }, { 104, 73 }, { 135, 104 }, + { 42, 11 }, { 11, 11 }, { 166, 135 }, { 196, 165 }, + { 226, 195 }, { 256, 225 }, { 74, 43 }, { 105, 74 }, + { 136, 105 }, { 227, 196 }, { 43, 12 }, { 197, 166 }, + { 167, 136 }, { 257, 226 }, { 256, 256 }, { 12, 12 }, + { 228, 197 }, { 75, 44 }, { 106, 75 }, { 198, 167 }, + { 137, 106 }, { 258, 227 }, { 168, 137 }, { 288, 257 }, + { 44, 13 }, { 229, 198 }, { 259, 228 }, { 199, 168 }, + { 107, 76 }, { 13, 13 }, { 169, 138 }, { 138, 107 }, + { 288, 288 }, { 289, 258 }, { 76, 45 }, { 230, 199 }, + { 260, 229 }, { 45, 14 }, { 200, 169 }, { 139, 108 }, + { 290, 259 }, { 108, 77 }, { 231, 200 }, { 320, 289 }, + { 261, 230 }, { 170, 139 }, { 77, 46 }, { 291, 260 }, + { 14, 14 }, { 321, 290 }, { 201, 170 }, { 262, 231 }, + { 320, 320 }, { 171, 140 }, { 292, 261 }, { 232, 201 }, + { 140, 109 }, { 322, 291 }, { 109, 78 }, { 46, 15 }, + { 202, 171 }, { 263, 232 }, { 233, 202 }, { 293, 262 }, + { 352, 321 }, { 323, 292 }, { 15, 15 }, { 78, 47 }, + { 203, 172 }, { 264, 233 }, { 294, 263 }, { 324, 293 }, + { 172, 141 }, { 353, 322 }, { 141, 110 }, { 234, 203 }, + { 352, 352 }, { 47, 16 }, { 295, 264 }, { 110, 79 }, + { 265, 234 }, { 354, 323 }, { 325, 294 }, { 79, 48 }, + { 16, 16 }, { 204, 173 }, { 235, 204 }, { 173, 142 }, + { 355, 324 }, { 384, 353 }, { 326, 295 }, { 142, 111 }, + { 296, 265 }, { 266, 235 }, { 356, 325 }, { 385, 354 }, + { 111, 80 }, { 48, 17 }, { 327, 296 }, { 297, 266 }, + { 205, 174 }, { 384, 384 }, { 236, 205 }, { 357, 326 }, + { 386, 355 }, { 80, 49 }, { 174, 143 }, { 17, 17 }, + { 328, 297 }, { 358, 327 }, { 387, 356 }, { 298, 267 }, + { 329, 298 }, { 388, 357 }, { 112, 81 }, { 416, 385 }, + { 237, 206 }, { 359, 328 }, { 49, 18 }, { 206, 175 }, + { 417, 386 }, { 389, 358 }, { 330, 299 }, { 18, 18 }, + { 416, 416 }, { 360, 329 }, { 81, 50 }, { 418, 387 }, + { 390, 359 }, { 238, 207 }, { 50, 19 }, { 361, 330 }, + { 419, 388 }, { 113, 82 }, { 448, 417 }, { 448, 448 }, + { 420, 389 }, { 82, 51 }, { 362, 331 }, { 449, 418 }, + { 421, 390 }, { 480, 480 }, { 450, 419 }, { 422, 391 }, + { 114, 83 }, { 451, 420 }, { 480, 449 }, { 452, 421 }, + { 481, 450 }, { 453, 422 }, { 512, 512 }, { 482, 451 }, + { 454, 423 }, { 512, 481 }, { 483, 452 }, { 513, 482 }, + { 484, 453 }, { 514, 483 }, { 485, 454 }, { 544, 513 }, + { 544, 544 }, { 486, 455 }, { 545, 514 }, { 546, 515 }, + { 576, 576 }, { 576, 545 }, { 577, 546 }, { 578, 547 }, + { 608, 577 }, { 609, 578 }, { 610, 579 }, { 19, 19 }, + { 143, 112 }, { 267, 236 }, { 391, 360 }, { 515, 484 }, + { 608, 608 }, { 20, 20 }, { 51, 20 }, { 144, 113 }, + { 175, 144 }, { 268, 237 }, { 299, 268 }, { 392, 361 }, + { 423, 392 }, { 516, 485 }, { 547, 516 }, { 640, 609 }, + { 640, 640 }, { 21, 21 }, { 52, 21 }, { 83, 52 }, + { 145, 114 }, { 176, 145 }, { 207, 176 }, { 269, 238 }, + { 300, 269 }, { 331, 300 }, { 393, 362 }, { 424, 393 }, + { 455, 424 }, { 517, 486 }, { 548, 517 }, { 579, 548 }, + { 641, 610 }, { 672, 641 }, { 672, 672 }, { 22, 22 }, + { 53, 22 }, { 84, 53 }, { 115, 84 }, { 146, 115 }, + { 177, 146 }, { 208, 177 }, { 239, 208 }, { 270, 239 }, + { 301, 270 }, { 332, 301 }, { 363, 332 }, { 394, 363 }, + { 425, 394 }, { 456, 425 }, { 487, 456 }, { 518, 487 }, + { 549, 518 }, { 580, 549 }, { 611, 580 }, { 642, 611 }, + { 673, 642 }, { 704, 673 }, { 704, 704 }, { 54, 23 }, + { 85, 54 }, { 116, 85 }, { 178, 147 }, { 209, 178 }, + { 240, 209 }, { 302, 271 }, { 333, 302 }, { 364, 333 }, + { 426, 395 }, { 457, 426 }, { 488, 457 }, { 550, 519 }, + { 581, 550 }, { 612, 581 }, { 674, 643 }, { 705, 674 }, + { 736, 705 }, { 86, 55 }, { 117, 86 }, { 210, 179 }, + { 241, 210 }, { 334, 303 }, { 365, 334 }, { 458, 427 }, + { 489, 458 }, { 582, 551 }, { 613, 582 }, { 706, 675 }, + { 737, 706 }, { 118, 87 }, { 242, 211 }, { 366, 335 }, + { 490, 459 }, { 614, 583 }, { 738, 707 }, { 23, 23 }, + { 147, 116 }, { 271, 240 }, { 395, 364 }, { 519, 488 }, + { 643, 612 }, { 736, 736 }, { 24, 24 }, { 55, 24 }, + { 148, 117 }, { 179, 148 }, { 272, 241 }, { 303, 272 }, + { 396, 365 }, { 427, 396 }, { 520, 489 }, { 551, 520 }, + { 644, 613 }, { 675, 644 }, { 768, 737 }, { 768, 768 }, + { 25, 25 }, { 56, 25 }, { 87, 56 }, { 149, 118 }, + { 180, 149 }, { 211, 180 }, { 273, 242 }, { 304, 273 }, + { 335, 304 }, { 397, 366 }, { 428, 397 }, { 459, 428 }, + { 521, 490 }, { 552, 521 }, { 583, 552 }, { 645, 614 }, + { 676, 645 }, { 707, 676 }, { 769, 738 }, { 800, 769 }, + { 800, 800 }, { 26, 26 }, { 57, 26 }, { 88, 57 }, + { 119, 88 }, { 150, 119 }, { 181, 150 }, { 212, 181 }, + { 243, 212 }, { 274, 243 }, { 305, 274 }, { 336, 305 }, + { 367, 336 }, { 398, 367 }, { 429, 398 }, { 460, 429 }, + { 491, 460 }, { 522, 491 }, { 553, 522 }, { 584, 553 }, + { 615, 584 }, { 646, 615 }, { 677, 646 }, { 708, 677 }, + { 739, 708 }, { 770, 739 }, { 801, 770 }, { 832, 801 }, + { 832, 832 }, { 58, 27 }, { 89, 58 }, { 120, 89 }, + { 182, 151 }, { 213, 182 }, { 244, 213 }, { 306, 275 }, + { 337, 306 }, { 368, 337 }, { 430, 399 }, { 461, 430 }, + { 492, 461 }, { 554, 523 }, { 585, 554 }, { 616, 585 }, + { 678, 647 }, { 709, 678 }, { 740, 709 }, { 802, 771 }, + { 833, 802 }, { 864, 833 }, { 90, 59 }, { 121, 90 }, + { 214, 183 }, { 245, 214 }, { 338, 307 }, { 369, 338 }, + { 462, 431 }, { 493, 462 }, { 586, 555 }, { 617, 586 }, + { 710, 679 }, { 741, 710 }, { 834, 803 }, { 865, 834 }, + { 122, 91 }, { 246, 215 }, { 370, 339 }, { 494, 463 }, + { 618, 587 }, { 742, 711 }, { 866, 835 }, { 27, 27 }, + { 151, 120 }, { 275, 244 }, { 399, 368 }, { 523, 492 }, + { 647, 616 }, { 771, 740 }, { 864, 864 }, { 28, 28 }, + { 59, 28 }, { 152, 121 }, { 183, 152 }, { 276, 245 }, + { 307, 276 }, { 400, 369 }, { 431, 400 }, { 524, 493 }, + { 555, 524 }, { 648, 617 }, { 679, 648 }, { 772, 741 }, + { 803, 772 }, { 896, 865 }, { 896, 896 }, { 29, 29 }, + { 60, 29 }, { 91, 60 }, { 153, 122 }, { 184, 153 }, + { 215, 184 }, { 277, 246 }, { 308, 277 }, { 339, 308 }, + { 401, 370 }, { 432, 401 }, { 463, 432 }, { 525, 494 }, + { 556, 525 }, { 587, 556 }, { 649, 618 }, { 680, 649 }, + { 711, 680 }, { 773, 742 }, { 804, 773 }, { 835, 804 }, + { 897, 866 }, { 928, 897 }, { 928, 928 }, { 30, 30 }, + { 61, 30 }, { 92, 61 }, { 123, 92 }, { 154, 123 }, + { 185, 154 }, { 216, 185 }, { 247, 216 }, { 278, 247 }, + { 309, 278 }, { 340, 309 }, { 371, 340 }, { 402, 371 }, + { 433, 402 }, { 464, 433 }, { 495, 464 }, { 526, 495 }, + { 557, 526 }, { 588, 557 }, { 619, 588 }, { 650, 619 }, + { 681, 650 }, { 712, 681 }, { 743, 712 }, { 774, 743 }, + { 805, 774 }, { 836, 805 }, { 867, 836 }, { 898, 867 }, + { 929, 898 }, { 960, 929 }, { 960, 960 }, { 62, 31 }, + { 93, 62 }, { 124, 93 }, { 186, 155 }, { 217, 186 }, + { 248, 217 }, { 310, 279 }, { 341, 310 }, { 372, 341 }, + { 434, 403 }, { 465, 434 }, { 496, 465 }, { 558, 527 }, + { 589, 558 }, { 620, 589 }, { 682, 651 }, { 713, 682 }, + { 744, 713 }, { 806, 775 }, { 837, 806 }, { 868, 837 }, + { 930, 899 }, { 961, 930 }, { 992, 961 }, { 94, 63 }, + { 125, 94 }, { 218, 187 }, { 249, 218 }, { 342, 311 }, + { 373, 342 }, { 466, 435 }, { 497, 466 }, { 590, 559 }, + { 621, 590 }, { 714, 683 }, { 745, 714 }, { 838, 807 }, + { 869, 838 }, { 962, 931 }, { 993, 962 }, { 126, 95 }, + { 250, 219 }, { 374, 343 }, { 498, 467 }, { 622, 591 }, + { 746, 715 }, { 870, 839 }, { 994, 963 }, { 155, 124 }, + { 279, 248 }, { 403, 372 }, { 527, 496 }, { 651, 620 }, + { 775, 744 }, { 899, 868 }, { 156, 125 }, { 187, 156 }, + { 280, 249 }, { 311, 280 }, { 404, 373 }, { 435, 404 }, + { 528, 497 }, { 559, 528 }, { 652, 621 }, { 683, 652 }, + { 776, 745 }, { 807, 776 }, { 900, 869 }, { 931, 900 }, + { 157, 126 }, { 188, 157 }, { 219, 188 }, { 281, 250 }, + { 312, 281 }, { 343, 312 }, { 405, 374 }, { 436, 405 }, + { 467, 436 }, { 529, 498 }, { 560, 529 }, { 591, 560 }, + { 653, 622 }, { 684, 653 }, { 715, 684 }, { 777, 746 }, + { 808, 777 }, { 839, 808 }, { 901, 870 }, { 932, 901 }, + { 963, 932 }, { 158, 127 }, { 189, 158 }, { 220, 189 }, + { 251, 220 }, { 282, 251 }, { 313, 282 }, { 344, 313 }, + { 375, 344 }, { 406, 375 }, { 437, 406 }, { 468, 437 }, + { 499, 468 }, { 530, 499 }, { 561, 530 }, { 592, 561 }, + { 623, 592 }, { 654, 623 }, { 685, 654 }, { 716, 685 }, + { 747, 716 }, { 778, 747 }, { 809, 778 }, { 840, 809 }, + { 871, 840 }, { 902, 871 }, { 933, 902 }, { 964, 933 }, + { 995, 964 }, { 190, 159 }, { 221, 190 }, { 252, 221 }, + { 314, 283 }, { 345, 314 }, { 376, 345 }, { 438, 407 }, + { 469, 438 }, { 500, 469 }, { 562, 531 }, { 593, 562 }, + { 624, 593 }, { 686, 655 }, { 717, 686 }, { 748, 717 }, + { 810, 779 }, { 841, 810 }, { 872, 841 }, { 934, 903 }, + { 965, 934 }, { 996, 965 }, { 222, 191 }, { 253, 222 }, + { 346, 315 }, { 377, 346 }, { 470, 439 }, { 501, 470 }, + { 594, 563 }, { 625, 594 }, { 718, 687 }, { 749, 718 }, + { 842, 811 }, { 873, 842 }, { 966, 935 }, { 997, 966 }, + { 254, 223 }, { 378, 347 }, { 502, 471 }, { 626, 595 }, + { 750, 719 }, { 874, 843 }, { 998, 967 }, { 283, 252 }, + { 407, 376 }, { 531, 500 }, { 655, 624 }, { 779, 748 }, + { 903, 872 }, { 284, 253 }, { 315, 284 }, { 408, 377 }, + { 439, 408 }, { 532, 501 }, { 563, 532 }, { 656, 625 }, + { 687, 656 }, { 780, 749 }, { 811, 780 }, { 904, 873 }, + { 935, 904 }, { 285, 254 }, { 316, 285 }, { 347, 316 }, + { 409, 378 }, { 440, 409 }, { 471, 440 }, { 533, 502 }, + { 564, 533 }, { 595, 564 }, { 657, 626 }, { 688, 657 }, + { 719, 688 }, { 781, 750 }, { 812, 781 }, { 843, 812 }, + { 905, 874 }, { 936, 905 }, { 967, 936 }, { 286, 255 }, + { 317, 286 }, { 348, 317 }, { 379, 348 }, { 410, 379 }, + { 441, 410 }, { 472, 441 }, { 503, 472 }, { 534, 503 }, + { 565, 534 }, { 596, 565 }, { 627, 596 }, { 658, 627 }, + { 689, 658 }, { 720, 689 }, { 751, 720 }, { 782, 751 }, + { 813, 782 }, { 844, 813 }, { 875, 844 }, { 906, 875 }, + { 937, 906 }, { 968, 937 }, { 999, 968 }, { 318, 287 }, + { 349, 318 }, { 380, 349 }, { 442, 411 }, { 473, 442 }, + { 504, 473 }, { 566, 535 }, { 597, 566 }, { 628, 597 }, + { 690, 659 }, { 721, 690 }, { 752, 721 }, { 814, 783 }, + { 845, 814 }, { 876, 845 }, { 938, 907 }, { 969, 938 }, + { 1000, 969 }, { 350, 319 }, { 381, 350 }, { 474, 443 }, + { 505, 474 }, { 598, 567 }, { 629, 598 }, { 722, 691 }, + { 753, 722 }, { 846, 815 }, { 877, 846 }, { 970, 939 }, + { 1001, 970 }, { 382, 351 }, { 506, 475 }, { 630, 599 }, + { 754, 723 }, { 878, 847 }, { 1002, 971 }, { 411, 380 }, + { 535, 504 }, { 659, 628 }, { 783, 752 }, { 907, 876 }, + { 412, 381 }, { 443, 412 }, { 536, 505 }, { 567, 536 }, + { 660, 629 }, { 691, 660 }, { 784, 753 }, { 815, 784 }, + { 908, 877 }, { 939, 908 }, { 413, 382 }, { 444, 413 }, + { 475, 444 }, { 537, 506 }, { 568, 537 }, { 599, 568 }, + { 661, 630 }, { 692, 661 }, { 723, 692 }, { 785, 754 }, + { 816, 785 }, { 847, 816 }, { 909, 878 }, { 940, 909 }, + { 971, 940 }, { 414, 383 }, { 445, 414 }, { 476, 445 }, + { 507, 476 }, { 538, 507 }, { 569, 538 }, { 600, 569 }, + { 631, 600 }, { 662, 631 }, { 693, 662 }, { 724, 693 }, + { 755, 724 }, { 786, 755 }, { 817, 786 }, { 848, 817 }, + { 879, 848 }, { 910, 879 }, { 941, 910 }, { 972, 941 }, + { 1003, 972 }, { 446, 415 }, { 477, 446 }, { 508, 477 }, + { 570, 539 }, { 601, 570 }, { 632, 601 }, { 694, 663 }, + { 725, 694 }, { 756, 725 }, { 818, 787 }, { 849, 818 }, + { 880, 849 }, { 942, 911 }, { 973, 942 }, { 1004, 973 }, + { 478, 447 }, { 509, 478 }, { 602, 571 }, { 633, 602 }, + { 726, 695 }, { 757, 726 }, { 850, 819 }, { 881, 850 }, + { 974, 943 }, { 1005, 974 }, { 510, 479 }, { 634, 603 }, + { 758, 727 }, { 882, 851 }, { 1006, 975 }, { 539, 508 }, + { 663, 632 }, { 787, 756 }, { 911, 880 }, { 540, 509 }, + { 571, 540 }, { 664, 633 }, { 695, 664 }, { 788, 757 }, + { 819, 788 }, { 912, 881 }, { 943, 912 }, { 541, 510 }, + { 572, 541 }, { 603, 572 }, { 665, 634 }, { 696, 665 }, + { 727, 696 }, { 789, 758 }, { 820, 789 }, { 851, 820 }, + { 913, 882 }, { 944, 913 }, { 975, 944 }, { 542, 511 }, + { 573, 542 }, { 604, 573 }, { 635, 604 }, { 666, 635 }, + { 697, 666 }, { 728, 697 }, { 759, 728 }, { 790, 759 }, + { 821, 790 }, { 852, 821 }, { 883, 852 }, { 914, 883 }, + { 945, 914 }, { 976, 945 }, { 1007, 976 }, { 574, 543 }, + { 605, 574 }, { 636, 605 }, { 698, 667 }, { 729, 698 }, + { 760, 729 }, { 822, 791 }, { 853, 822 }, { 884, 853 }, + { 946, 915 }, { 977, 946 }, { 1008, 977 }, { 606, 575 }, + { 637, 606 }, { 730, 699 }, { 761, 730 }, { 854, 823 }, + { 885, 854 }, { 978, 947 }, { 1009, 978 }, { 638, 607 }, + { 762, 731 }, { 886, 855 }, { 1010, 979 }, { 667, 636 }, + { 791, 760 }, { 915, 884 }, { 668, 637 }, { 699, 668 }, + { 792, 761 }, { 823, 792 }, { 916, 885 }, { 947, 916 }, + { 669, 638 }, { 700, 669 }, { 731, 700 }, { 793, 762 }, + { 824, 793 }, { 855, 824 }, { 917, 886 }, { 948, 917 }, + { 979, 948 }, { 670, 639 }, { 701, 670 }, { 732, 701 }, + { 763, 732 }, { 794, 763 }, { 825, 794 }, { 856, 825 }, + { 887, 856 }, { 918, 887 }, { 949, 918 }, { 980, 949 }, + { 1011, 980 }, { 702, 671 }, { 733, 702 }, { 764, 733 }, + { 826, 795 }, { 857, 826 }, { 888, 857 }, { 950, 919 }, + { 981, 950 }, { 1012, 981 }, { 734, 703 }, { 765, 734 }, + { 858, 827 }, { 889, 858 }, { 982, 951 }, { 1013, 982 }, + { 766, 735 }, { 890, 859 }, { 1014, 983 }, { 795, 764 }, + { 919, 888 }, { 796, 765 }, { 827, 796 }, { 920, 889 }, + { 951, 920 }, { 797, 766 }, { 828, 797 }, { 859, 828 }, + { 921, 890 }, { 952, 921 }, { 983, 952 }, { 798, 767 }, + { 829, 798 }, { 860, 829 }, { 891, 860 }, { 922, 891 }, + { 953, 922 }, { 984, 953 }, { 1015, 984 }, { 830, 799 }, + { 861, 830 }, { 892, 861 }, { 954, 923 }, { 985, 954 }, + { 1016, 985 }, { 862, 831 }, { 893, 862 }, { 986, 955 }, + { 1017, 986 }, { 894, 863 }, { 1018, 987 }, { 923, 892 }, + { 924, 893 }, { 955, 924 }, { 925, 894 }, { 956, 925 }, + { 987, 956 }, { 926, 895 }, { 957, 926 }, { 988, 957 }, + { 1019, 988 }, { 958, 927 }, { 989, 958 }, { 1020, 989 }, + { 990, 959 }, { 1021, 990 }, { 1022, 991 }, { 0, 0 }, +}; + +const int16_t (*ff_vp9_scans_nb[5][4])[2] = { + { + ff_vp9_default_scan_4x4_nb, ff_vp9_col_scan_4x4_nb, + ff_vp9_row_scan_4x4_nb, ff_vp9_default_scan_4x4_nb + }, { + ff_vp9_default_scan_8x8_nb, ff_vp9_col_scan_8x8_nb, + ff_vp9_row_scan_8x8_nb, ff_vp9_default_scan_8x8_nb + }, { + ff_vp9_default_scan_16x16_nb, ff_vp9_col_scan_16x16_nb, + ff_vp9_row_scan_16x16_nb, ff_vp9_default_scan_16x16_nb + }, { + ff_vp9_default_scan_32x32_nb, ff_vp9_default_scan_32x32_nb, + ff_vp9_default_scan_32x32_nb, ff_vp9_default_scan_32x32_nb + }, { // lossless + ff_vp9_default_scan_4x4_nb, ff_vp9_default_scan_4x4_nb, + ff_vp9_default_scan_4x4_nb, ff_vp9_default_scan_4x4_nb + } +}; + +const uint8_t ff_vp9_model_pareto8[256][8] = { + { 6, 86, 128, 11, 87, 42, 91, 52 }, + { 3, 86, 128, 6, 86, 23, 88, 29 }, + { 6, 86, 128, 11, 87, 42, 91, 52 }, + { 9, 86, 129, 17, 88, 61, 94, 76 }, + { 12, 86, 129, 22, 88, 77, 97, 93 }, + { 15, 87, 129, 28, 89, 93, 100, 110 }, + { 17, 87, 129, 33, 90, 105, 103, 123 }, + { 20, 88, 130, 38, 91, 118, 106, 136 }, + { 23, 88, 130, 43, 91, 128, 108, 146 }, + { 26, 89, 131, 48, 92, 139, 111, 156 }, + { 28, 89, 131, 53, 93, 147, 114, 163 }, + { 31, 90, 131, 58, 94, 156, 117, 171 }, + { 34, 90, 131, 62, 94, 163, 119, 177 }, + { 37, 90, 132, 66, 95, 171, 122, 184 }, + { 39, 90, 132, 70, 96, 177, 124, 189 }, + { 42, 91, 132, 75, 97, 183, 127, 194 }, + { 44, 91, 132, 79, 97, 188, 129, 198 }, + { 47, 92, 133, 83, 98, 193, 132, 202 }, + { 49, 92, 133, 86, 99, 197, 134, 205 }, + { 52, 93, 133, 90, 100, 201, 137, 208 }, + { 54, 93, 133, 94, 100, 204, 139, 211 }, + { 57, 94, 134, 98, 101, 208, 142, 214 }, + { 59, 94, 134, 101, 102, 211, 144, 216 }, + { 62, 94, 135, 105, 103, 214, 146, 218 }, + { 64, 94, 135, 108, 103, 216, 148, 220 }, + { 66, 95, 135, 111, 104, 219, 151, 222 }, + { 68, 95, 135, 114, 105, 221, 153, 223 }, + { 71, 96, 136, 117, 106, 224, 155, 225 }, + { 73, 96, 136, 120, 106, 225, 157, 226 }, + { 76, 97, 136, 123, 107, 227, 159, 228 }, + { 78, 97, 136, 126, 108, 229, 160, 229 }, + { 80, 98, 137, 129, 109, 231, 162, 231 }, + { 82, 98, 137, 131, 109, 232, 164, 232 }, + { 84, 98, 138, 134, 110, 234, 166, 233 }, + { 86, 98, 138, 137, 111, 235, 168, 234 }, + { 89, 99, 138, 140, 112, 236, 170, 235 }, + { 91, 99, 138, 142, 112, 237, 171, 235 }, + { 93, 100, 139, 145, 113, 238, 173, 236 }, + { 95, 100, 139, 147, 114, 239, 174, 237 }, + { 97, 101, 140, 149, 115, 240, 176, 238 }, + { 99, 101, 140, 151, 115, 241, 177, 238 }, + { 101, 102, 140, 154, 116, 242, 179, 239 }, + { 103, 102, 140, 156, 117, 242, 180, 239 }, + { 105, 103, 141, 158, 118, 243, 182, 240 }, + { 107, 103, 141, 160, 118, 243, 183, 240 }, + { 109, 104, 141, 162, 119, 244, 185, 241 }, + { 111, 104, 141, 164, 119, 244, 186, 241 }, + { 113, 104, 142, 166, 120, 245, 187, 242 }, + { 114, 104, 142, 168, 121, 245, 188, 242 }, + { 116, 105, 143, 170, 122, 246, 190, 243 }, + { 118, 105, 143, 171, 122, 246, 191, 243 }, + { 120, 106, 143, 173, 123, 247, 192, 244 }, + { 121, 106, 143, 175, 124, 247, 193, 244 }, + { 123, 107, 144, 177, 125, 248, 195, 244 }, + { 125, 107, 144, 178, 125, 248, 196, 244 }, + { 127, 108, 145, 180, 126, 249, 197, 245 }, + { 128, 108, 145, 181, 127, 249, 198, 245 }, + { 130, 109, 145, 183, 128, 249, 199, 245 }, + { 132, 109, 145, 184, 128, 249, 200, 245 }, + { 134, 110, 146, 186, 129, 250, 201, 246 }, + { 135, 110, 146, 187, 130, 250, 202, 246 }, + { 137, 111, 147, 189, 131, 251, 203, 246 }, + { 138, 111, 147, 190, 131, 251, 204, 246 }, + { 140, 112, 147, 192, 132, 251, 205, 247 }, + { 141, 112, 147, 193, 132, 251, 206, 247 }, + { 143, 113, 148, 194, 133, 251, 207, 247 }, + { 144, 113, 148, 195, 134, 251, 207, 247 }, + { 146, 114, 149, 197, 135, 252, 208, 248 }, + { 147, 114, 149, 198, 135, 252, 209, 248 }, + { 149, 115, 149, 199, 136, 252, 210, 248 }, + { 150, 115, 149, 200, 137, 252, 210, 248 }, + { 152, 115, 150, 201, 138, 252, 211, 248 }, + { 153, 115, 150, 202, 138, 252, 212, 248 }, + { 155, 116, 151, 204, 139, 253, 213, 249 }, + { 156, 116, 151, 205, 139, 253, 213, 249 }, + { 158, 117, 151, 206, 140, 253, 214, 249 }, + { 159, 117, 151, 207, 141, 253, 215, 249 }, + { 161, 118, 152, 208, 142, 253, 216, 249 }, + { 162, 118, 152, 209, 142, 253, 216, 249 }, + { 163, 119, 153, 210, 143, 253, 217, 249 }, + { 164, 119, 153, 211, 143, 253, 217, 249 }, + { 166, 120, 153, 212, 144, 254, 218, 250 }, + { 167, 120, 153, 212, 145, 254, 219, 250 }, + { 168, 121, 154, 213, 146, 254, 220, 250 }, + { 169, 121, 154, 214, 146, 254, 220, 250 }, + { 171, 122, 155, 215, 147, 254, 221, 250 }, + { 172, 122, 155, 216, 147, 254, 221, 250 }, + { 173, 123, 155, 217, 148, 254, 222, 250 }, + { 174, 123, 155, 217, 149, 254, 222, 250 }, + { 176, 124, 156, 218, 150, 254, 223, 250 }, + { 177, 124, 156, 219, 150, 254, 223, 250 }, + { 178, 125, 157, 220, 151, 254, 224, 251 }, + { 179, 125, 157, 220, 151, 254, 224, 251 }, + { 180, 126, 157, 221, 152, 254, 225, 251 }, + { 181, 126, 157, 221, 152, 254, 225, 251 }, + { 183, 127, 158, 222, 153, 254, 226, 251 }, + { 184, 127, 158, 223, 154, 254, 226, 251 }, + { 185, 128, 159, 224, 155, 255, 227, 251 }, + { 186, 128, 159, 224, 155, 255, 227, 251 }, + { 187, 129, 160, 225, 156, 255, 228, 251 }, + { 188, 130, 160, 225, 156, 255, 228, 251 }, + { 189, 131, 160, 226, 157, 255, 228, 251 }, + { 190, 131, 160, 226, 158, 255, 228, 251 }, + { 191, 132, 161, 227, 159, 255, 229, 251 }, + { 192, 132, 161, 227, 159, 255, 229, 251 }, + { 193, 133, 162, 228, 160, 255, 230, 252 }, + { 194, 133, 162, 229, 160, 255, 230, 252 }, + { 195, 134, 163, 230, 161, 255, 231, 252 }, + { 196, 134, 163, 230, 161, 255, 231, 252 }, + { 197, 135, 163, 231, 162, 255, 231, 252 }, + { 198, 135, 163, 231, 162, 255, 231, 252 }, + { 199, 136, 164, 232, 163, 255, 232, 252 }, + { 200, 136, 164, 232, 164, 255, 232, 252 }, + { 201, 137, 165, 233, 165, 255, 233, 252 }, + { 201, 137, 165, 233, 165, 255, 233, 252 }, + { 202, 138, 166, 233, 166, 255, 233, 252 }, + { 203, 138, 166, 233, 166, 255, 233, 252 }, + { 204, 139, 166, 234, 167, 255, 234, 252 }, + { 205, 139, 166, 234, 167, 255, 234, 252 }, + { 206, 140, 167, 235, 168, 255, 235, 252 }, + { 206, 140, 167, 235, 168, 255, 235, 252 }, + { 207, 141, 168, 236, 169, 255, 235, 252 }, + { 208, 141, 168, 236, 170, 255, 235, 252 }, + { 209, 142, 169, 237, 171, 255, 236, 252 }, + { 209, 143, 169, 237, 171, 255, 236, 252 }, + { 210, 144, 169, 237, 172, 255, 236, 252 }, + { 211, 144, 169, 237, 172, 255, 236, 252 }, + { 212, 145, 170, 238, 173, 255, 237, 252 }, + { 213, 145, 170, 238, 173, 255, 237, 252 }, + { 214, 146, 171, 239, 174, 255, 237, 253 }, + { 214, 146, 171, 239, 174, 255, 237, 253 }, + { 215, 147, 172, 240, 175, 255, 238, 253 }, + { 215, 147, 172, 240, 175, 255, 238, 253 }, + { 216, 148, 173, 240, 176, 255, 238, 253 }, + { 217, 148, 173, 240, 176, 255, 238, 253 }, + { 218, 149, 173, 241, 177, 255, 239, 253 }, + { 218, 149, 173, 241, 178, 255, 239, 253 }, + { 219, 150, 174, 241, 179, 255, 239, 253 }, + { 219, 151, 174, 241, 179, 255, 239, 253 }, + { 220, 152, 175, 242, 180, 255, 240, 253 }, + { 221, 152, 175, 242, 180, 255, 240, 253 }, + { 222, 153, 176, 242, 181, 255, 240, 253 }, + { 222, 153, 176, 242, 181, 255, 240, 253 }, + { 223, 154, 177, 243, 182, 255, 240, 253 }, + { 223, 154, 177, 243, 182, 255, 240, 253 }, + { 224, 155, 178, 244, 183, 255, 241, 253 }, + { 224, 155, 178, 244, 183, 255, 241, 253 }, + { 225, 156, 178, 244, 184, 255, 241, 253 }, + { 225, 157, 178, 244, 184, 255, 241, 253 }, + { 226, 158, 179, 244, 185, 255, 242, 253 }, + { 227, 158, 179, 244, 185, 255, 242, 253 }, + { 228, 159, 180, 245, 186, 255, 242, 253 }, + { 228, 159, 180, 245, 186, 255, 242, 253 }, + { 229, 160, 181, 245, 187, 255, 242, 253 }, + { 229, 160, 181, 245, 187, 255, 242, 253 }, + { 230, 161, 182, 246, 188, 255, 243, 253 }, + { 230, 162, 182, 246, 188, 255, 243, 253 }, + { 231, 163, 183, 246, 189, 255, 243, 253 }, + { 231, 163, 183, 246, 189, 255, 243, 253 }, + { 232, 164, 184, 247, 190, 255, 243, 253 }, + { 232, 164, 184, 247, 190, 255, 243, 253 }, + { 233, 165, 185, 247, 191, 255, 244, 253 }, + { 233, 165, 185, 247, 191, 255, 244, 253 }, + { 234, 166, 185, 247, 192, 255, 244, 253 }, + { 234, 167, 185, 247, 192, 255, 244, 253 }, + { 235, 168, 186, 248, 193, 255, 244, 253 }, + { 235, 168, 186, 248, 193, 255, 244, 253 }, + { 236, 169, 187, 248, 194, 255, 244, 253 }, + { 236, 169, 187, 248, 194, 255, 244, 253 }, + { 236, 170, 188, 248, 195, 255, 245, 253 }, + { 236, 170, 188, 248, 195, 255, 245, 253 }, + { 237, 171, 189, 249, 196, 255, 245, 254 }, + { 237, 172, 189, 249, 196, 255, 245, 254 }, + { 238, 173, 190, 249, 197, 255, 245, 254 }, + { 238, 173, 190, 249, 197, 255, 245, 254 }, + { 239, 174, 191, 249, 198, 255, 245, 254 }, + { 239, 174, 191, 249, 198, 255, 245, 254 }, + { 240, 175, 192, 249, 199, 255, 246, 254 }, + { 240, 176, 192, 249, 199, 255, 246, 254 }, + { 240, 177, 193, 250, 200, 255, 246, 254 }, + { 240, 177, 193, 250, 200, 255, 246, 254 }, + { 241, 178, 194, 250, 201, 255, 246, 254 }, + { 241, 178, 194, 250, 201, 255, 246, 254 }, + { 242, 179, 195, 250, 202, 255, 246, 254 }, + { 242, 180, 195, 250, 202, 255, 246, 254 }, + { 242, 181, 196, 250, 203, 255, 247, 254 }, + { 242, 181, 196, 250, 203, 255, 247, 254 }, + { 243, 182, 197, 251, 204, 255, 247, 254 }, + { 243, 183, 197, 251, 204, 255, 247, 254 }, + { 244, 184, 198, 251, 205, 255, 247, 254 }, + { 244, 184, 198, 251, 205, 255, 247, 254 }, + { 244, 185, 199, 251, 206, 255, 247, 254 }, + { 244, 185, 199, 251, 206, 255, 247, 254 }, + { 245, 186, 200, 251, 207, 255, 247, 254 }, + { 245, 187, 200, 251, 207, 255, 247, 254 }, + { 246, 188, 201, 252, 207, 255, 248, 254 }, + { 246, 188, 201, 252, 207, 255, 248, 254 }, + { 246, 189, 202, 252, 208, 255, 248, 254 }, + { 246, 190, 202, 252, 208, 255, 248, 254 }, + { 247, 191, 203, 252, 209, 255, 248, 254 }, + { 247, 191, 203, 252, 209, 255, 248, 254 }, + { 247, 192, 204, 252, 210, 255, 248, 254 }, + { 247, 193, 204, 252, 210, 255, 248, 254 }, + { 248, 194, 205, 252, 211, 255, 248, 254 }, + { 248, 194, 205, 252, 211, 255, 248, 254 }, + { 248, 195, 206, 252, 212, 255, 249, 254 }, + { 248, 196, 206, 252, 212, 255, 249, 254 }, + { 249, 197, 207, 253, 213, 255, 249, 254 }, + { 249, 197, 207, 253, 213, 255, 249, 254 }, + { 249, 198, 208, 253, 214, 255, 249, 254 }, + { 249, 199, 209, 253, 214, 255, 249, 254 }, + { 250, 200, 210, 253, 215, 255, 249, 254 }, + { 250, 200, 210, 253, 215, 255, 249, 254 }, + { 250, 201, 211, 253, 215, 255, 249, 254 }, + { 250, 202, 211, 253, 215, 255, 249, 254 }, + { 250, 203, 212, 253, 216, 255, 249, 254 }, + { 250, 203, 212, 253, 216, 255, 249, 254 }, + { 251, 204, 213, 253, 217, 255, 250, 254 }, + { 251, 205, 213, 253, 217, 255, 250, 254 }, + { 251, 206, 214, 254, 218, 255, 250, 254 }, + { 251, 206, 215, 254, 218, 255, 250, 254 }, + { 252, 207, 216, 254, 219, 255, 250, 254 }, + { 252, 208, 216, 254, 219, 255, 250, 254 }, + { 252, 209, 217, 254, 220, 255, 250, 254 }, + { 252, 210, 217, 254, 220, 255, 250, 254 }, + { 252, 211, 218, 254, 221, 255, 250, 254 }, + { 252, 212, 218, 254, 221, 255, 250, 254 }, + { 253, 213, 219, 254, 222, 255, 250, 254 }, + { 253, 213, 220, 254, 222, 255, 250, 254 }, + { 253, 214, 221, 254, 223, 255, 250, 254 }, + { 253, 215, 221, 254, 223, 255, 250, 254 }, + { 253, 216, 222, 254, 224, 255, 251, 254 }, + { 253, 217, 223, 254, 224, 255, 251, 254 }, + { 253, 218, 224, 254, 225, 255, 251, 254 }, + { 253, 219, 224, 254, 225, 255, 251, 254 }, + { 254, 220, 225, 254, 225, 255, 251, 254 }, + { 254, 221, 226, 254, 225, 255, 251, 254 }, + { 254, 222, 227, 255, 226, 255, 251, 254 }, + { 254, 223, 227, 255, 226, 255, 251, 254 }, + { 254, 224, 228, 255, 227, 255, 251, 254 }, + { 254, 225, 229, 255, 227, 255, 251, 254 }, + { 254, 226, 230, 255, 228, 255, 251, 254 }, + { 254, 227, 230, 255, 229, 255, 251, 254 }, + { 255, 228, 231, 255, 230, 255, 251, 254 }, + { 255, 229, 232, 255, 230, 255, 251, 254 }, + { 255, 230, 233, 255, 231, 255, 252, 254 }, + { 255, 231, 234, 255, 231, 255, 252, 254 }, + { 255, 232, 235, 255, 232, 255, 252, 254 }, + { 255, 233, 236, 255, 232, 255, 252, 254 }, + { 255, 235, 237, 255, 233, 255, 252, 254 }, + { 255, 236, 238, 255, 234, 255, 252, 254 }, + { 255, 238, 240, 255, 235, 255, 252, 255 }, + { 255, 239, 241, 255, 235, 255, 252, 254 }, + { 255, 241, 243, 255, 236, 255, 252, 254 }, + { 255, 243, 245, 255, 237, 255, 252, 254 }, + { 255, 246, 247, 255, 239, 255, 253, 255 }, +}; + +const ProbContext ff_vp9_default_probs = { + { /* y_mode */ + { 65, 32, 18, 144, 162, 194, 41, 51, 98 } /* bsize < 8x8 */, + { 132, 68, 18, 165, 217, 196, 45, 40, 78 } /* bsize < 16x16 */, + { 173, 80, 19, 176, 240, 193, 64, 35, 46 } /* bsize < 32x32 */, + { 221, 135, 38, 194, 248, 121, 96, 85, 29 } /* bsize >= 32x32 */ + }, { /* uv_mode */ + { 48, 12, 154, 155, 139, 90, 34, 117, 119 } /* y = v */, + { 67, 6, 25, 204, 243, 158, 13, 21, 96 } /* y = h */, + { 120, 7, 76, 176, 208, 126, 28, 54, 103 } /* y = dc */, + { 97, 5, 44, 131, 176, 139, 48, 68, 97 } /* y = d45 */, + { 83, 5, 42, 156, 111, 152, 26, 49, 152 } /* y = d135 */, + { 80, 5, 58, 178, 74, 83, 33, 62, 145 } /* y = d117 */, + { 86, 5, 32, 154, 192, 168, 14, 22, 163 } /* y = d153 */, + { 77, 7, 64, 116, 132, 122, 37, 126, 120 } /* y = d63 */, + { 85, 5, 32, 156, 216, 148, 19, 29, 73 } /* y = d27 */, + { 101, 21, 107, 181, 192, 103, 19, 67, 125 } /* y = tm */ + }, { /* filter */ + { 235, 162, }, + { 36, 255, }, + { 34, 3, }, + { 149, 144, }, + }, { /* mv_mode */ + { 2, 173, 34 }, // 0 = both zero mv + { 7, 145, 85 }, // 1 = one zero mv + one a predicted mv + { 7, 166, 63 }, // 2 = two predicted mvs + { 7, 94, 66 }, // 3 = one predicted/zero and one new mv + { 8, 64, 46 }, // 4 = two new mvs + { 17, 81, 31 }, // 5 = one intra neighbor + x + { 25, 29, 30 }, // 6 = two intra neighbors + }, { /* intra */ + 9, 102, 187, 225 + }, { /* comp */ + 239, 183, 119, 96, 41 + }, { /* single_ref */ + { 33, 16 }, + { 77, 74 }, + { 142, 142 }, + { 172, 170 }, + { 238, 247 } + }, { /* comp_ref */ + 50, 126, 123, 221, 226 + }, { /* tx32p */ + { 3, 136, 37, }, + { 5, 52, 13, }, + }, { /* tx16p */ + { 20, 152, }, + { 15, 101, }, + }, { /* tx8p */ + 100, 66 + }, { /* skip */ + 192, 128, 64 + }, { /* mv_joint */ + 32, 64, 96 + }, { + { /* mv vertical component */ + 128, /* sign */ + { 224, 144, 192, 168, 192, 176, 192, 198, 198, 245 }, /* class */ + 216, /* class0 */ + { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240 }, /* bits */ + { /* class0_fp */ + { 128, 128, 64 }, + { 96, 112, 64 } + }, + { 64, 96, 64 }, /* fp */ + 160, /* class0_hp bit */ + 128, /* hp */ + }, { /* mv horizontal component */ + 128, /* sign */ + { 216, 128, 176, 160, 176, 176, 192, 198, 198, 208 }, /* class */ + 208, /* class0 */ + { 136, 140, 148, 160, 176, 192, 224, 234, 234, 240 }, /* bits */ + { /* class0_fp */ + { 128, 128, 64 }, + { 96, 112, 64 } + }, + { 64, 96, 64 }, /* fp */ + 160, /* class0_hp bit */ + 128, /* hp */ + } + }, { /* partition */ + { /* 64x64 -> 32x32 */ + { 222, 34, 30 } /* a/l both not split */, + { 72, 16, 44 } /* a split, l not split */, + { 58, 32, 12 } /* l split, a not split */, + { 10, 7, 6 } /* a/l both split */, + }, { /* 32x32 -> 16x16 */ + { 177, 58, 59 } /* a/l both not split */, + { 68, 26, 63 } /* a split, l not split */, + { 52, 79, 25 } /* l split, a not split */, + { 17, 14, 12 } /* a/l both split */, + }, { /* 16x16 -> 8x8 */ + { 174, 73, 87 } /* a/l both not split */, + { 92, 41, 83 } /* a split, l not split */, + { 82, 99, 50 } /* l split, a not split */, + { 53, 39, 39 } /* a/l both split */, + }, { /* 8x8 -> 4x4 */ + { 199, 122, 141 } /* a/l both not split */, + { 147, 63, 159 } /* a split, l not split */, + { 148, 133, 118 } /* l split, a not split */, + { 121, 104, 114 } /* a/l both split */, + } + }, +}; + +const uint8_t ff_vp9_default_coef_probs[4][2][2][6][6][3] = { + { /* tx = 4x4 */ + { /* block Type 0 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 195, 29, 183 }, + { 84, 49, 136 }, + { 8, 42, 71 } + }, { /* Coeff Band 1 */ + { 31, 107, 169 }, + { 35, 99, 159 }, + { 17, 82, 140 }, + { 8, 66, 114 }, + { 2, 44, 76 }, + { 1, 19, 32 } + }, { /* Coeff Band 2 */ + { 40, 132, 201 }, + { 29, 114, 187 }, + { 13, 91, 157 }, + { 7, 75, 127 }, + { 3, 58, 95 }, + { 1, 28, 47 } + }, { /* Coeff Band 3 */ + { 69, 142, 221 }, + { 42, 122, 201 }, + { 15, 91, 159 }, + { 6, 67, 121 }, + { 1, 42, 77 }, + { 1, 17, 31 } + }, { /* Coeff Band 4 */ + { 102, 148, 228 }, + { 67, 117, 204 }, + { 17, 82, 154 }, + { 6, 59, 114 }, + { 2, 39, 75 }, + { 1, 15, 29 } + }, { /* Coeff Band 5 */ + { 156, 57, 233 }, + { 119, 57, 212 }, + { 58, 48, 163 }, + { 29, 40, 124 }, + { 12, 30, 81 }, + { 3, 12, 31 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 191, 107, 226 }, + { 124, 117, 204 }, + { 25, 99, 155 } + }, { /* Coeff Band 1 */ + { 29, 148, 210 }, + { 37, 126, 194 }, + { 8, 93, 157 }, + { 2, 68, 118 }, + { 1, 39, 69 }, + { 1, 17, 33 } + }, { /* Coeff Band 2 */ + { 41, 151, 213 }, + { 27, 123, 193 }, + { 3, 82, 144 }, + { 1, 58, 105 }, + { 1, 32, 60 }, + { 1, 13, 26 } + }, { /* Coeff Band 3 */ + { 59, 159, 220 }, + { 23, 126, 198 }, + { 4, 88, 151 }, + { 1, 66, 114 }, + { 1, 38, 71 }, + { 1, 18, 34 } + }, { /* Coeff Band 4 */ + { 114, 136, 232 }, + { 51, 114, 207 }, + { 11, 83, 155 }, + { 3, 56, 105 }, + { 1, 33, 65 }, + { 1, 17, 34 } + }, { /* Coeff Band 5 */ + { 149, 65, 234 }, + { 121, 57, 215 }, + { 61, 49, 166 }, + { 28, 36, 114 }, + { 12, 25, 76 }, + { 3, 16, 42 } + } + } + }, { /* block Type 1 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 214, 49, 220 }, + { 132, 63, 188 }, + { 42, 65, 137 } + }, { /* Coeff Band 1 */ + { 85, 137, 221 }, + { 104, 131, 216 }, + { 49, 111, 192 }, + { 21, 87, 155 }, + { 2, 49, 87 }, + { 1, 16, 28 } + }, { /* Coeff Band 2 */ + { 89, 163, 230 }, + { 90, 137, 220 }, + { 29, 100, 183 }, + { 10, 70, 135 }, + { 2, 42, 81 }, + { 1, 17, 33 } + }, { /* Coeff Band 3 */ + { 108, 167, 237 }, + { 55, 133, 222 }, + { 15, 97, 179 }, + { 4, 72, 135 }, + { 1, 45, 85 }, + { 1, 19, 38 } + }, { /* Coeff Band 4 */ + { 124, 146, 240 }, + { 66, 124, 224 }, + { 17, 88, 175 }, + { 4, 58, 122 }, + { 1, 36, 75 }, + { 1, 18, 37 } + }, { /* Coeff Band 5 */ + { 141, 79, 241 }, + { 126, 70, 227 }, + { 66, 58, 182 }, + { 30, 44, 136 }, + { 12, 34, 96 }, + { 2, 20, 47 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 229, 99, 249 }, + { 143, 111, 235 }, + { 46, 109, 192 } + }, { /* Coeff Band 1 */ + { 82, 158, 236 }, + { 94, 146, 224 }, + { 25, 117, 191 }, + { 9, 87, 149 }, + { 3, 56, 99 }, + { 1, 33, 57 } + }, { /* Coeff Band 2 */ + { 83, 167, 237 }, + { 68, 145, 222 }, + { 10, 103, 177 }, + { 2, 72, 131 }, + { 1, 41, 79 }, + { 1, 20, 39 } + }, { /* Coeff Band 3 */ + { 99, 167, 239 }, + { 47, 141, 224 }, + { 10, 104, 178 }, + { 2, 73, 133 }, + { 1, 44, 85 }, + { 1, 22, 47 } + }, { /* Coeff Band 4 */ + { 127, 145, 243 }, + { 71, 129, 228 }, + { 17, 93, 177 }, + { 3, 61, 124 }, + { 1, 41, 84 }, + { 1, 21, 52 } + }, { /* Coeff Band 5 */ + { 157, 78, 244 }, + { 140, 72, 231 }, + { 69, 58, 184 }, + { 31, 44, 137 }, + { 14, 38, 105 }, + { 8, 23, 61 } + } + } + } + }, { /* tx = 8x8 */ + { /* block Type 0 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 125, 34, 187 }, + { 52, 41, 133 }, + { 6, 31, 56 } + }, { /* Coeff Band 1 */ + { 37, 109, 153 }, + { 51, 102, 147 }, + { 23, 87, 128 }, + { 8, 67, 101 }, + { 1, 41, 63 }, + { 1, 19, 29 } + }, { /* Coeff Band 2 */ + { 31, 154, 185 }, + { 17, 127, 175 }, + { 6, 96, 145 }, + { 2, 73, 114 }, + { 1, 51, 82 }, + { 1, 28, 45 } + }, { /* Coeff Band 3 */ + { 23, 163, 200 }, + { 10, 131, 185 }, + { 2, 93, 148 }, + { 1, 67, 111 }, + { 1, 41, 69 }, + { 1, 14, 24 } + }, { /* Coeff Band 4 */ + { 29, 176, 217 }, + { 12, 145, 201 }, + { 3, 101, 156 }, + { 1, 69, 111 }, + { 1, 39, 63 }, + { 1, 14, 23 } + }, { /* Coeff Band 5 */ + { 57, 192, 233 }, + { 25, 154, 215 }, + { 6, 109, 167 }, + { 3, 78, 118 }, + { 1, 48, 69 }, + { 1, 21, 29 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 202, 105, 245 }, + { 108, 106, 216 }, + { 18, 90, 144 } + }, { /* Coeff Band 1 */ + { 33, 172, 219 }, + { 64, 149, 206 }, + { 14, 117, 177 }, + { 5, 90, 141 }, + { 2, 61, 95 }, + { 1, 37, 57 } + }, { /* Coeff Band 2 */ + { 33, 179, 220 }, + { 11, 140, 198 }, + { 1, 89, 148 }, + { 1, 60, 104 }, + { 1, 33, 57 }, + { 1, 12, 21 } + }, { /* Coeff Band 3 */ + { 30, 181, 221 }, + { 8, 141, 198 }, + { 1, 87, 145 }, + { 1, 58, 100 }, + { 1, 31, 55 }, + { 1, 12, 20 } + }, { /* Coeff Band 4 */ + { 32, 186, 224 }, + { 7, 142, 198 }, + { 1, 86, 143 }, + { 1, 58, 100 }, + { 1, 31, 55 }, + { 1, 12, 22 } + }, { /* Coeff Band 5 */ + { 57, 192, 227 }, + { 20, 143, 204 }, + { 3, 96, 154 }, + { 1, 68, 112 }, + { 1, 42, 69 }, + { 1, 19, 32 } + } + } + }, { /* block Type 1 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 212, 35, 215 }, + { 113, 47, 169 }, + { 29, 48, 105 } + }, { /* Coeff Band 1 */ + { 74, 129, 203 }, + { 106, 120, 203 }, + { 49, 107, 178 }, + { 19, 84, 144 }, + { 4, 50, 84 }, + { 1, 15, 25 } + }, { /* Coeff Band 2 */ + { 71, 172, 217 }, + { 44, 141, 209 }, + { 15, 102, 173 }, + { 6, 76, 133 }, + { 2, 51, 89 }, + { 1, 24, 42 } + }, { /* Coeff Band 3 */ + { 64, 185, 231 }, + { 31, 148, 216 }, + { 8, 103, 175 }, + { 3, 74, 131 }, + { 1, 46, 81 }, + { 1, 18, 30 } + }, { /* Coeff Band 4 */ + { 65, 196, 235 }, + { 25, 157, 221 }, + { 5, 105, 174 }, + { 1, 67, 120 }, + { 1, 38, 69 }, + { 1, 15, 30 } + }, { /* Coeff Band 5 */ + { 65, 204, 238 }, + { 30, 156, 224 }, + { 7, 107, 177 }, + { 2, 70, 124 }, + { 1, 42, 73 }, + { 1, 18, 34 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 225, 86, 251 }, + { 144, 104, 235 }, + { 42, 99, 181 } + }, { /* Coeff Band 1 */ + { 85, 175, 239 }, + { 112, 165, 229 }, + { 29, 136, 200 }, + { 12, 103, 162 }, + { 6, 77, 123 }, + { 2, 53, 84 } + }, { /* Coeff Band 2 */ + { 75, 183, 239 }, + { 30, 155, 221 }, + { 3, 106, 171 }, + { 1, 74, 128 }, + { 1, 44, 76 }, + { 1, 17, 28 } + }, { /* Coeff Band 3 */ + { 73, 185, 240 }, + { 27, 159, 222 }, + { 2, 107, 172 }, + { 1, 75, 127 }, + { 1, 42, 73 }, + { 1, 17, 29 } + }, { /* Coeff Band 4 */ + { 62, 190, 238 }, + { 21, 159, 222 }, + { 2, 107, 172 }, + { 1, 72, 122 }, + { 1, 40, 71 }, + { 1, 18, 32 } + }, { /* Coeff Band 5 */ + { 61, 199, 240 }, + { 27, 161, 226 }, + { 4, 113, 180 }, + { 1, 76, 129 }, + { 1, 46, 80 }, + { 1, 23, 41 } + } + } + } + }, { /* tx = 16x16 */ + { /* block Type 0 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 7, 27, 153 }, + { 5, 30, 95 }, + { 1, 16, 30 } + }, { /* Coeff Band 1 */ + { 50, 75, 127 }, + { 57, 75, 124 }, + { 27, 67, 108 }, + { 10, 54, 86 }, + { 1, 33, 52 }, + { 1, 12, 18 } + }, { /* Coeff Band 2 */ + { 43, 125, 151 }, + { 26, 108, 148 }, + { 7, 83, 122 }, + { 2, 59, 89 }, + { 1, 38, 60 }, + { 1, 17, 27 } + }, { /* Coeff Band 3 */ + { 23, 144, 163 }, + { 13, 112, 154 }, + { 2, 75, 117 }, + { 1, 50, 81 }, + { 1, 31, 51 }, + { 1, 14, 23 } + }, { /* Coeff Band 4 */ + { 18, 162, 185 }, + { 6, 123, 171 }, + { 1, 78, 125 }, + { 1, 51, 86 }, + { 1, 31, 54 }, + { 1, 14, 23 } + }, { /* Coeff Band 5 */ + { 15, 199, 227 }, + { 3, 150, 204 }, + { 1, 91, 146 }, + { 1, 55, 95 }, + { 1, 30, 53 }, + { 1, 11, 20 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 19, 55, 240 }, + { 19, 59, 196 }, + { 3, 52, 105 } + }, { /* Coeff Band 1 */ + { 41, 166, 207 }, + { 104, 153, 199 }, + { 31, 123, 181 }, + { 14, 101, 152 }, + { 5, 72, 106 }, + { 1, 36, 52 } + }, { /* Coeff Band 2 */ + { 35, 176, 211 }, + { 12, 131, 190 }, + { 2, 88, 144 }, + { 1, 60, 101 }, + { 1, 36, 60 }, + { 1, 16, 28 } + }, { /* Coeff Band 3 */ + { 28, 183, 213 }, + { 8, 134, 191 }, + { 1, 86, 142 }, + { 1, 56, 96 }, + { 1, 30, 53 }, + { 1, 12, 20 } + }, { /* Coeff Band 4 */ + { 20, 190, 215 }, + { 4, 135, 192 }, + { 1, 84, 139 }, + { 1, 53, 91 }, + { 1, 28, 49 }, + { 1, 11, 20 } + }, { /* Coeff Band 5 */ + { 13, 196, 216 }, + { 2, 137, 192 }, + { 1, 86, 143 }, + { 1, 57, 99 }, + { 1, 32, 56 }, + { 1, 13, 24 } + } + } + }, { /* block Type 1 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 211, 29, 217 }, + { 96, 47, 156 }, + { 22, 43, 87 } + }, { /* Coeff Band 1 */ + { 78, 120, 193 }, + { 111, 116, 186 }, + { 46, 102, 164 }, + { 15, 80, 128 }, + { 2, 49, 76 }, + { 1, 18, 28 } + }, { /* Coeff Band 2 */ + { 71, 161, 203 }, + { 42, 132, 192 }, + { 10, 98, 150 }, + { 3, 69, 109 }, + { 1, 44, 70 }, + { 1, 18, 29 } + }, { /* Coeff Band 3 */ + { 57, 186, 211 }, + { 30, 140, 196 }, + { 4, 93, 146 }, + { 1, 62, 102 }, + { 1, 38, 65 }, + { 1, 16, 27 } + }, { /* Coeff Band 4 */ + { 47, 199, 217 }, + { 14, 145, 196 }, + { 1, 88, 142 }, + { 1, 57, 98 }, + { 1, 36, 62 }, + { 1, 15, 26 } + }, { /* Coeff Band 5 */ + { 26, 219, 229 }, + { 5, 155, 207 }, + { 1, 94, 151 }, + { 1, 60, 104 }, + { 1, 36, 62 }, + { 1, 16, 28 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 233, 29, 248 }, + { 146, 47, 220 }, + { 43, 52, 140 } + }, { /* Coeff Band 1 */ + { 100, 163, 232 }, + { 179, 161, 222 }, + { 63, 142, 204 }, + { 37, 113, 174 }, + { 26, 89, 137 }, + { 18, 68, 97 } + }, { /* Coeff Band 2 */ + { 85, 181, 230 }, + { 32, 146, 209 }, + { 7, 100, 164 }, + { 3, 71, 121 }, + { 1, 45, 77 }, + { 1, 18, 30 } + }, { /* Coeff Band 3 */ + { 65, 187, 230 }, + { 20, 148, 207 }, + { 2, 97, 159 }, + { 1, 68, 116 }, + { 1, 40, 70 }, + { 1, 14, 29 } + }, { /* Coeff Band 4 */ + { 40, 194, 227 }, + { 8, 147, 204 }, + { 1, 94, 155 }, + { 1, 65, 112 }, + { 1, 39, 66 }, + { 1, 14, 26 } + }, { /* Coeff Band 5 */ + { 16, 208, 228 }, + { 3, 151, 207 }, + { 1, 98, 160 }, + { 1, 67, 117 }, + { 1, 41, 74 }, + { 1, 17, 31 } + } + } + } + }, { /* tx = 32x32 */ + { /* block Type 0 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 17, 38, 140 }, + { 7, 34, 80 }, + { 1, 17, 29 } + }, { /* Coeff Band 1 */ + { 37, 75, 128 }, + { 41, 76, 128 }, + { 26, 66, 116 }, + { 12, 52, 94 }, + { 2, 32, 55 }, + { 1, 10, 16 } + }, { /* Coeff Band 2 */ + { 50, 127, 154 }, + { 37, 109, 152 }, + { 16, 82, 121 }, + { 5, 59, 85 }, + { 1, 35, 54 }, + { 1, 13, 20 } + }, { /* Coeff Band 3 */ + { 40, 142, 167 }, + { 17, 110, 157 }, + { 2, 71, 112 }, + { 1, 44, 72 }, + { 1, 27, 45 }, + { 1, 11, 17 } + }, { /* Coeff Band 4 */ + { 30, 175, 188 }, + { 9, 124, 169 }, + { 1, 74, 116 }, + { 1, 48, 78 }, + { 1, 30, 49 }, + { 1, 11, 18 } + }, { /* Coeff Band 5 */ + { 10, 222, 223 }, + { 2, 150, 194 }, + { 1, 83, 128 }, + { 1, 48, 79 }, + { 1, 27, 45 }, + { 1, 11, 17 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 36, 41, 235 }, + { 29, 36, 193 }, + { 10, 27, 111 } + }, { /* Coeff Band 1 */ + { 85, 165, 222 }, + { 177, 162, 215 }, + { 110, 135, 195 }, + { 57, 113, 168 }, + { 23, 83, 120 }, + { 10, 49, 61 } + }, { /* Coeff Band 2 */ + { 85, 190, 223 }, + { 36, 139, 200 }, + { 5, 90, 146 }, + { 1, 60, 103 }, + { 1, 38, 65 }, + { 1, 18, 30 } + }, { /* Coeff Band 3 */ + { 72, 202, 223 }, + { 23, 141, 199 }, + { 2, 86, 140 }, + { 1, 56, 97 }, + { 1, 36, 61 }, + { 1, 16, 27 } + }, { /* Coeff Band 4 */ + { 55, 218, 225 }, + { 13, 145, 200 }, + { 1, 86, 141 }, + { 1, 57, 99 }, + { 1, 35, 61 }, + { 1, 13, 22 } + }, { /* Coeff Band 5 */ + { 15, 235, 212 }, + { 1, 132, 184 }, + { 1, 84, 139 }, + { 1, 57, 97 }, + { 1, 34, 56 }, + { 1, 14, 23 } + } + } + }, { /* block Type 1 */ + { /* Intra */ + { /* Coeff Band 0 */ + { 181, 21, 201 }, + { 61, 37, 123 }, + { 10, 38, 71 } + }, { /* Coeff Band 1 */ + { 47, 106, 172 }, + { 95, 104, 173 }, + { 42, 93, 159 }, + { 18, 77, 131 }, + { 4, 50, 81 }, + { 1, 17, 23 } + }, { /* Coeff Band 2 */ + { 62, 147, 199 }, + { 44, 130, 189 }, + { 28, 102, 154 }, + { 18, 75, 115 }, + { 2, 44, 65 }, + { 1, 12, 19 } + }, { /* Coeff Band 3 */ + { 55, 153, 210 }, + { 24, 130, 194 }, + { 3, 93, 146 }, + { 1, 61, 97 }, + { 1, 31, 50 }, + { 1, 10, 16 } + }, { /* Coeff Band 4 */ + { 49, 186, 223 }, + { 17, 148, 204 }, + { 1, 96, 142 }, + { 1, 53, 83 }, + { 1, 26, 44 }, + { 1, 11, 17 } + }, { /* Coeff Band 5 */ + { 13, 217, 212 }, + { 2, 136, 180 }, + { 1, 78, 124 }, + { 1, 50, 83 }, + { 1, 29, 49 }, + { 1, 14, 23 } + } + }, { /* Inter */ + { /* Coeff Band 0 */ + { 197, 13, 247 }, + { 82, 17, 222 }, + { 25, 17, 162 } + }, { /* Coeff Band 1 */ + { 126, 186, 247 }, + { 234, 191, 243 }, + { 176, 177, 234 }, + { 104, 158, 220 }, + { 66, 128, 186 }, + { 55, 90, 137 } + }, { /* Coeff Band 2 */ + { 111, 197, 242 }, + { 46, 158, 219 }, + { 9, 104, 171 }, + { 2, 65, 125 }, + { 1, 44, 80 }, + { 1, 17, 91 } + }, { /* Coeff Band 3 */ + { 104, 208, 245 }, + { 39, 168, 224 }, + { 3, 109, 162 }, + { 1, 79, 124 }, + { 1, 50, 102 }, + { 1, 43, 102 } + }, { /* Coeff Band 4 */ + { 84, 220, 246 }, + { 31, 177, 231 }, + { 2, 115, 180 }, + { 1, 79, 134 }, + { 1, 55, 77 }, + { 1, 60, 79 } + }, { /* Coeff Band 5 */ + { 43, 243, 240 }, + { 8, 180, 217 }, + { 1, 115, 166 }, + { 1, 84, 121 }, + { 1, 51, 67 }, + { 1, 16, 6 } + } + } + } + } +}; + +const int8_t ff_vp9_mv_joint_tree[3][2] = { + { -MV_JOINT_ZERO, 1 }, // '0' + { -MV_JOINT_H, 2 }, // '10' + { -MV_JOINT_V, -MV_JOINT_HV }, // '11x' +}; + +const int8_t ff_vp9_mv_class_tree[10][2] = { + { -0, 1 }, // '0' + { -1, 2 }, // '10' + { 3, 4 }, + { -2, -3 }, // '110x' + { 5, 6 }, + { -4, -5 }, // '1110x' + { -6, 7 }, // '11110' + { 8, 9 }, + { -7, -8 }, // '111110x' + { -9, -10 }, // '111111x' +}; + +const int8_t ff_vp9_mv_fp_tree[3][2] = { + { -0, 1 }, // '0' + { -1, 2 }, // '10' + { -2, -3 }, // '11x' +}; diff --git a/libavcodec/vp9data.h b/libavcodec/vp9data.h new file mode 100644 index 0000000000..a52cc0a353 --- /dev/null +++ b/libavcodec/vp9data.h @@ -0,0 +1,70 @@ +/* + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVCODEC_VP9DATA_H +#define AVCODEC_VP9DATA_H + +#include + +#include "vp9.h" + +extern const int8_t ff_vp9_partition_tree[3][2]; +extern const uint8_t ff_vp9_default_kf_partition_probs[4][4][3]; +extern const int8_t ff_vp9_segmentation_tree[7][2]; +extern const int8_t ff_vp9_intramode_tree[9][2]; +extern const uint8_t ff_vp9_default_kf_ymode_probs[10][10][9]; +extern const uint8_t ff_vp9_default_kf_uvmode_probs[10][9]; +extern const int8_t ff_vp9_inter_mode_tree[3][2]; +extern const int8_t ff_vp9_filter_tree[2][2]; +extern const enum FilterMode ff_vp9_filter_lut[3]; +extern const int16_t ff_vp9_dc_qlookup[256]; +extern const int16_t ff_vp9_ac_qlookup[256]; +extern const enum TxfmType ff_vp9_intra_txfm_type[14]; +extern const int16_t ff_vp9_default_scan_4x4[16]; +extern const int16_t ff_vp9_col_scan_4x4[16]; +extern const int16_t ff_vp9_row_scan_4x4[16]; +extern const int16_t ff_vp9_default_scan_8x8[64]; +extern const int16_t ff_vp9_col_scan_8x8[64]; +extern const int16_t ff_vp9_row_scan_8x8[64]; +extern const int16_t ff_vp9_default_scan_16x16[256]; +extern const int16_t ff_vp9_col_scan_16x16[256]; +extern const int16_t ff_vp9_row_scan_16x16[256]; +extern const int16_t ff_vp9_default_scan_32x32[1024]; +extern const int16_t *ff_vp9_scans[5][4]; +extern const int16_t ff_vp9_default_scan_4x4_nb[16][2]; +extern const int16_t ff_vp9_col_scan_4x4_nb[16][2]; +extern const int16_t ff_vp9_row_scan_4x4_nb[16][2]; +extern const int16_t ff_vp9_default_scan_8x8_nb[64][2]; +extern const int16_t ff_vp9_col_scan_8x8_nb[64][2]; +extern const int16_t ff_vp9_row_scan_8x8_nb[64][2]; +extern const int16_t ff_vp9_default_scan_16x16_nb[256][2]; +extern const int16_t ff_vp9_col_scan_16x16_nb[256][2]; +extern const int16_t ff_vp9_row_scan_16x16_nb[256][2]; +extern const int16_t ff_vp9_default_scan_32x32_nb[1024][2]; +extern const int16_t (*ff_vp9_scans_nb[5][4])[2]; +extern const uint8_t ff_vp9_model_pareto8[256][8]; +extern const ProbContext ff_vp9_default_probs; +extern const uint8_t ff_vp9_default_coef_probs[4][2][2][6][6][3]; +extern const int8_t ff_vp9_mv_joint_tree[3][2]; +extern const int8_t ff_vp9_mv_class_tree[10][2]; +extern const int8_t ff_vp9_mv_fp_tree[3][2]; + +#endif /* AVCODEC_VP9DATA_H */ diff --git a/libavcodec/vp9dsp.c b/libavcodec/vp9dsp.c new file mode 100644 index 0000000000..c83defeda3 --- /dev/null +++ b/libavcodec/vp9dsp.c @@ -0,0 +1,2174 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/common.h" +#include "libavutil/intreadwrite.h" + +#include "rnd_avg.h" +#include "vp9.h" + +// FIXME see whether we can merge parts of this (perhaps at least 4x4 and 8x8) +// back with h264pred.[ch] + +static void vert_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + unsigned p4 = AV_RN32A(top); + + AV_WN32A(dst + stride * 0, p4); + AV_WN32A(dst + stride * 1, p4); + AV_WN32A(dst + stride * 2, p4); + AV_WN32A(dst + stride * 3, p4); +} + +static void vert_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t p8 = AV_RN64A(top); + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, p8); + dst += stride; + } +} + +static void vert_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t p8a = AV_RN64A(top + 0), p8b = AV_RN64A(top + 8); + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, p8a); + AV_WN64A(dst + 8, p8b); + dst += stride; + } +} + +static void vert_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t p8a = AV_RN64A(top + 0), p8b = AV_RN64A(top + 8), + p8c = AV_RN64A(top + 16), p8d = AV_RN64A(top + 24); + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, p8a); + AV_WN64A(dst + 8, p8b); + AV_WN64A(dst + 16, p8c); + AV_WN64A(dst + 24, p8d); + dst += stride; + } +} + +static void hor_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + AV_WN32A(dst + stride * 0, left[0] * 0x01010101U); + AV_WN32A(dst + stride * 1, left[1] * 0x01010101U); + AV_WN32A(dst + stride * 2, left[2] * 0x01010101U); + AV_WN32A(dst + stride * 3, left[3] * 0x01010101U); +} + +static void hor_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, left[y] * 0x0101010101010101ULL); + dst += stride; + } +} + +static void hor_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 16; y++) { + uint64_t p8 = left[y] * 0x0101010101010101ULL; + + AV_WN64A(dst + 0, p8); + AV_WN64A(dst + 8, p8); + dst += stride; + } +} + +static void hor_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 32; y++) { + uint64_t p8 = left[y] * 0x0101010101010101ULL; + + AV_WN64A(dst + 0, p8); + AV_WN64A(dst + 8, p8); + AV_WN64A(dst + 16, p8); + AV_WN64A(dst + 24, p8); + dst += stride; + } +} + +static void tm_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y, tl = top[-1]; + + for (y = 0; y < 4; y++) { + int l_m_tl = left[y] - tl; + + dst[0] = av_clip_uint8(top[0] + l_m_tl); + dst[1] = av_clip_uint8(top[1] + l_m_tl); + dst[2] = av_clip_uint8(top[2] + l_m_tl); + dst[3] = av_clip_uint8(top[3] + l_m_tl); + dst += stride; + } +} + +static void tm_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y, tl = top[-1]; + + for (y = 0; y < 8; y++) { + int l_m_tl = left[y] - tl; + + dst[0] = av_clip_uint8(top[0] + l_m_tl); + dst[1] = av_clip_uint8(top[1] + l_m_tl); + dst[2] = av_clip_uint8(top[2] + l_m_tl); + dst[3] = av_clip_uint8(top[3] + l_m_tl); + dst[4] = av_clip_uint8(top[4] + l_m_tl); + dst[5] = av_clip_uint8(top[5] + l_m_tl); + dst[6] = av_clip_uint8(top[6] + l_m_tl); + dst[7] = av_clip_uint8(top[7] + l_m_tl); + dst += stride; + } +} + +static void tm_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y, tl = top[-1]; + + for (y = 0; y < 16; y++) { + int l_m_tl = left[y] - tl; + + dst[0] = av_clip_uint8(top[0] + l_m_tl); + dst[1] = av_clip_uint8(top[1] + l_m_tl); + dst[2] = av_clip_uint8(top[2] + l_m_tl); + dst[3] = av_clip_uint8(top[3] + l_m_tl); + dst[4] = av_clip_uint8(top[4] + l_m_tl); + dst[5] = av_clip_uint8(top[5] + l_m_tl); + dst[6] = av_clip_uint8(top[6] + l_m_tl); + dst[7] = av_clip_uint8(top[7] + l_m_tl); + dst[8] = av_clip_uint8(top[8] + l_m_tl); + dst[9] = av_clip_uint8(top[9] + l_m_tl); + dst[10] = av_clip_uint8(top[10] + l_m_tl); + dst[11] = av_clip_uint8(top[11] + l_m_tl); + dst[12] = av_clip_uint8(top[12] + l_m_tl); + dst[13] = av_clip_uint8(top[13] + l_m_tl); + dst[14] = av_clip_uint8(top[14] + l_m_tl); + dst[15] = av_clip_uint8(top[15] + l_m_tl); + dst += stride; + } +} + +static void tm_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y, tl = top[-1]; + + for (y = 0; y < 32; y++) { + int l_m_tl = left[y] - tl; + + dst[0] = av_clip_uint8(top[0] + l_m_tl); + dst[1] = av_clip_uint8(top[1] + l_m_tl); + dst[2] = av_clip_uint8(top[2] + l_m_tl); + dst[3] = av_clip_uint8(top[3] + l_m_tl); + dst[4] = av_clip_uint8(top[4] + l_m_tl); + dst[5] = av_clip_uint8(top[5] + l_m_tl); + dst[6] = av_clip_uint8(top[6] + l_m_tl); + dst[7] = av_clip_uint8(top[7] + l_m_tl); + dst[8] = av_clip_uint8(top[8] + l_m_tl); + dst[9] = av_clip_uint8(top[9] + l_m_tl); + dst[10] = av_clip_uint8(top[10] + l_m_tl); + dst[11] = av_clip_uint8(top[11] + l_m_tl); + dst[12] = av_clip_uint8(top[12] + l_m_tl); + dst[13] = av_clip_uint8(top[13] + l_m_tl); + dst[14] = av_clip_uint8(top[14] + l_m_tl); + dst[15] = av_clip_uint8(top[15] + l_m_tl); + dst[16] = av_clip_uint8(top[16] + l_m_tl); + dst[17] = av_clip_uint8(top[17] + l_m_tl); + dst[18] = av_clip_uint8(top[18] + l_m_tl); + dst[19] = av_clip_uint8(top[19] + l_m_tl); + dst[20] = av_clip_uint8(top[20] + l_m_tl); + dst[21] = av_clip_uint8(top[21] + l_m_tl); + dst[22] = av_clip_uint8(top[22] + l_m_tl); + dst[23] = av_clip_uint8(top[23] + l_m_tl); + dst[24] = av_clip_uint8(top[24] + l_m_tl); + dst[25] = av_clip_uint8(top[25] + l_m_tl); + dst[26] = av_clip_uint8(top[26] + l_m_tl); + dst[27] = av_clip_uint8(top[27] + l_m_tl); + dst[28] = av_clip_uint8(top[28] + l_m_tl); + dst[29] = av_clip_uint8(top[29] + l_m_tl); + dst[30] = av_clip_uint8(top[30] + l_m_tl); + dst[31] = av_clip_uint8(top[31] + l_m_tl); + dst += stride; + } +} + +static void dc_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + unsigned dc = 0x01010101U * + ((left[0] + left[1] + left[2] + left[3] + + top[0] + top[1] + top[2] + top[3] + 4) >> 3); + + AV_WN32A(dst + stride * 0, dc); + AV_WN32A(dst + stride * 1, dc); + AV_WN32A(dst + stride * 2, dc); + AV_WN32A(dst + stride * 3, dc); +} + +static void dc_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + + top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + 8) >> 4); + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, dc); + dst += stride; + } +} + +static void dc_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + + left[8] + left[9] + left[10] + left[11] + + left[12] + left[13] + left[14] + left[15] + + top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + + top[8] + top[9] + top[10] + top[11] + + top[12] + top[13] + top[14] + top[15] + 16) >> 5); + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + dst += stride; + } +} + +static void dc_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + + left[8] + left[9] + left[10] + left[11] + + left[12] + left[13] + left[14] + left[15] + + left[16] + left[17] + left[18] + left[19] + + left[20] + left[21] + left[22] + left[23] + + left[24] + left[25] + left[26] + left[27] + + left[28] + left[29] + left[30] + left[31] + + top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + + top[8] + top[9] + top[10] + top[11] + + top[12] + top[13] + top[14] + top[15] + + top[16] + top[17] + top[18] + top[19] + + top[20] + top[21] + top[22] + top[23] + + top[24] + top[25] + top[26] + top[27] + + top[28] + top[29] + top[30] + top[31] + 32) >> 6); + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + AV_WN64A(dst + 16, dc); + AV_WN64A(dst + 24, dc); + dst += stride; + } +} + +static void dc_left_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + unsigned dc = 0x01010101U * + ((left[0] + left[1] + left[2] + left[3] + 2) >> 2); + + AV_WN32A(dst + stride * 0, dc); + AV_WN32A(dst + stride * 1, dc); + AV_WN32A(dst + stride * 2, dc); + AV_WN32A(dst + stride * 3, dc); +} + +static void dc_left_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + 4) >> 3); + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, dc); + dst += stride; + } +} + +static void dc_left_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + + left[8] + left[9] + left[10] + left[11] + + left[12] + left[13] + left[14] + left[15] + 8) >> 4); + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + dst += stride; + } +} + +static void dc_left_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((left[0] + left[1] + left[2] + left[3] + + left[4] + left[5] + left[6] + left[7] + + left[8] + left[9] + left[10] + left[11] + + left[12] + left[13] + left[14] + left[15] + + left[16] + left[17] + left[18] + left[19] + + left[20] + left[21] + left[22] + left[23] + + left[24] + left[25] + left[26] + left[27] + + left[28] + left[29] + left[30] + left[31] + 16) >> 5); + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + AV_WN64A(dst + 16, dc); + AV_WN64A(dst + 24, dc); + dst += stride; + } +} + +static void dc_top_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + unsigned dc = 0x01010101U * ((top[0] + top[1] + top[2] + top[3] + 2) >> 2); + + AV_WN32A(dst + stride * 0, dc); + AV_WN32A(dst + stride * 1, dc); + AV_WN32A(dst + stride * 2, dc); + AV_WN32A(dst + stride * 3, dc); +} + +static void dc_top_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + 4) >> 3); + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, dc); + dst += stride; + } +} + +static void dc_top_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + + top[8] + top[9] + top[10] + top[11] + + top[12] + top[13] + top[14] + top[15] + 8) >> 4); + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + dst += stride; + } +} + +static void dc_top_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + uint64_t dc = 0x0101010101010101ULL * + ((top[0] + top[1] + top[2] + top[3] + + top[4] + top[5] + top[6] + top[7] + + top[8] + top[9] + top[10] + top[11] + + top[12] + top[13] + top[14] + top[15] + + top[16] + top[17] + top[18] + top[19] + + top[20] + top[21] + top[22] + top[23] + + top[24] + top[25] + top[26] + top[27] + + top[28] + top[29] + top[30] + top[31] + 16) >> 5); + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, dc); + AV_WN64A(dst + 8, dc); + AV_WN64A(dst + 16, dc); + AV_WN64A(dst + 24, dc); + dst += stride; + } +} + +static void dc_128_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + AV_WN32A(dst + stride * 0, 0x80808080U); + AV_WN32A(dst + stride * 1, 0x80808080U); + AV_WN32A(dst + stride * 2, 0x80808080U); + AV_WN32A(dst + stride * 3, 0x80808080U); +} + +static void dc_128_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, 0x8080808080808080ULL); + dst += stride; + } +} + +static void dc_128_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, 0x8080808080808080ULL); + AV_WN64A(dst + 8, 0x8080808080808080ULL); + dst += stride; + } +} + +static void dc_128_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, 0x8080808080808080ULL); + AV_WN64A(dst + 8, 0x8080808080808080ULL); + AV_WN64A(dst + 16, 0x8080808080808080ULL); + AV_WN64A(dst + 24, 0x8080808080808080ULL); + dst += stride; + } +} + +static void dc_127_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + AV_WN32A(dst + stride * 0, 0x7F7F7F7FU); + AV_WN32A(dst + stride * 1, 0x7F7F7F7FU); + AV_WN32A(dst + stride * 2, 0x7F7F7F7FU); + AV_WN32A(dst + stride * 3, 0x7F7F7F7FU); +} + +static void dc_127_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, 0x7F7F7F7F7F7F7F7FULL); + dst += stride; + } +} + +static void dc_127_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, 0x7F7F7F7F7F7F7F7FULL); + AV_WN64A(dst + 8, 0x7F7F7F7F7F7F7F7FULL); + dst += stride; + } +} + +static void dc_127_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, 0x7F7F7F7F7F7F7F7FULL); + AV_WN64A(dst + 8, 0x7F7F7F7F7F7F7F7FULL); + AV_WN64A(dst + 16, 0x7F7F7F7F7F7F7F7FULL); + AV_WN64A(dst + 24, 0x7F7F7F7F7F7F7F7FULL); + dst += stride; + } +} + +static void dc_129_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + AV_WN32A(dst + stride * 0, 0x81818181U); + AV_WN32A(dst + stride * 1, 0x81818181U); + AV_WN32A(dst + stride * 2, 0x81818181U); + AV_WN32A(dst + stride * 3, 0x81818181U); +} + +static void dc_129_8x8_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 8; y++) { + AV_WN64A(dst, 0x8181818181818181ULL); + dst += stride; + } +} + +static void dc_129_16x16_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 16; y++) { + AV_WN64A(dst + 0, 0x8181818181818181ULL); + AV_WN64A(dst + 8, 0x8181818181818181ULL); + dst += stride; + } +} + +static void dc_129_32x32_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int y; + + for (y = 0; y < 32; y++) { + AV_WN64A(dst + 0, 0x8181818181818181ULL); + AV_WN64A(dst + 8, 0x8181818181818181ULL); + AV_WN64A(dst + 16, 0x8181818181818181ULL); + AV_WN64A(dst + 24, 0x8181818181818181ULL); + dst += stride; + } +} + +#define DST(x, y) dst[(x) + (y) * stride] + +static void diag_downleft_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3], + a4 = top[4], a5 = top[5], a6 = top[6], a7 = top[7]; + + DST(0, 0) = (a0 + a1 * 2 + a2 + 2) >> 2; + DST(1, 0) = + DST(0, 1) = (a1 + a2 * 2 + a3 + 2) >> 2; + DST(2, 0) = + DST(1, 1) = + DST(0, 2) = (a2 + a3 * 2 + a4 + 2) >> 2; + DST(3, 0) = + DST(2, 1) = + DST(1, 2) = + DST(0, 3) = (a3 + a4 * 2 + a5 + 2) >> 2; + DST(3, 1) = + DST(2, 2) = + DST(1, 3) = (a4 + a5 * 2 + a6 + 2) >> 2; + DST(3, 2) = + DST(2, 3) = (a5 + a6 * 2 + a7 + 2) >> 2; + DST(3, 3) = a7; // note: this is different from vp8 and such +} + +#define def_diag_downleft(size) \ +static void diag_downleft_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t v[size - 1]; \ + \ + for (i = 0; i < size - 2; i++) \ + v[i] = (top[i] + top[i + 1] * 2 + top[i + 2] + 2) >> 2; \ + v[size - 2] = (top[size - 2] + top[size - 1] * 3 + 2) >> 2; \ + \ + for (j = 0; j < size; j++) { \ + memcpy(dst + j * stride, v + j, size - 1 - j); \ + memset(dst + j * stride + size - 1 - j, top[size - 1], j + 1); \ + } \ +} + +def_diag_downleft(8) +def_diag_downleft(16) +def_diag_downleft(32) + +static void diag_downright_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3], + l0 = left[0], l1 = left[1], l2 = left[2], l3 = left[3]; + + DST(0, 3) = (l1 + l2 * 2 + l3 + 2) >> 2; + DST(0, 2) = + DST(1, 3) = (l0 + l1 * 2 + l2 + 2) >> 2; + DST(0, 1) = + DST(1, 2) = + DST(2, 3) = (tl + l0 * 2 + l1 + 2) >> 2; + DST(0, 0) = + DST(1, 1) = + DST(2, 2) = + DST(3, 3) = (l0 + tl * 2 + a0 + 2) >> 2; + DST(1, 0) = + DST(2, 1) = + DST(3, 2) = (tl + a0 * 2 + a1 + 2) >> 2; + DST(2, 0) = + DST(3, 1) = (a0 + a1 * 2 + a2 + 2) >> 2; + DST(3, 0) = (a1 + a2 * 2 + a3 + 2) >> 2; +} + +#define def_diag_downright(size) \ +static void diag_downright_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t v[size + size - 1]; \ + \ + for (i = 0; i < size - 2; i++) { \ + v[i] = (left[size - 1 - i] + \ + left[size - 2 - i] * 2 + \ + left[size - 3 - i] + 2) >> 2; \ + v[size + 1 + i] = (top[i] + \ + top[i + 1] * 2 + \ + top[i + 2] + 2) >> 2; \ + } \ + v[size - 2] = (left[1] + left[0] * 2 + top[-1] + 2) >> 2; \ + v[size - 1] = (left[0] + top[-1] * 2 + top[0] + 2) >> 2; \ + v[size] = (top[-1] + top[0] * 2 + top[1] + 2) >> 2; \ + \ + for (j = 0; j < size; j++) \ + memcpy(dst + j * stride, v + size - 1 - j, size); \ +} + +def_diag_downright(8) +def_diag_downright(16) +def_diag_downright(32) + +static void vert_right_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3], + l0 = left[0], l1 = left[1], l2 = left[2]; + + DST(0, 3) = (l0 + l1 * 2 + l2 + 2) >> 2; + DST(0, 2) = (tl + l0 * 2 + l1 + 2) >> 2; + DST(0, 0) = + DST(1, 2) = (tl + a0 + 1) >> 1; + DST(0, 1) = + DST(1, 3) = (l0 + tl * 2 + a0 + 2) >> 2; + DST(1, 0) = + DST(2, 2) = (a0 + a1 + 1) >> 1; + DST(1, 1) = + DST(2, 3) = (tl + a0 * 2 + a1 + 2) >> 2; + DST(2, 0) = + DST(3, 2) = (a1 + a2 + 1) >> 1; + DST(2, 1) = + DST(3, 3) = (a0 + a1 * 2 + a2 + 2) >> 2; + DST(3, 0) = (a2 + a3 + 1) >> 1; + DST(3, 1) = (a1 + a2 * 2 + a3 + 2) >> 2; +} + +#define def_vert_right(size) \ +static void vert_right_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t ve[size + size / 2 - 1], vo[size + size / 2 - 1]; \ + \ + for (i = 0; i < size / 2 - 2; i++) { \ + vo[i] = (left[size - 4 - i * 2] + \ + left[size - 3 - i * 2] * 2 + \ + left[size - 2 - i * 2] + 2) >> 2; \ + ve[i] = (left[size - 5 - i * 2] + \ + left[size - 4 - i * 2] * 2 + \ + left[size - 3 - i * 2] + 2) >> 2; \ + } \ + vo[size / 2 - 2] = (left[0] + left[1] * 2 + left[2] + 2) >> 2; \ + ve[size / 2 - 2] = (top[-1] + left[0] * 2 + left[1] + 2) >> 2; \ + \ + ve[size / 2 - 1] = (top[-1] + top[0] + 1) >> 1; \ + vo[size / 2 - 1] = (left[0] + top[-1] * 2 + top[0] + 2) >> 2; \ + for (i = 0; i < size - 1; i++) { \ + ve[size / 2 + i] = (top[i] + top[i + 1] + 1) >> 1; \ + vo[size / 2 + i] = (top[i - 1] + top[i] * 2 + top[i + 1] + 2) >> 2; \ + } \ + \ + for (j = 0; j < size / 2; j++) { \ + memcpy(dst + j * 2 * stride, ve + size / 2 - 1 - j, size); \ + memcpy(dst + (j * 2 + 1) * stride, vo + size / 2 - 1 - j, size); \ + } \ +} + +def_vert_right(8) +def_vert_right(16) +def_vert_right(32) + +static void hor_down_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int l0 = left[0], l1 = left[1], l2 = left[2], l3 = left[3], + tl = top[-1], a0 = top[0], a1 = top[1], a2 = top[2]; + + DST(2, 0) = (tl + a0 * 2 + a1 + 2) >> 2; + DST(3, 0) = (a0 + a1 * 2 + a2 + 2) >> 2; + DST(0, 0) = + DST(2, 1) = (tl + l0 + 1) >> 1; + DST(1, 0) = + DST(3, 1) = (a0 + tl * 2 + l0 + 2) >> 2; + DST(0, 1) = + DST(2, 2) = (l0 + l1 + 1) >> 1; + DST(1, 1) = + DST(3, 2) = (tl + l0 * 2 + l1 + 2) >> 2; + DST(0, 2) = + DST(2, 3) = (l1 + l2 + 1) >> 1; + DST(1, 2) = + DST(3, 3) = (l0 + l1 * 2 + l2 + 2) >> 2; + DST(0, 3) = (l2 + l3 + 1) >> 1; + DST(1, 3) = (l1 + l2 * 2 + l3 + 2) >> 2; +} + +#define def_hor_down(size) \ +static void hor_down_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t v[size * 3 - 2]; \ + \ + for (i = 0; i < size - 2; i++) { \ + v[i * 2] = (left[size - 2 - i] + \ + left[size - 1 - i] + 1) >> 1; \ + v[i * 2 + 1] = (left[size - 3 - i] + \ + left[size - 2 - i] * 2 + \ + left[size - 1 - i] + 2) >> 2; \ + v[size * 2 + i] = (top[i - 1] + \ + top[i] * 2 + \ + top[i + 1] + 2) >> 2; \ + } \ + v[size * 2 - 2] = (top[-1] + left[0] + 1) >> 1; \ + v[size * 2 - 4] = (left[0] + left[1] + 1) >> 1; \ + v[size * 2 - 1] = (top[0] + top[-1] * 2 + left[0] + 2) >> 2; \ + v[size * 2 - 3] = (top[-1] + left[0] * 2 + left[1] + 2) >> 2; \ + \ + for (j = 0; j < size; j++) \ + memcpy(dst + j * stride, v + size * 2 - 2 - j * 2, size); \ +} + +def_hor_down(8) +def_hor_down(16) +def_hor_down(32) + +static void vert_left_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int a0 = top[0], a1 = top[1], a2 = top[2], a3 = top[3], + a4 = top[4], a5 = top[5], a6 = top[6]; + + DST(0, 0) = (a0 + a1 + 1) >> 1; + DST(0, 1) = (a0 + a1 * 2 + a2 + 2) >> 2; + DST(1, 0) = + DST(0, 2) = (a1 + a2 + 1) >> 1; + DST(1, 1) = + DST(0, 3) = (a1 + a2 * 2 + a3 + 2) >> 2; + DST(2, 0) = + DST(1, 2) = (a2 + a3 + 1) >> 1; + DST(2, 1) = + DST(1, 3) = (a2 + a3 * 2 + a4 + 2) >> 2; + DST(3, 0) = + DST(2, 2) = (a3 + a4 + 1) >> 1; + DST(3, 1) = + DST(2, 3) = (a3 + a4 * 2 + a5 + 2) >> 2; + DST(3, 2) = (a4 + a5 + 1) >> 1; + DST(3, 3) = (a4 + a5 * 2 + a6 + 2) >> 2; +} + +#define def_vert_left(size) \ +static void vert_left_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t ve[size - 1], vo[size - 1]; \ + \ + for (i = 0; i < size - 2; i++) { \ + ve[i] = (top[i] + top[i + 1] + 1) >> 1; \ + vo[i] = (top[i] + top[i + 1] * 2 + top[i + 2] + 2) >> 2; \ + } \ + ve[size - 2] = (top[size - 2] + top[size - 1] + 1) >> 1; \ + vo[size - 2] = (top[size - 2] + top[size - 1] * 3 + 2) >> 2; \ + \ + for (j = 0; j < size / 2; j++) { \ + memcpy(dst + j * 2 * stride, ve + j, size - (j + 1)); \ + memset(dst + j * 2 * stride + size - j - 1, \ + top[size - 1], j + 1); \ + memcpy(dst + (j * 2 + 1) * stride, vo + j, size - (j + 1)); \ + memset(dst + (j * 2 + 1) * stride + size - j - 1, \ + top[size - 1], j + 1); \ + } \ +} + +def_vert_left(8) +def_vert_left(16) +def_vert_left(32) + +static void hor_up_4x4_c(uint8_t *dst, ptrdiff_t stride, + const uint8_t *left, const uint8_t *top) +{ + int l0 = left[0], l1 = left[1], l2 = left[2], l3 = left[3]; + + DST(0, 0) = (l0 + l1 + 1) >> 1; + DST(1, 0) = (l0 + l1 * 2 + l2 + 2) >> 2; + DST(0, 1) = + DST(2, 0) = (l1 + l2 + 1) >> 1; + DST(1, 1) = + DST(3, 0) = (l1 + l2 * 2 + l3 + 2) >> 2; + DST(0, 2) = + DST(2, 1) = (l2 + l3 + 1) >> 1; + DST(1, 2) = + DST(3, 1) = (l2 + l3 * 3 + 2) >> 2; + DST(0, 3) = + DST(1, 3) = + DST(2, 2) = + DST(2, 3) = + DST(3, 2) = + DST(3, 3) = l3; +} + +#define def_hor_up(size) \ +static void hor_up_ ## size ## x ## size ## _c(uint8_t *dst, \ + ptrdiff_t stride, \ + const uint8_t *left, \ + const uint8_t *top) \ +{ \ + int i, j; \ + uint8_t v[size * 2 - 2]; \ + \ + for (i = 0; i < size - 2; i++) { \ + v[i * 2] = (left[i] + left[i + 1] + 1) >> 1; \ + v[i * 2 + 1] = (left[i] + left[i + 1] * 2 + left[i + 2] + 2) >> 2; \ + } \ + v[size * 2 - 4] = (left[size - 2] + left[size - 1] + 1) >> 1; \ + v[size * 2 - 3] = (left[size - 2] + left[size - 1] * 3 + 2) >> 2; \ + \ + for (j = 0; j < size / 2; j++) \ + memcpy(dst + j * stride, v + j * 2, size); \ + for (j = size / 2; j < size; j++) { \ + memcpy(dst + j * stride, v + j * 2, size * 2 - 2 - j * 2); \ + memset(dst + j * stride + size * 2 - 2 - j * 2, left[size - 1], \ + 2 + j * 2 - size); \ + } \ +} + +def_hor_up(8) +def_hor_up(16) +def_hor_up(32) + +#undef DST + +static av_cold void vp9dsp_intrapred_init(VP9DSPContext *dsp) +{ +#define init_intra_pred(tx, sz) \ + dsp->intra_pred[tx][VERT_PRED] = vert_ ## sz ## _c; \ + dsp->intra_pred[tx][HOR_PRED] = hor_ ## sz ## _c; \ + dsp->intra_pred[tx][DC_PRED] = dc_ ## sz ## _c; \ + dsp->intra_pred[tx][DIAG_DOWN_LEFT_PRED] = diag_downleft_ ## sz ## _c; \ + dsp->intra_pred[tx][DIAG_DOWN_RIGHT_PRED] = diag_downright_ ## sz ## _c; \ + dsp->intra_pred[tx][VERT_RIGHT_PRED] = vert_right_ ## sz ## _c; \ + dsp->intra_pred[tx][HOR_DOWN_PRED] = hor_down_ ## sz ## _c; \ + dsp->intra_pred[tx][VERT_LEFT_PRED] = vert_left_ ## sz ## _c; \ + dsp->intra_pred[tx][HOR_UP_PRED] = hor_up_ ## sz ## _c; \ + dsp->intra_pred[tx][TM_VP8_PRED] = tm_ ## sz ## _c; \ + dsp->intra_pred[tx][LEFT_DC_PRED] = dc_left_ ## sz ## _c; \ + dsp->intra_pred[tx][TOP_DC_PRED] = dc_top_ ## sz ## _c; \ + dsp->intra_pred[tx][DC_128_PRED] = dc_128_ ## sz ## _c; \ + dsp->intra_pred[tx][DC_127_PRED] = dc_127_ ## sz ## _c; \ + dsp->intra_pred[tx][DC_129_PRED] = dc_129_ ## sz ## _c + + init_intra_pred(TX_4X4, 4x4); + init_intra_pred(TX_8X8, 8x8); + init_intra_pred(TX_16X16, 16x16); + init_intra_pred(TX_32X32, 32x32); + +#undef init_intra_pred +} + +#define itxfm_wrapper(type_a, type_b, sz, bits) \ +static void \ +type_a ## _ ## type_b ## _ ## sz ## x ## sz ## _add_c(uint8_t *dst, \ + ptrdiff_t stride, \ + int16_t *block, \ + int eob) \ +{ \ + int i, j; \ + int16_t tmp[sz * sz], out[sz]; \ + for (i = 0; i < sz; i++) \ + type_a ## sz ## _1d(tmp + i * sz, block + i, sz, 0); \ + memset(block, 0, sz * sz * sizeof(*block)); \ + for (i = 0; i < sz; i++) { \ + type_b ## sz ## _1d(out, tmp + i, sz, 1); \ + for (j = 0; j < sz; j++) \ + dst[j * stride] = \ + av_clip_uint8(dst[j * stride] + \ + (bits ? (out[j] + (1 << (bits - 1))) >> bits \ + : out[j])); \ + dst++; \ + } \ +} + +#define itxfm_wrap(sz, bits) \ + itxfm_wrapper(idct, idct, sz, bits) \ + itxfm_wrapper(iadst, idct, sz, bits) \ + itxfm_wrapper(idct, iadst, sz, bits) \ + itxfm_wrapper(iadst, iadst, sz, bits) + +#define IN(x) in[x * stride] + +static av_always_inline void idct4_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t1, t2, t3; + + t0 = ((IN(0) + IN(2)) * 11585 + (1 << 13)) >> 14; + t1 = ((IN(0) - IN(2)) * 11585 + (1 << 13)) >> 14; + t2 = (IN(1) * 6270 - IN(3) * 15137 + (1 << 13)) >> 14; + t3 = (IN(1) * 15137 + IN(3) * 6270 + (1 << 13)) >> 14; + + out[0] = t0 + t3; + out[1] = t1 + t2; + out[2] = t1 - t2; + out[3] = t0 - t3; +} + +static av_always_inline void iadst4_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t1, t2, t3; + + t0 = 5283 * IN(0) + 15212 * IN(2) + 9929 * IN(3); + t1 = 9929 * IN(0) - 5283 * IN(2) - 15212 * IN(3); + t2 = 13377 * (IN(0) - IN(2) + IN(3)); + t3 = 13377 * IN(1); + + out[0] = (t0 + t3 + (1 << 13)) >> 14; + out[1] = (t1 + t3 + (1 << 13)) >> 14; + out[2] = (t2 + (1 << 13)) >> 14; + out[3] = (t0 + t1 - t3 + (1 << 13)) >> 14; +} + +itxfm_wrap(4, 4) + +static av_always_inline void idct8_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t0a, t1, t1a, t2, t2a, t3, t3a, t4, t4a, t5, t5a, t6, t6a, t7, t7a; + + t0a = ((IN(0) + IN(4)) * 11585 + (1 << 13)) >> 14; + t1a = ((IN(0) - IN(4)) * 11585 + (1 << 13)) >> 14; + t2a = (IN(2) * 6270 - IN(6) * 15137 + (1 << 13)) >> 14; + t3a = (IN(2) * 15137 + IN(6) * 6270 + (1 << 13)) >> 14; + t4a = (IN(1) * 3196 - IN(7) * 16069 + (1 << 13)) >> 14; + t5a = (IN(5) * 13623 - IN(3) * 9102 + (1 << 13)) >> 14; + t6a = (IN(5) * 9102 + IN(3) * 13623 + (1 << 13)) >> 14; + t7a = (IN(1) * 16069 + IN(7) * 3196 + (1 << 13)) >> 14; + + t0 = t0a + t3a; + t1 = t1a + t2a; + t2 = t1a - t2a; + t3 = t0a - t3a; + t4 = t4a + t5a; + t5a = t4a - t5a; + t7 = t7a + t6a; + t6a = t7a - t6a; + + t5 = ((t6a - t5a) * 11585 + (1 << 13)) >> 14; + t6 = ((t6a + t5a) * 11585 + (1 << 13)) >> 14; + + out[0] = t0 + t7; + out[1] = t1 + t6; + out[2] = t2 + t5; + out[3] = t3 + t4; + out[4] = t3 - t4; + out[5] = t2 - t5; + out[6] = t1 - t6; + out[7] = t0 - t7; +} + +static av_always_inline void iadst8_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t0a, t1, t1a, t2, t2a, t3, t3a, t4, t4a, t5, t5a, t6, t6a, t7, t7a; + + t0a = 16305 * IN(7) + 1606 * IN(0); + t1a = 1606 * IN(7) - 16305 * IN(0); + t2a = 14449 * IN(5) + 7723 * IN(2); + t3a = 7723 * IN(5) - 14449 * IN(2); + t4a = 10394 * IN(3) + 12665 * IN(4); + t5a = 12665 * IN(3) - 10394 * IN(4); + t6a = 4756 * IN(1) + 15679 * IN(6); + t7a = 15679 * IN(1) - 4756 * IN(6); + + t0 = (t0a + t4a + (1 << 13)) >> 14; + t1 = (t1a + t5a + (1 << 13)) >> 14; + t2 = (t2a + t6a + (1 << 13)) >> 14; + t3 = (t3a + t7a + (1 << 13)) >> 14; + t4 = (t0a - t4a + (1 << 13)) >> 14; + t5 = (t1a - t5a + (1 << 13)) >> 14; + t6 = (t2a - t6a + (1 << 13)) >> 14; + t7 = (t3a - t7a + (1 << 13)) >> 14; + + t4a = 15137 * t4 + 6270 * t5; + t5a = 6270 * t4 - 15137 * t5; + t6a = 15137 * t7 - 6270 * t6; + t7a = 6270 * t7 + 15137 * t6; + + out[0] = t0 + t2; + out[7] = -(t1 + t3); + t2 = t0 - t2; + t3 = t1 - t3; + + out[1] = -((t4a + t6a + (1 << 13)) >> 14); + out[6] = (t5a + t7a + (1 << 13)) >> 14; + t6 = (t4a - t6a + (1 << 13)) >> 14; + t7 = (t5a - t7a + (1 << 13)) >> 14; + + out[3] = -(((t2 + t3) * 11585 + (1 << 13)) >> 14); + out[4] = ((t2 - t3) * 11585 + (1 << 13)) >> 14; + out[2] = ((t6 + t7) * 11585 + (1 << 13)) >> 14; + out[5] = -(((t6 - t7) * 11585 + (1 << 13)) >> 14); +} + +itxfm_wrap(8, 5) + +static av_always_inline void idct16_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15; + int t0a, t1a, t2a, t3a, t4a, t5a, t6a, t7a; + int t8a, t9a, t10a, t11a, t12a, t13a, t14a, t15a; + + t0a = ((IN(0) + IN(8)) * 11585 + (1 << 13)) >> 14; + t1a = ((IN(0) - IN(8)) * 11585 + (1 << 13)) >> 14; + t2a = (IN(4) * 6270 - IN(12) * 15137 + (1 << 13)) >> 14; + t3a = (IN(4) * 15137 + IN(12) * 6270 + (1 << 13)) >> 14; + t4a = (IN(2) * 3196 - IN(14) * 16069 + (1 << 13)) >> 14; + t7a = (IN(2) * 16069 + IN(14) * 3196 + (1 << 13)) >> 14; + t5a = (IN(10) * 13623 - IN(6) * 9102 + (1 << 13)) >> 14; + t6a = (IN(10) * 9102 + IN(6) * 13623 + (1 << 13)) >> 14; + t8a = (IN(1) * 1606 - IN(15) * 16305 + (1 << 13)) >> 14; + t15a = (IN(1) * 16305 + IN(15) * 1606 + (1 << 13)) >> 14; + t9a = (IN(9) * 12665 - IN(7) * 10394 + (1 << 13)) >> 14; + t14a = (IN(9) * 10394 + IN(7) * 12665 + (1 << 13)) >> 14; + t10a = (IN(5) * 7723 - IN(11) * 14449 + (1 << 13)) >> 14; + t13a = (IN(5) * 14449 + IN(11) * 7723 + (1 << 13)) >> 14; + t11a = (IN(13) * 15679 - IN(3) * 4756 + (1 << 13)) >> 14; + t12a = (IN(13) * 4756 + IN(3) * 15679 + (1 << 13)) >> 14; + + t0 = t0a + t3a; + t1 = t1a + t2a; + t2 = t1a - t2a; + t3 = t0a - t3a; + t4 = t4a + t5a; + t5 = t4a - t5a; + t6 = t7a - t6a; + t7 = t7a + t6a; + t8 = t8a + t9a; + t9 = t8a - t9a; + t10 = t11a - t10a; + t11 = t11a + t10a; + t12 = t12a + t13a; + t13 = t12a - t13a; + t14 = t15a - t14a; + t15 = t15a + t14a; + + t5a = ((t6 - t5) * 11585 + (1 << 13)) >> 14; + t6a = ((t6 + t5) * 11585 + (1 << 13)) >> 14; + t9a = (t14 * 6270 - t9 * 15137 + (1 << 13)) >> 14; + t14a = (t14 * 15137 + t9 * 6270 + (1 << 13)) >> 14; + t10a = (-(t13 * 15137 + t10 * 6270) + (1 << 13)) >> 14; + t13a = (t13 * 6270 - t10 * 15137 + (1 << 13)) >> 14; + + t0a = t0 + t7; + t1a = t1 + t6a; + t2a = t2 + t5a; + t3a = t3 + t4; + t4 = t3 - t4; + t5 = t2 - t5a; + t6 = t1 - t6a; + t7 = t0 - t7; + t8a = t8 + t11; + t9 = t9a + t10a; + t10 = t9a - t10a; + t11a = t8 - t11; + t12a = t15 - t12; + t13 = t14a - t13a; + t14 = t14a + t13a; + t15a = t15 + t12; + + t10a = ((t13 - t10) * 11585 + (1 << 13)) >> 14; + t13a = ((t13 + t10) * 11585 + (1 << 13)) >> 14; + t11 = ((t12a - t11a) * 11585 + (1 << 13)) >> 14; + t12 = ((t12a + t11a) * 11585 + (1 << 13)) >> 14; + + out[0] = t0a + t15a; + out[1] = t1a + t14; + out[2] = t2a + t13a; + out[3] = t3a + t12; + out[4] = t4 + t11; + out[5] = t5 + t10a; + out[6] = t6 + t9; + out[7] = t7 + t8a; + out[8] = t7 - t8a; + out[9] = t6 - t9; + out[10] = t5 - t10a; + out[11] = t4 - t11; + out[12] = t3a - t12; + out[13] = t2a - t13a; + out[14] = t1a - t14; + out[15] = t0a - t15a; +} + +static av_always_inline void iadst16_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15; + int t0a, t1a, t2a, t3a, t4a, t5a, t6a, t7a; + int t8a, t9a, t10a, t11a, t12a, t13a, t14a, t15a; + + t0 = IN(15) * 16364 + IN(0) * 804; + t1 = IN(15) * 804 - IN(0) * 16364; + t2 = IN(13) * 15893 + IN(2) * 3981; + t3 = IN(13) * 3981 - IN(2) * 15893; + t4 = IN(11) * 14811 + IN(4) * 7005; + t5 = IN(11) * 7005 - IN(4) * 14811; + t6 = IN(9) * 13160 + IN(6) * 9760; + t7 = IN(9) * 9760 - IN(6) * 13160; + t8 = IN(7) * 11003 + IN(8) * 12140; + t9 = IN(7) * 12140 - IN(8) * 11003; + t10 = IN(5) * 8423 + IN(10) * 14053; + t11 = IN(5) * 14053 - IN(10) * 8423; + t12 = IN(3) * 5520 + IN(12) * 15426; + t13 = IN(3) * 15426 - IN(12) * 5520; + t14 = IN(1) * 2404 + IN(14) * 16207; + t15 = IN(1) * 16207 - IN(14) * 2404; + + t0a = (t0 + t8 + (1 << 13)) >> 14; + t1a = (t1 + t9 + (1 << 13)) >> 14; + t2a = (t2 + t10 + (1 << 13)) >> 14; + t3a = (t3 + t11 + (1 << 13)) >> 14; + t4a = (t4 + t12 + (1 << 13)) >> 14; + t5a = (t5 + t13 + (1 << 13)) >> 14; + t6a = (t6 + t14 + (1 << 13)) >> 14; + t7a = (t7 + t15 + (1 << 13)) >> 14; + t8a = (t0 - t8 + (1 << 13)) >> 14; + t9a = (t1 - t9 + (1 << 13)) >> 14; + t10a = (t2 - t10 + (1 << 13)) >> 14; + t11a = (t3 - t11 + (1 << 13)) >> 14; + t12a = (t4 - t12 + (1 << 13)) >> 14; + t13a = (t5 - t13 + (1 << 13)) >> 14; + t14a = (t6 - t14 + (1 << 13)) >> 14; + t15a = (t7 - t15 + (1 << 13)) >> 14; + + t8 = t8a * 16069 + t9a * 3196; + t9 = t8a * 3196 - t9a * 16069; + t10 = t10a * 9102 + t11a * 13623; + t11 = t10a * 13623 - t11a * 9102; + t12 = t13a * 16069 - t12a * 3196; + t13 = t13a * 3196 + t12a * 16069; + t14 = t15a * 9102 - t14a * 13623; + t15 = t15a * 13623 + t14a * 9102; + + t0 = t0a + t4a; + t1 = t1a + t5a; + t2 = t2a + t6a; + t3 = t3a + t7a; + t4 = t0a - t4a; + t5 = t1a - t5a; + t6 = t2a - t6a; + t7 = t3a - t7a; + t8a = (t8 + t12 + (1 << 13)) >> 14; + t9a = (t9 + t13 + (1 << 13)) >> 14; + t10a = (t10 + t14 + (1 << 13)) >> 14; + t11a = (t11 + t15 + (1 << 13)) >> 14; + t12a = (t8 - t12 + (1 << 13)) >> 14; + t13a = (t9 - t13 + (1 << 13)) >> 14; + t14a = (t10 - t14 + (1 << 13)) >> 14; + t15a = (t11 - t15 + (1 << 13)) >> 14; + + t4a = t4 * 15137 + t5 * 6270; + t5a = t4 * 6270 - t5 * 15137; + t6a = t7 * 15137 - t6 * 6270; + t7a = t7 * 6270 + t6 * 15137; + t12 = t12a * 15137 + t13a * 6270; + t13 = t12a * 6270 - t13a * 15137; + t14 = t15a * 15137 - t14a * 6270; + t15 = t15a * 6270 + t14a * 15137; + + out[0] = t0 + t2; + out[15] = -(t1 + t3); + t2a = t0 - t2; + t3a = t1 - t3; + out[3] = -((t4a + t6a + (1 << 13)) >> 14); + out[12] = (t5a + t7a + (1 << 13)) >> 14; + t6 = (t4a - t6a + (1 << 13)) >> 14; + t7 = (t5a - t7a + (1 << 13)) >> 14; + out[1] = -(t8a + t10a); + out[14] = t9a + t11a; + t10 = t8a - t10a; + t11 = t9a - t11a; + out[2] = (t12 + t14 + (1 << 13)) >> 14; + out[13] = -((t13 + t15 + (1 << 13)) >> 14); + t14a = (t12 - t14 + (1 << 13)) >> 14; + t15a = (t13 - t15 + (1 << 13)) >> 14; + + out[7] = ((t2a + t3a) * -11585 + (1 << 13)) >> 14; + out[8] = ((t2a - t3a) * 11585 + (1 << 13)) >> 14; + out[4] = ((t7 + t6) * 11585 + (1 << 13)) >> 14; + out[11] = ((t7 - t6) * 11585 + (1 << 13)) >> 14; + out[6] = ((t11 + t10) * 11585 + (1 << 13)) >> 14; + out[9] = ((t11 - t10) * 11585 + (1 << 13)) >> 14; + out[5] = ((t14a + t15a) * -11585 + (1 << 13)) >> 14; + out[10] = ((t14a - t15a) * 11585 + (1 << 13)) >> 14; +} + +itxfm_wrap(16, 6) + +static av_always_inline void idct32_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0a = ((IN(0) + IN(16)) * 11585 + (1 << 13)) >> 14; + int t1a = ((IN(0) - IN(16)) * 11585 + (1 << 13)) >> 14; + int t2a = (IN(8) * 6270 - IN(24) * 15137 + (1 << 13)) >> 14; + int t3a = (IN(8) * 15137 + IN(24) * 6270 + (1 << 13)) >> 14; + int t4a = (IN(4) * 3196 - IN(28) * 16069 + (1 << 13)) >> 14; + int t7a = (IN(4) * 16069 + IN(28) * 3196 + (1 << 13)) >> 14; + int t5a = (IN(20) * 13623 - IN(12) * 9102 + (1 << 13)) >> 14; + int t6a = (IN(20) * 9102 + IN(12) * 13623 + (1 << 13)) >> 14; + int t8a = (IN(2) * 1606 - IN(30) * 16305 + (1 << 13)) >> 14; + int t15a = (IN(2) * 16305 + IN(30) * 1606 + (1 << 13)) >> 14; + int t9a = (IN(18) * 12665 - IN(14) * 10394 + (1 << 13)) >> 14; + int t14a = (IN(18) * 10394 + IN(14) * 12665 + (1 << 13)) >> 14; + int t10a = (IN(10) * 7723 - IN(22) * 14449 + (1 << 13)) >> 14; + int t13a = (IN(10) * 14449 + IN(22) * 7723 + (1 << 13)) >> 14; + int t11a = (IN(26) * 15679 - IN(6) * 4756 + (1 << 13)) >> 14; + int t12a = (IN(26) * 4756 + IN(6) * 15679 + (1 << 13)) >> 14; + int t16a = (IN(1) * 804 - IN(31) * 16364 + (1 << 13)) >> 14; + int t31a = (IN(1) * 16364 + IN(31) * 804 + (1 << 13)) >> 14; + int t17a = (IN(17) * 12140 - IN(15) * 11003 + (1 << 13)) >> 14; + int t30a = (IN(17) * 11003 + IN(15) * 12140 + (1 << 13)) >> 14; + int t18a = (IN(9) * 7005 - IN(23) * 14811 + (1 << 13)) >> 14; + int t29a = (IN(9) * 14811 + IN(23) * 7005 + (1 << 13)) >> 14; + int t19a = (IN(25) * 15426 - IN(7) * 5520 + (1 << 13)) >> 14; + int t28a = (IN(25) * 5520 + IN(7) * 15426 + (1 << 13)) >> 14; + int t20a = (IN(5) * 3981 - IN(27) * 15893 + (1 << 13)) >> 14; + int t27a = (IN(5) * 15893 + IN(27) * 3981 + (1 << 13)) >> 14; + int t21a = (IN(21) * 14053 - IN(11) * 8423 + (1 << 13)) >> 14; + int t26a = (IN(21) * 8423 + IN(11) * 14053 + (1 << 13)) >> 14; + int t22a = (IN(13) * 9760 - IN(19) * 13160 + (1 << 13)) >> 14; + int t25a = (IN(13) * 13160 + IN(19) * 9760 + (1 << 13)) >> 14; + int t23a = (IN(29) * 16207 - IN(3) * 2404 + (1 << 13)) >> 14; + int t24a = (IN(29) * 2404 + IN(3) * 16207 + (1 << 13)) >> 14; + + int t0 = t0a + t3a; + int t1 = t1a + t2a; + int t2 = t1a - t2a; + int t3 = t0a - t3a; + int t4 = t4a + t5a; + int t5 = t4a - t5a; + int t6 = t7a - t6a; + int t7 = t7a + t6a; + int t8 = t8a + t9a; + int t9 = t8a - t9a; + int t10 = t11a - t10a; + int t11 = t11a + t10a; + int t12 = t12a + t13a; + int t13 = t12a - t13a; + int t14 = t15a - t14a; + int t15 = t15a + t14a; + int t16 = t16a + t17a; + int t17 = t16a - t17a; + int t18 = t19a - t18a; + int t19 = t19a + t18a; + int t20 = t20a + t21a; + int t21 = t20a - t21a; + int t22 = t23a - t22a; + int t23 = t23a + t22a; + int t24 = t24a + t25a; + int t25 = t24a - t25a; + int t26 = t27a - t26a; + int t27 = t27a + t26a; + int t28 = t28a + t29a; + int t29 = t28a - t29a; + int t30 = t31a - t30a; + int t31 = t31a + t30a; + + t5a = ((t6 - t5) * 11585 + (1 << 13)) >> 14; + t6a = ((t6 + t5) * 11585 + (1 << 13)) >> 14; + t9a = (t14 * 6270 - t9 * 15137 + (1 << 13)) >> 14; + t14a = (t14 * 15137 + t9 * 6270 + (1 << 13)) >> 14; + t10a = (-(t13 * 15137 + t10 * 6270) + (1 << 13)) >> 14; + t13a = (t13 * 6270 - t10 * 15137 + (1 << 13)) >> 14; + t17a = (t30 * 3196 - t17 * 16069 + (1 << 13)) >> 14; + t30a = (t30 * 16069 + t17 * 3196 + (1 << 13)) >> 14; + t18a = (-(t29 * 16069 + t18 * 3196) + (1 << 13)) >> 14; + t29a = (t29 * 3196 - t18 * 16069 + (1 << 13)) >> 14; + t21a = (t26 * 13623 - t21 * 9102 + (1 << 13)) >> 14; + t26a = (t26 * 9102 + t21 * 13623 + (1 << 13)) >> 14; + t22a = (-(t25 * 9102 + t22 * 13623) + (1 << 13)) >> 14; + t25a = (t25 * 13623 - t22 * 9102 + (1 << 13)) >> 14; + + t0a = t0 + t7; + t1a = t1 + t6a; + t2a = t2 + t5a; + t3a = t3 + t4; + t4a = t3 - t4; + t5 = t2 - t5a; + t6 = t1 - t6a; + t7a = t0 - t7; + t8a = t8 + t11; + t9 = t9a + t10a; + t10 = t9a - t10a; + t11a = t8 - t11; + t12a = t15 - t12; + t13 = t14a - t13a; + t14 = t14a + t13a; + t15a = t15 + t12; + t16a = t16 + t19; + t17 = t17a + t18a; + t18 = t17a - t18a; + t19a = t16 - t19; + t20a = t23 - t20; + t21 = t22a - t21a; + t22 = t22a + t21a; + t23a = t23 + t20; + t24a = t24 + t27; + t25 = t25a + t26a; + t26 = t25a - t26a; + t27a = t24 - t27; + t28a = t31 - t28; + t29 = t30a - t29a; + t30 = t30a + t29a; + t31a = t31 + t28; + + t10a = ((t13 - t10) * 11585 + (1 << 13)) >> 14; + t13a = ((t13 + t10) * 11585 + (1 << 13)) >> 14; + t11 = ((t12a - t11a) * 11585 + (1 << 13)) >> 14; + t12 = ((t12a + t11a) * 11585 + (1 << 13)) >> 14; + t18a = (t29 * 6270 - t18 * 15137 + (1 << 13)) >> 14; + t29a = (t29 * 15137 + t18 * 6270 + (1 << 13)) >> 14; + t19 = (t28a * 6270 - t19a * 15137 + (1 << 13)) >> 14; + t28 = (t28a * 15137 + t19a * 6270 + (1 << 13)) >> 14; + t20 = (-(t27a * 15137 + t20a * 6270) + (1 << 13)) >> 14; + t27 = (t27a * 6270 - t20a * 15137 + (1 << 13)) >> 14; + t21a = (-(t26 * 15137 + t21 * 6270) + (1 << 13)) >> 14; + t26a = (t26 * 6270 - t21 * 15137 + (1 << 13)) >> 14; + + t0 = t0a + t15a; + t1 = t1a + t14; + t2 = t2a + t13a; + t3 = t3a + t12; + t4 = t4a + t11; + t5a = t5 + t10a; + t6a = t6 + t9; + t7 = t7a + t8a; + t8 = t7a - t8a; + t9a = t6 - t9; + t10 = t5 - t10a; + t11a = t4a - t11; + t12a = t3a - t12; + t13 = t2a - t13a; + t14a = t1a - t14; + t15 = t0a - t15a; + t16 = t16a + t23a; + t17a = t17 + t22; + t18 = t18a + t21a; + t19a = t19 + t20; + t20a = t19 - t20; + t21 = t18a - t21a; + t22a = t17 - t22; + t23 = t16a - t23a; + t24 = t31a - t24a; + t25a = t30 - t25; + t26 = t29a - t26a; + t27a = t28 - t27; + t28a = t28 + t27; + t29 = t29a + t26a; + t30a = t30 + t25; + t31 = t31a + t24a; + + t20 = ((t27a - t20a) * 11585 + (1 << 13)) >> 14; + t27 = ((t27a + t20a) * 11585 + (1 << 13)) >> 14; + t21a = ((t26 - t21) * 11585 + (1 << 13)) >> 14; + t26a = ((t26 + t21) * 11585 + (1 << 13)) >> 14; + t22 = ((t25a - t22a) * 11585 + (1 << 13)) >> 14; + t25 = ((t25a + t22a) * 11585 + (1 << 13)) >> 14; + t23a = ((t24 - t23) * 11585 + (1 << 13)) >> 14; + t24a = ((t24 + t23) * 11585 + (1 << 13)) >> 14; + + out[0] = t0 + t31; + out[1] = t1 + t30a; + out[2] = t2 + t29; + out[3] = t3 + t28a; + out[4] = t4 + t27; + out[5] = t5a + t26a; + out[6] = t6a + t25; + out[7] = t7 + t24a; + out[8] = t8 + t23a; + out[9] = t9a + t22; + out[10] = t10 + t21a; + out[11] = t11a + t20; + out[12] = t12a + t19a; + out[13] = t13 + t18; + out[14] = t14a + t17a; + out[15] = t15 + t16; + out[16] = t15 - t16; + out[17] = t14a - t17a; + out[18] = t13 - t18; + out[19] = t12a - t19a; + out[20] = t11a - t20; + out[21] = t10 - t21a; + out[22] = t9a - t22; + out[23] = t8 - t23a; + out[24] = t7 - t24a; + out[25] = t6a - t25; + out[26] = t5a - t26a; + out[27] = t4 - t27; + out[28] = t3 - t28a; + out[29] = t2 - t29; + out[30] = t1 - t30a; + out[31] = t0 - t31; +} + +itxfm_wrapper(idct, idct, 32, 6) + +static av_always_inline void iwht4_1d(int16_t *out, const int16_t *in, + ptrdiff_t stride, int pass) +{ + int t0, t1, t2, t3, t4; + + if (pass == 0) { + t0 = IN(0) >> 2; + t1 = IN(3) >> 2; + t2 = IN(1) >> 2; + t3 = IN(2) >> 2; + } else { + t0 = IN(0); + t1 = IN(3); + t2 = IN(1); + t3 = IN(2); + } + + t0 += t2; + t3 -= t1; + t4 = (t0 - t3) >> 1; + t1 = t4 - t1; + t2 = t4 - t2; + t0 -= t1; + t3 += t2; + + out[0] = t0; + out[1] = t1; + out[2] = t2; + out[3] = t3; +} + +itxfm_wrapper(iwht, iwht, 4, 0) + +#undef IN +#undef itxfm_wrapper +#undef itxfm_wrap + +static av_cold void vp9dsp_itxfm_init(VP9DSPContext *dsp) +{ +#define init_itxfm(tx, sz) \ + dsp->itxfm_add[tx][DCT_DCT] = idct_idct_ ## sz ## _add_c; \ + dsp->itxfm_add[tx][DCT_ADST] = iadst_idct_ ## sz ## _add_c; \ + dsp->itxfm_add[tx][ADST_DCT] = idct_iadst_ ## sz ## _add_c; \ + dsp->itxfm_add[tx][ADST_ADST] = iadst_iadst_ ## sz ## _add_c + +#define init_idct(tx, nm) \ + dsp->itxfm_add[tx][DCT_DCT] = \ + dsp->itxfm_add[tx][ADST_DCT] = \ + dsp->itxfm_add[tx][DCT_ADST] = \ + dsp->itxfm_add[tx][ADST_ADST] = nm ## _add_c + + init_itxfm(TX_4X4, 4x4); + init_itxfm(TX_8X8, 8x8); + init_itxfm(TX_16X16, 16x16); + init_idct(TX_32X32, idct_idct_32x32); + init_idct(4 /* lossless */, iwht_iwht_4x4); + +#undef init_itxfm +#undef init_idct +} + +static av_always_inline void loop_filter(uint8_t *dst, ptrdiff_t stride, + int E, int I, int H, + ptrdiff_t stridea, ptrdiff_t strideb, + int wd) +{ + int i; + + for (i = 0; i < 8; i++, dst += stridea) { + int p7, p6, p5, p4; + int p3 = dst[strideb * -4], p2 = dst[strideb * -3]; + int p1 = dst[strideb * -2], p0 = dst[strideb * -1]; + int q0 = dst[strideb * +0], q1 = dst[strideb * +1]; + int q2 = dst[strideb * +2], q3 = dst[strideb * +3]; + int q4, q5, q6, q7; + int fm = FFABS(p3 - p2) <= I && FFABS(p2 - p1) <= I && + FFABS(p1 - p0) <= I && FFABS(q1 - q0) <= I && + FFABS(q2 - q1) <= I && FFABS(q3 - q2) <= I && + FFABS(p0 - q0) * 2 + (FFABS(p1 - q1) >> 1) <= E; + int flat8out, flat8in; + + if (!fm) + continue; + + if (wd >= 16) { + p7 = dst[strideb * -8]; + p6 = dst[strideb * -7]; + p5 = dst[strideb * -6]; + p4 = dst[strideb * -5]; + q4 = dst[strideb * +4]; + q5 = dst[strideb * +5]; + q6 = dst[strideb * +6]; + q7 = dst[strideb * +7]; + + flat8out = FFABS(p7 - p0) <= 1 && FFABS(p6 - p0) <= 1 && + FFABS(p5 - p0) <= 1 && FFABS(p4 - p0) <= 1 && + FFABS(q4 - q0) <= 1 && FFABS(q5 - q0) <= 1 && + FFABS(q6 - q0) <= 1 && FFABS(q7 - q0) <= 1; + } + + if (wd >= 8) + flat8in = FFABS(p3 - p0) <= 1 && FFABS(p2 - p0) <= 1 && + FFABS(p1 - p0) <= 1 && FFABS(q1 - q0) <= 1 && + FFABS(q2 - q0) <= 1 && FFABS(q3 - q0) <= 1; + + if (wd >= 16 && flat8out && flat8in) { + dst[strideb * -7] = (p7 + p7 + p7 + p7 + p7 + p7 + p7 + p6 * 2 + + p5 + p4 + p3 + p2 + p1 + p0 + q0 + 8) >> 4; + dst[strideb * -6] = (p7 + p7 + p7 + p7 + p7 + p7 + p6 + p5 * 2 + + p4 + p3 + p2 + p1 + p0 + q0 + q1 + 8) >> 4; + dst[strideb * -5] = (p7 + p7 + p7 + p7 + p7 + p6 + p5 + p4 * 2 + + p3 + p2 + p1 + p0 + q0 + q1 + q2 + 8) >> 4; + dst[strideb * -4] = (p7 + p7 + p7 + p7 + p6 + p5 + p4 + p3 * 2 + + p2 + p1 + p0 + q0 + q1 + q2 + q3 + 8) >> 4; + dst[strideb * -3] = (p7 + p7 + p7 + p6 + p5 + p4 + p3 + p2 * 2 + + p1 + p0 + q0 + q1 + q2 + q3 + q4 + 8) >> 4; + dst[strideb * -2] = (p7 + p7 + p6 + p5 + p4 + p3 + p2 + p1 * 2 + + p0 + q0 + q1 + q2 + q3 + q4 + q5 + 8) >> 4; + dst[strideb * -1] = (p7 + p6 + p5 + p4 + p3 + p2 + p1 + p0 * 2 + + q0 + q1 + q2 + q3 + q4 + q5 + q6 + 8) >> 4; + dst[strideb * +0] = (p6 + p5 + p4 + p3 + p2 + p1 + p0 + q0 * 2 + + q1 + q2 + q3 + q4 + q5 + q6 + q7 + 8) >> 4; + dst[strideb * +1] = (p5 + p4 + p3 + p2 + p1 + p0 + q0 + q1 * 2 + + q2 + q3 + q4 + q5 + q6 + q7 + q7 + 8) >> 4; + dst[strideb * +2] = (p4 + p3 + p2 + p1 + p0 + q0 + q1 + q2 * 2 + + q3 + q4 + q5 + q6 + q7 + q7 + q7 + 8) >> 4; + dst[strideb * +3] = (p3 + p2 + p1 + p0 + q0 + q1 + q2 + q3 * 2 + + q4 + q5 + q6 + q7 + q7 + q7 + q7 + 8) >> 4; + dst[strideb * +4] = (p2 + p1 + p0 + q0 + q1 + q2 + q3 + q4 * 2 + + q5 + q6 + q7 + q7 + q7 + q7 + q7 + 8) >> 4; + dst[strideb * +5] = (p1 + p0 + q0 + q1 + q2 + q3 + q4 + q5 * 2 + + q6 + q7 + q7 + q7 + q7 + q7 + q7 + 8) >> 4; + dst[strideb * +6] = (p0 + q0 + q1 + q2 + q3 + q4 + q5 + q6 * 2 + + q7 + q7 + q7 + q7 + q7 + q7 + q7 + 8) >> 4; + } else if (wd >= 8 && flat8in) { + dst[strideb * -3] = (p3 + p3 + p3 + 2 * p2 + p1 + p0 + q0 + 4) >> 3; + dst[strideb * -2] = (p3 + p3 + p2 + 2 * p1 + p0 + q0 + q1 + 4) >> 3; + dst[strideb * -1] = (p3 + p2 + p1 + 2 * p0 + q0 + q1 + q2 + 4) >> 3; + dst[strideb * +0] = (p2 + p1 + p0 + 2 * q0 + q1 + q2 + q3 + 4) >> 3; + dst[strideb * +1] = (p1 + p0 + q0 + 2 * q1 + q2 + q3 + q3 + 4) >> 3; + dst[strideb * +2] = (p0 + q0 + q1 + 2 * q2 + q3 + q3 + q3 + 4) >> 3; + } else { + int hev = FFABS(p1 - p0) > H || FFABS(q1 - q0) > H; + + if (hev) { + int f = av_clip_int8(3 * (q0 - p0) + av_clip_int8(p1 - q1)); + int f1 = FFMIN(f + 4, 127) >> 3; + int f2 = FFMIN(f + 3, 127) >> 3; + + dst[strideb * -1] = av_clip_uint8(p0 + f2); + dst[strideb * +0] = av_clip_uint8(q0 - f1); + } else { + int f = av_clip_int8(3 * (q0 - p0)); + int f1 = FFMIN(f + 4, 127) >> 3; + int f2 = FFMIN(f + 3, 127) >> 3; + + dst[strideb * -1] = av_clip_uint8(p0 + f2); + dst[strideb * +0] = av_clip_uint8(q0 - f1); + + f = (f1 + 1) >> 1; + dst[strideb * -2] = av_clip_uint8(p1 + f); + dst[strideb * +1] = av_clip_uint8(q1 - f); + } + } + } +} + +#define lf_8_fn(dir, wd, stridea, strideb) \ +static void loop_filter_ ## dir ## _ ## wd ## _8_c(uint8_t *dst, \ + ptrdiff_t stride, \ + int E, int I, int H) \ +{ \ + loop_filter(dst, stride, E, I, H, stridea, strideb, wd); \ +} + +#define lf_8_fns(wd) \ + lf_8_fn(h, wd, stride, 1) \ + lf_8_fn(v, wd, 1, stride) + +lf_8_fns(4) +lf_8_fns(8) +lf_8_fns(16) + +#undef lf_8_fn +#undef lf_8_fns + +#define lf_16_fn(dir, stridea) \ +static void loop_filter_ ## dir ## _16_16_c(uint8_t *dst, \ + ptrdiff_t stride, \ + int E, int I, int H) \ +{ \ + loop_filter_ ## dir ## _16_8_c(dst, stride, E, I, H); \ + loop_filter_ ## dir ## _16_8_c(dst + 8 * stridea, stride, E, I, H); \ +} + +lf_16_fn(h, stride) +lf_16_fn(v, 1) + +#undef lf_16_fn + +#define lf_mix_fn(dir, wd1, wd2, stridea) \ +static void loop_filter_ ## dir ## _ ## wd1 ## wd2 ## _16_c(uint8_t *dst, \ + ptrdiff_t stride, \ + int E, int I, \ + int H) \ +{ \ + loop_filter_ ## dir ## _ ## wd1 ## _8_c(dst, stride, E & 0xff, \ + I & 0xff, H & 0xff); \ + loop_filter_ ## dir ## _ ## wd2 ## _8_c(dst + 8 * stridea, stride, \ + E >> 8, I >> 8, H >> 8); \ +} + +#define lf_mix_fns(wd1, wd2) \ + lf_mix_fn(h, wd1, wd2, stride) \ + lf_mix_fn(v, wd1, wd2, 1) + +lf_mix_fns(4, 4) +lf_mix_fns(4, 8) +lf_mix_fns(8, 4) +lf_mix_fns(8, 8) + +#undef lf_mix_fn +#undef lf_mix_fns + +static av_cold void vp9dsp_loopfilter_init(VP9DSPContext *dsp) +{ + dsp->loop_filter_8[0][0] = loop_filter_h_4_8_c; + dsp->loop_filter_8[0][1] = loop_filter_v_4_8_c; + dsp->loop_filter_8[1][0] = loop_filter_h_8_8_c; + dsp->loop_filter_8[1][1] = loop_filter_v_8_8_c; + dsp->loop_filter_8[2][0] = loop_filter_h_16_8_c; + dsp->loop_filter_8[2][1] = loop_filter_v_16_8_c; + + dsp->loop_filter_16[0] = loop_filter_h_16_16_c; + dsp->loop_filter_16[1] = loop_filter_v_16_16_c; + + dsp->loop_filter_mix2[0][0][0] = loop_filter_h_44_16_c; + dsp->loop_filter_mix2[0][0][1] = loop_filter_v_44_16_c; + dsp->loop_filter_mix2[0][1][0] = loop_filter_h_48_16_c; + dsp->loop_filter_mix2[0][1][1] = loop_filter_v_48_16_c; + dsp->loop_filter_mix2[1][0][0] = loop_filter_h_84_16_c; + dsp->loop_filter_mix2[1][0][1] = loop_filter_v_84_16_c; + dsp->loop_filter_mix2[1][1][0] = loop_filter_h_88_16_c; + dsp->loop_filter_mix2[1][1][1] = loop_filter_v_88_16_c; +} + +static av_always_inline void copy_c(uint8_t *dst, const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h) +{ + do { + memcpy(dst, src, w); + + dst += dst_stride; + src += src_stride; + } while (--h); +} + +static av_always_inline void avg_c(uint8_t *dst, const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h) +{ + do { + int x; + + for (x = 0; x < w; x += 4) + AV_WN32A(&dst[x], rnd_avg32(AV_RN32A(&dst[x]), AV_RN32(&src[x]))); + + dst += dst_stride; + src += src_stride; + } while (--h); +} + +#define fpel_fn(type, sz) \ +static void type ## sz ## _c(uint8_t *dst, const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + type ## _c(dst, src, dst_stride, src_stride, sz, h); \ +} + +#define copy_avg_fn(sz) \ + fpel_fn(copy, sz) \ + fpel_fn(avg, sz) + +copy_avg_fn(64) +copy_avg_fn(32) +copy_avg_fn(16) +copy_avg_fn(8) +copy_avg_fn(4) + +#undef fpel_fn +#undef copy_avg_fn + +static const int8_t vp9_subpel_filters[3][15][8] = { + [FILTER_8TAP_REGULAR] = { + { 0, 1, -5, 126, 8, -3, 1, 0 }, + { -1, 3, -10, 122, 18, -6, 2, 0 }, + { -1, 4, -13, 118, 27, -9, 3, -1 }, + { -1, 4, -16, 112, 37, -11, 4, -1 }, + { -1, 5, -18, 105, 48, -14, 4, -1 }, + { -1, 5, -19, 97, 58, -16, 5, -1 }, + { -1, 6, -19, 88, 68, -18, 5, -1 }, + { -1, 6, -19, 78, 78, -19, 6, -1 }, + { -1, 5, -18, 68, 88, -19, 6, -1 }, + { -1, 5, -16, 58, 97, -19, 5, -1 }, + { -1, 4, -14, 48, 105, -18, 5, -1 }, + { -1, 4, -11, 37, 112, -16, 4, -1 }, + { -1, 3, -9, 27, 118, -13, 4, -1 }, + { 0, 2, -6, 18, 122, -10, 3, -1 }, + { 0, 1, -3, 8, 126, -5, 1, 0 }, + }, [FILTER_8TAP_SHARP] = { + { -1, 3, -7, 127, 8, -3, 1, 0 }, + { -2, 5, -13, 125, 17, -6, 3, -1 }, + { -3, 7, -17, 121, 27, -10, 5, -2 }, + { -4, 9, -20, 115, 37, -13, 6, -2 }, + { -4, 10, -23, 108, 48, -16, 8, -3 }, + { -4, 10, -24, 100, 59, -19, 9, -3 }, + { -4, 11, -24, 90, 70, -21, 10, -4 }, + { -4, 11, -23, 80, 80, -23, 11, -4 }, + { -4, 10, -21, 70, 90, -24, 11, -4 }, + { -3, 9, -19, 59, 100, -24, 10, -4 }, + { -3, 8, -16, 48, 108, -23, 10, -4 }, + { -2, 6, -13, 37, 115, -20, 9, -4 }, + { -2, 5, -10, 27, 121, -17, 7, -3 }, + { -1, 3, -6, 17, 125, -13, 5, -2 }, + { 0, 1, -3, 8, 127, -7, 3, -1 }, + }, [FILTER_8TAP_SMOOTH] = { + { -3, -1, 32, 64, 38, 1, -3, 0 }, + { -2, -2, 29, 63, 41, 2, -3, 0 }, + { -2, -2, 26, 63, 43, 4, -4, 0 }, + { -2, -3, 24, 62, 46, 5, -4, 0 }, + { -2, -3, 21, 60, 49, 7, -4, 0 }, + { -1, -4, 18, 59, 51, 9, -4, 0 }, + { -1, -4, 16, 57, 53, 12, -4, -1 }, + { -1, -4, 14, 55, 55, 14, -4, -1 }, + { -1, -4, 12, 53, 57, 16, -4, -1 }, + { 0, -4, 9, 51, 59, 18, -4, -1 }, + { 0, -4, 7, 49, 60, 21, -3, -2 }, + { 0, -4, 5, 46, 62, 24, -3, -2 }, + { 0, -4, 4, 43, 63, 26, -2, -2 }, + { 0, -3, 2, 41, 63, 29, -2, -2 }, + { 0, -3, 1, 38, 64, 32, -1, -3 }, + } +}; + +#define FILTER_8TAP(src, x, F, stride) \ + av_clip_uint8((F[0] * src[x + -3 * stride] + \ + F[1] * src[x + -2 * stride] + \ + F[2] * src[x + -1 * stride] + \ + F[3] * src[x + +0 * stride] + \ + F[4] * src[x + +1 * stride] + \ + F[5] * src[x + +2 * stride] + \ + F[6] * src[x + +3 * stride] + \ + F[7] * src[x + +4 * stride] + 64) >> 7) + +static av_always_inline void do_8tap_1d_c(uint8_t *dst, const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h, ptrdiff_t ds, + const int8_t *filter, int avg) +{ + do { + int x; + + for (x = 0; x < w; x++) + if (avg) + dst[x] = (dst[x] + FILTER_8TAP(src, x, filter, ds) + 1) >> 1; + else + dst[x] = FILTER_8TAP(src, x, filter, ds); + + dst += dst_stride; + src += src_stride; + } while (--h); +} + +#define filter_8tap_1d_fn(opn, opa, dir, ds) \ +static av_noinline void opn ## _8tap_1d_ ## dir ## _c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int w, int h, \ + const int8_t *filter) \ +{ \ + do_8tap_1d_c(dst, src, dst_stride, src_stride, w, h, ds, filter, opa); \ +} + +filter_8tap_1d_fn(put, 0, v, src_stride) +filter_8tap_1d_fn(put, 0, h, 1) +filter_8tap_1d_fn(avg, 1, v, src_stride) +filter_8tap_1d_fn(avg, 1, h, 1) + +#undef filter_8tap_1d_fn + +static av_always_inline void do_8tap_2d_c(uint8_t *dst, const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h, const int8_t *filterx, + const int8_t *filtery, int avg) +{ + int tmp_h = h + 7; + uint8_t tmp[64 * 71], *tmp_ptr = tmp; + + src -= src_stride * 3; + do { + int x; + + for (x = 0; x < w; x++) + tmp_ptr[x] = FILTER_8TAP(src, x, filterx, 1); + + tmp_ptr += 64; + src += src_stride; + } while (--tmp_h); + + tmp_ptr = tmp + 64 * 3; + do { + int x; + + for (x = 0; x < w; x++) + if (avg) + dst[x] = (dst[x] + FILTER_8TAP(tmp_ptr, x, filtery, 64) + 1) >> 1; + else + dst[x] = FILTER_8TAP(tmp_ptr, x, filtery, 64); + + tmp_ptr += 64; + dst += dst_stride; + } while (--h); +} + +#define filter_8tap_2d_fn(opn, opa) \ +static av_noinline void opn ## _8tap_2d_hv_c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int w, int h, \ + const int8_t *filterx, \ + const int8_t *filtery) \ +{ \ + do_8tap_2d_c(dst, src, dst_stride, src_stride, \ + w, h, filterx, filtery, opa); \ +} + +filter_8tap_2d_fn(put, 0) +filter_8tap_2d_fn(avg, 1) + +#undef filter_8tap_2d_fn + +#undef FILTER_8TAP + +#define filter_fn_1d(sz, dir, dir_m, type, type_idx, avg) \ +static void \ +avg ## _8tap_ ## type ## _ ## sz ## dir ## _c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + avg ## _8tap_1d_ ## dir ## _c(dst, src, dst_stride, src_stride, sz, h, \ + vp9_subpel_filters[type_idx][dir_m - 1]); \ +} + +#define filter_fn_2d(sz, type, type_idx, avg) \ +static void avg ## _8tap_ ## type ## _ ## sz ## hv_c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + avg ## _8tap_2d_hv_c(dst, src, dst_stride, src_stride, sz, h, \ + vp9_subpel_filters[type_idx][mx - 1], \ + vp9_subpel_filters[type_idx][my - 1]); \ +} + +#define FILTER_BILIN(src, x, mxy, stride) \ + (src[x] + ((mxy * (src[x + stride] - src[x]) + 8) >> 4)) + +static av_always_inline void do_bilin_1d_c(uint8_t *dst, + const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h, ptrdiff_t ds, + int mxy, int avg) +{ + do { + int x; + + for (x = 0; x < w; x++) + if (avg) + dst[x] = (dst[x] + FILTER_BILIN(src, x, mxy, ds) + 1) >> 1; + else + dst[x] = FILTER_BILIN(src, x, mxy, ds); + + dst += dst_stride; + src += src_stride; + } while (--h); +} + +#define bilin_1d_fn(opn, opa, dir, ds) \ +static av_noinline void opn ## _bilin_1d_ ## dir ## _c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int w, int h, int mxy) \ +{ \ + do_bilin_1d_c(dst, src, dst_stride, src_stride, w, h, ds, mxy, opa); \ +} + +bilin_1d_fn(put, 0, v, src_stride) +bilin_1d_fn(put, 0, h, 1) +bilin_1d_fn(avg, 1, v, src_stride) +bilin_1d_fn(avg, 1, h, 1) + +#undef bilin_1d_fn + +static av_always_inline void do_bilin_2d_c(uint8_t *dst, + const uint8_t *src, + ptrdiff_t dst_stride, + ptrdiff_t src_stride, + int w, int h, int mx, int my, + int avg) +{ + uint8_t tmp[64 * 65], *tmp_ptr = tmp; + int tmp_h = h + 1; + + do { + int x; + + for (x = 0; x < w; x++) + tmp_ptr[x] = FILTER_BILIN(src, x, mx, 1); + + tmp_ptr += 64; + src += src_stride; + } while (--tmp_h); + + tmp_ptr = tmp; + do { + int x; + + for (x = 0; x < w; x++) + if (avg) + dst[x] = (dst[x] + FILTER_BILIN(tmp_ptr, x, my, 64) + 1) >> 1; + else + dst[x] = FILTER_BILIN(tmp_ptr, x, my, 64); + + tmp_ptr += 64; + dst += dst_stride; + } while (--h); +} + +#define bilin_2d_fn(opn, opa) \ +static av_noinline void opn ## _bilin_2d_hv_c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int w, int h, \ + int mx, int my) \ +{ \ + do_bilin_2d_c(dst, src, dst_stride, src_stride, w, h, mx, my, opa); \ +} + +bilin_2d_fn(put, 0) +bilin_2d_fn(avg, 1) + +#undef bilin_2d_fn + +#undef FILTER_BILIN + +#define bilinf_fn_1d(sz, dir, dir_m, avg) \ +static void avg ## _bilin_ ## sz ## dir ## _c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + avg ## _bilin_1d_ ## dir ## _c(dst, src, dst_stride, src_stride, \ + sz, h, dir_m); \ +} + +#define bilinf_fn_2d(sz, avg) \ +static void avg ## _bilin_ ## sz ## hv_c(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + avg ## _bilin_2d_hv_c(dst, src, dst_stride, src_stride, \ + sz, h, mx, my); \ +} + +#define filter_fn(sz, avg) \ + filter_fn_1d(sz, h, mx, regular, FILTER_8TAP_REGULAR, avg) \ + filter_fn_1d(sz, v, my, regular, FILTER_8TAP_REGULAR, avg) \ + filter_fn_2d(sz, regular, FILTER_8TAP_REGULAR, avg) \ + filter_fn_1d(sz, h, mx, smooth, FILTER_8TAP_SMOOTH, avg) \ + filter_fn_1d(sz, v, my, smooth, FILTER_8TAP_SMOOTH, avg) \ + filter_fn_2d(sz, smooth, FILTER_8TAP_SMOOTH, avg) \ + filter_fn_1d(sz, h, mx, sharp, FILTER_8TAP_SHARP, avg) \ + filter_fn_1d(sz, v, my, sharp, FILTER_8TAP_SHARP, avg) \ + filter_fn_2d(sz, sharp, FILTER_8TAP_SHARP, avg) \ + bilinf_fn_1d(sz, h, mx, avg) \ + bilinf_fn_1d(sz, v, my, avg) \ + bilinf_fn_2d(sz, avg) + +#define filter_fn_set(avg) \ + filter_fn(64, avg) \ + filter_fn(32, avg) \ + filter_fn(16, avg) \ + filter_fn(8, avg) \ + filter_fn(4, avg) + +filter_fn_set(put) +filter_fn_set(avg) + +#undef filter_fn +#undef filter_fn_set +#undef filter_fn_1d +#undef filter_fn_2d +#undef bilinf_fn_1d +#undef bilinf_fn_2d + +static av_cold void vp9dsp_mc_init(VP9DSPContext *dsp) +{ +#define init_fpel(idx1, idx2, sz, type) \ + dsp->mc[idx1][FILTER_8TAP_SMOOTH][idx2][0][0] = type ## sz ## _c; \ + dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = type ## sz ## _c; \ + dsp->mc[idx1][FILTER_8TAP_SHARP][idx2][0][0] = type ## sz ## _c; \ + dsp->mc[idx1][FILTER_BILINEAR][idx2][0][0] = type ## sz ## _c + +#define init_copy_avg(idx, sz) \ + init_fpel(idx, 0, sz, copy); \ + init_fpel(idx, 1, sz, avg) + + init_copy_avg(0, 64); + init_copy_avg(1, 32); + init_copy_avg(2, 16); + init_copy_avg(3, 8); + init_copy_avg(4, 4); + +#undef init_copy_avg +#undef init_fpel + +#define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type) \ + dsp->mc[idx1][FILTER_8TAP_SMOOTH][idx2][idxh][idxv] = type ## _8tap_smooth_ ## sz ## dir ## _c; \ + dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = type ## _8tap_regular_ ## sz ## dir ## _c; \ + dsp->mc[idx1][FILTER_8TAP_SHARP][idx2][idxh][idxv] = type ## _8tap_sharp_ ## sz ## dir ## _c; \ + dsp->mc[idx1][FILTER_BILINEAR][idx2][idxh][idxv] = type ## _bilin_ ## sz ## dir ## _c + +#define init_subpel2(idx, idxh, idxv, dir, type) \ + init_subpel1(0, idx, idxh, idxv, 64, dir, type); \ + init_subpel1(1, idx, idxh, idxv, 32, dir, type); \ + init_subpel1(2, idx, idxh, idxv, 16, dir, type); \ + init_subpel1(3, idx, idxh, idxv, 8, dir, type); \ + init_subpel1(4, idx, idxh, idxv, 4, dir, type) + +#define init_subpel3(idx, type) \ + init_subpel2(idx, 1, 1, hv, type); \ + init_subpel2(idx, 0, 1, v, type); \ + init_subpel2(idx, 1, 0, h, type) + + init_subpel3(0, put); + init_subpel3(1, avg); + +#undef init_subpel1 +#undef init_subpel2 +#undef init_subpel3 +} + +av_cold void ff_vp9dsp_init(VP9DSPContext *dsp) +{ + vp9dsp_intrapred_init(dsp); + vp9dsp_itxfm_init(dsp); + vp9dsp_loopfilter_init(dsp); + vp9dsp_mc_init(dsp); + + if (ARCH_X86) + ff_vp9dsp_init_x86(dsp); +} diff --git a/libavcodec/vp9mvs.c b/libavcodec/vp9mvs.c new file mode 100644 index 0000000000..2f37755046 --- /dev/null +++ b/libavcodec/vp9mvs.c @@ -0,0 +1,344 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "internal.h" +#include "vp56.h" +#include "vp9.h" +#include "vp9data.h" + +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 *const b = &s->b; + int row = b->row, col = b->col, row7 = b->row7; + const int8_t (*p)[2] = mv_ref_blk_off[b->bs]; +#define INVALID_MV 0x80008000U + uint32_t mem = 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; \ + clamp_mv(&tmp, &mv, s); \ + m = AV_RN32A(&tmp); \ + if (!idx) { \ + AV_WN32A(pmv, m); \ + return; \ + } else if (mem == INVALID_MV) { \ + mem = m; \ + } else if (m != mem) { \ + AV_WN32A(pmv, m); \ + 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) { + VP9MVRefPair *mv = &s->mv[0][(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->tiling.tile_col_start) { + VP9MVRefPair *mv = &s->mv[0][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 the neighborhood, using same reference frame + for (; i < 8; i++) { + int c = p[i][0] + col, r = p[i][1] + row; + + if (c >= s->tiling.tile_col_start && c < s->cols && + r >= 0 && r < s->rows) { + VP9MVRefPair *mv = &s->mv[0][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->use_last_frame_mvs) { + VP9MVRefPair *mv = &s->mv[1][row * s->sb_cols * 8 + col]; + + 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 the neighborhood, using different reference frame + for (i = 0; i < 8; i++) { + int c = p[i][0] + col, r = p[i][1] + row; + + if (c >= s->tiling.tile_col_start && c < s->cols && + r >= 0 && r < s->rows) { + VP9MVRefPair *mv = &s->mv[0][r * s->sb_cols * 8 + c]; + + if (mv->ref[0] != ref && mv->ref[0] >= 0) + RETURN_SCALE_MV(mv->mv[0], + s->signbias[mv->ref[0]] != s->signbias[ref]); + if (mv->ref[1] != ref && mv->ref[1] >= 0) + RETURN_SCALE_MV(mv->mv[1], + s->signbias[mv->ref[1]] != s->signbias[ref]); + } + } + + // MV at this position in previous frame, using different reference frame + if (s->use_last_frame_mvs) { + VP9MVRefPair *mv = &s->mv[1][row * s->sb_cols * 8 + col]; + + if (mv->ref[0] != ref && mv->ref[0] >= 0) + RETURN_SCALE_MV(mv->mv[0], + s->signbias[mv->ref[0]] != s->signbias[ref]); + if (mv->ref[1] != ref && mv->ref[1] >= 0) + RETURN_SCALE_MV(mv->mv[1], + s->signbias[mv->ref[1]] != s->signbias[ref]); + } + + AV_ZERO32(pmv); +#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, ff_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, ff_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, ff_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); +} + +void ff_vp9_fill_mv(VP9Context *s, VP56mv *mv, int mode, int sb) +{ + VP9Block *const b = &s->b; + + if (mode == ZEROMV) { + memset(mv, 0, sizeof(*mv) * 2); + } 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->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, ff_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->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, ff_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); + } + } + } +} diff --git a/libavcodec/vp9prob.c b/libavcodec/vp9prob.c new file mode 100644 index 0000000000..b8a7c22af4 --- /dev/null +++ b/libavcodec/vp9prob.c @@ -0,0 +1,274 @@ +/* + * VP9 compatible video decoder + * + * Copyright (C) 2013 Ronald S. Bultje + * Copyright (C) 2013 Clément Bœsch + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "vp56.h" +#include "vp9.h" +#include "vp9data.h" + +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; + + p1 = *p; + p2 = ((ct0 << 8) + (ct >> 1)) / ct; + p2 = av_clip(p2, 1, 255); + ct = FFMIN(ct, max_count); + update_factor = FASTDIV(update_factor * ct, max_count); + + // (p1 * (256 - update_factor) + p2 * update_factor + 128) >> 8 + *p = p1 + (((p2 - p1) * update_factor + 128) >> 8); +} + +void ff_vp9_adapt_probs(VP9Context *s) +{ + int i, j, k, l, m; + ProbContext *p = &s->prob_ctx[s->framectxid].p; + int uf = (s->keyframe || s->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->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->keyframe || s->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->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->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->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->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->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->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); + } +} diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index 0fe1c1af5e..6f4935bc3e 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -35,6 +35,7 @@ OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp_init.o OBJS-$(CONFIG_VP3DSP) += x86/vp3dsp_init.o OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp_init.o OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp_init.o +OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o OBJS-$(CONFIG_XMM_CLOBBER_TEST) += x86/w64xmmtest.o MMX-OBJS-$(CONFIG_DSPUTIL) += x86/dsputil_mmx.o \ @@ -90,3 +91,4 @@ YASM-OBJS-$(CONFIG_VP3DSP) += x86/vp3dsp.o YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp.o YASM-OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp.o \ x86/vp8dsp_loopfilter.o +YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp.o diff --git a/libavcodec/x86/vp9dsp.asm b/libavcodec/x86/vp9dsp.asm new file mode 100644 index 0000000000..6488f3092d --- /dev/null +++ b/libavcodec/x86/vp9dsp.asm @@ -0,0 +1,277 @@ +;****************************************************************************** +;* VP9 SIMD optimizations +;* +;* Copyright (c) 2013 Ronald S. Bultje +;* +;* This file is part of Libav. +;* +;* Libav is free software; you can redistribute it and/or +;* modify it under the terms of the GNU Lesser General Public +;* License as published by the Free Software Foundation; either +;* version 2.1 of the License, or (at your option) any later version. +;* +;* Libav is distributed in the hope that it will be useful, +;* but WITHOUT ANY WARRANTY; without even the implied warranty of +;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +;* Lesser General Public License for more details. +;* +;* You should have received a copy of the GNU Lesser General Public +;* License along with Libav; if not, write to the Free Software +;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +;****************************************************************************** + +%include "libavutil/x86/x86util.asm" + +SECTION_RODATA + +; FIXME share with vp8dsp.asm +pw_256: times 8 dw 256 + +%macro F8_TAPS 8 +times 8 db %1, %2 +times 8 db %3, %4 +times 8 db %5, %6 +times 8 db %7, %8 +%endmacro +; int8_t ff_filters_ssse3[3][15][4][16] +const filters_ssse3 ; smooth + F8_TAPS -3, -1, 32, 64, 38, 1, -3, 0 + F8_TAPS -2, -2, 29, 63, 41, 2, -3, 0 + F8_TAPS -2, -2, 26, 63, 43, 4, -4, 0 + F8_TAPS -2, -3, 24, 62, 46, 5, -4, 0 + F8_TAPS -2, -3, 21, 60, 49, 7, -4, 0 + F8_TAPS -1, -4, 18, 59, 51, 9, -4, 0 + F8_TAPS -1, -4, 16, 57, 53, 12, -4, -1 + F8_TAPS -1, -4, 14, 55, 55, 14, -4, -1 + F8_TAPS -1, -4, 12, 53, 57, 16, -4, -1 + F8_TAPS 0, -4, 9, 51, 59, 18, -4, -1 + F8_TAPS 0, -4, 7, 49, 60, 21, -3, -2 + F8_TAPS 0, -4, 5, 46, 62, 24, -3, -2 + F8_TAPS 0, -4, 4, 43, 63, 26, -2, -2 + F8_TAPS 0, -3, 2, 41, 63, 29, -2, -2 + F8_TAPS 0, -3, 1, 38, 64, 32, -1, -3 + ; regular + F8_TAPS 0, 1, -5, 126, 8, -3, 1, 0 + F8_TAPS -1, 3, -10, 122, 18, -6, 2, 0 + F8_TAPS -1, 4, -13, 118, 27, -9, 3, -1 + F8_TAPS -1, 4, -16, 112, 37, -11, 4, -1 + F8_TAPS -1, 5, -18, 105, 48, -14, 4, -1 + F8_TAPS -1, 5, -19, 97, 58, -16, 5, -1 + F8_TAPS -1, 6, -19, 88, 68, -18, 5, -1 + F8_TAPS -1, 6, -19, 78, 78, -19, 6, -1 + F8_TAPS -1, 5, -18, 68, 88, -19, 6, -1 + F8_TAPS -1, 5, -16, 58, 97, -19, 5, -1 + F8_TAPS -1, 4, -14, 48, 105, -18, 5, -1 + F8_TAPS -1, 4, -11, 37, 112, -16, 4, -1 + F8_TAPS -1, 3, -9, 27, 118, -13, 4, -1 + F8_TAPS 0, 2, -6, 18, 122, -10, 3, -1 + F8_TAPS 0, 1, -3, 8, 126, -5, 1, 0 + ; sharp + F8_TAPS -1, 3, -7, 127, 8, -3, 1, 0 + F8_TAPS -2, 5, -13, 125, 17, -6, 3, -1 + F8_TAPS -3, 7, -17, 121, 27, -10, 5, -2 + F8_TAPS -4, 9, -20, 115, 37, -13, 6, -2 + F8_TAPS -4, 10, -23, 108, 48, -16, 8, -3 + F8_TAPS -4, 10, -24, 100, 59, -19, 9, -3 + F8_TAPS -4, 11, -24, 90, 70, -21, 10, -4 + F8_TAPS -4, 11, -23, 80, 80, -23, 11, -4 + F8_TAPS -4, 10, -21, 70, 90, -24, 11, -4 + F8_TAPS -3, 9, -19, 59, 100, -24, 10, -4 + F8_TAPS -3, 8, -16, 48, 108, -23, 10, -4 + F8_TAPS -2, 6, -13, 37, 115, -20, 9, -4 + F8_TAPS -2, 5, -10, 27, 121, -17, 7, -3 + F8_TAPS -1, 3, -6, 17, 125, -13, 5, -2 + F8_TAPS 0, 1, -3, 8, 127, -7, 3, -1 + +SECTION .text + +%macro filter_h_fn 1 +%assign %%px mmsize/2 +cglobal %1_8tap_1d_h_ %+ %%px, 6, 6, 11, dst, src, dstride, sstride, h, filtery + mova m6, [pw_256] + mova m7, [filteryq+ 0] +%if ARCH_X86_64 && mmsize > 8 + mova m8, [filteryq+16] + mova m9, [filteryq+32] + mova m10, [filteryq+48] +%endif +.loop: + movh m0, [srcq-3] + movh m1, [srcq-2] + movh m2, [srcq-1] + movh m3, [srcq+0] + movh m4, [srcq+1] + movh m5, [srcq+2] + punpcklbw m0, m1 + punpcklbw m2, m3 + movh m1, [srcq+3] + movh m3, [srcq+4] + add srcq, sstrideq + punpcklbw m4, m5 + punpcklbw m1, m3 + pmaddubsw m0, m7 +%if ARCH_X86_64 && mmsize > 8 + pmaddubsw m2, m8 + pmaddubsw m4, m9 + pmaddubsw m1, m10 +%else + pmaddubsw m2, [filteryq+16] + pmaddubsw m4, [filteryq+32] + pmaddubsw m1, [filteryq+48] +%endif + paddw m0, m2 + paddw m4, m1 + paddsw m0, m4 + pmulhrsw m0, m6 +%ifidn %1, avg + movh m1, [dstq] +%endif + packuswb m0, m0 +%ifidn %1, avg + pavgb m0, m1 +%endif + movh [dstq], m0 + add dstq, dstrideq + dec hd + jg .loop + RET +%endmacro + +INIT_MMX ssse3 +filter_h_fn put +filter_h_fn avg + +INIT_XMM ssse3 +filter_h_fn put +filter_h_fn avg + +%macro filter_v_fn 1 +%assign %%px mmsize/2 +%if ARCH_X86_64 +cglobal %1_8tap_1d_v_ %+ %%px, 6, 8, 11, dst, src, dstride, sstride, h, filtery, src4, sstride3 +%else +cglobal %1_8tap_1d_v_ %+ %%px, 4, 7, 11, dst, src, dstride, sstride, filtery, src4, sstride3 + mov filteryq, r5mp +%define hd r4mp +%endif + sub srcq, sstrideq + lea sstride3q, [sstrideq*3] + sub srcq, sstrideq + mova m6, [pw_256] + sub srcq, sstrideq + mova m7, [filteryq+ 0] + lea src4q, [srcq+sstrideq*4] +%if ARCH_X86_64 && mmsize > 8 + mova m8, [filteryq+16] + mova m9, [filteryq+32] + mova m10, [filteryq+48] +%endif +.loop: + ; FIXME maybe reuse loads from previous rows, or just more generally + ; unroll this to prevent multiple loads of the same data? + movh m0, [srcq] + movh m1, [srcq+sstrideq] + movh m2, [srcq+sstrideq*2] + movh m3, [srcq+sstride3q] + movh m4, [src4q] + movh m5, [src4q+sstrideq] + punpcklbw m0, m1 + punpcklbw m2, m3 + movh m1, [src4q+sstrideq*2] + movh m3, [src4q+sstride3q] + add srcq, sstrideq + add src4q, sstrideq + punpcklbw m4, m5 + punpcklbw m1, m3 + pmaddubsw m0, m7 +%if ARCH_X86_64 && mmsize > 8 + pmaddubsw m2, m8 + pmaddubsw m4, m9 + pmaddubsw m1, m10 +%else + pmaddubsw m2, [filteryq+16] + pmaddubsw m4, [filteryq+32] + pmaddubsw m1, [filteryq+48] +%endif + paddw m0, m2 + paddw m4, m1 + paddsw m0, m4 + pmulhrsw m0, m6 +%ifidn %1, avg + movh m1, [dstq] +%endif + packuswb m0, m0 +%ifidn %1, avg + pavgb m0, m1 +%endif + movh [dstq], m0 + add dstq, dstrideq + dec hd + jg .loop + RET +%endmacro + +INIT_MMX ssse3 +filter_v_fn put +filter_v_fn avg + +INIT_XMM ssse3 +filter_v_fn put +filter_v_fn avg + +%macro fpel_fn 6 +%if %2 == 4 +%define %%srcfn movh +%define %%dstfn movh +%else +%define %%srcfn movu +%define %%dstfn mova +%endif + +%if %2 <= 16 +cglobal %1%2, 5, 7, 4, dst, src, dstride, sstride, h, dstride3, sstride3 + lea sstride3q, [sstrideq*3] + lea dstride3q, [dstrideq*3] +%else +cglobal %1%2, 5, 5, 4, dst, src, dstride, sstride, h +%endif +.loop: + %%srcfn m0, [srcq] + %%srcfn m1, [srcq+s%3] + %%srcfn m2, [srcq+s%4] + %%srcfn m3, [srcq+s%5] + lea srcq, [srcq+sstrideq*%6] +%ifidn %1, avg + pavgb m0, [dstq] + pavgb m1, [dstq+d%3] + pavgb m2, [dstq+d%4] + pavgb m3, [dstq+d%5] +%endif + %%dstfn [dstq], m0 + %%dstfn [dstq+d%3], m1 + %%dstfn [dstq+d%4], m2 + %%dstfn [dstq+d%5], m3 + lea dstq, [dstq+dstrideq*%6] + sub hd, %6 + jnz .loop + RET +%endmacro + +%define d16 16 +%define s16 16 +INIT_MMX mmx +fpel_fn put, 4, strideq, strideq*2, stride3q, 4 +fpel_fn put, 8, strideq, strideq*2, stride3q, 4 +INIT_MMX sse +fpel_fn avg, 4, strideq, strideq*2, stride3q, 4 +fpel_fn avg, 8, strideq, strideq*2, stride3q, 4 +INIT_XMM sse +fpel_fn put, 16, strideq, strideq*2, stride3q, 4 +fpel_fn put, 32, mmsize, strideq, strideq+mmsize, 2 +fpel_fn put, 64, mmsize, mmsize*2, mmsize*3, 1 +INIT_XMM sse2 +fpel_fn avg, 16, strideq, strideq*2, stride3q, 4 +fpel_fn avg, 32, mmsize, strideq, strideq+mmsize, 2 +fpel_fn avg, 64, mmsize, mmsize*2, mmsize*3, 1 +%undef s16 +%undef d16 diff --git a/libavcodec/x86/vp9dsp_init.c b/libavcodec/x86/vp9dsp_init.c new file mode 100644 index 0000000000..540dc3882f --- /dev/null +++ b/libavcodec/x86/vp9dsp_init.c @@ -0,0 +1,245 @@ +/* + * VP9 SIMD optimizations + * + * Copyright (c) 2013 Ronald S. Bultje + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/cpu.h" +#include "libavutil/internal.h" +#include "libavutil/mem.h" +#include "libavutil/x86/asm.h" +#include "libavutil/x86/cpu.h" +#include "libavcodec/vp9.h" + +#if HAVE_YASM + +#define fpel_func(avg, sz, opt) \ +void ff_ ## avg ## sz ## _ ## opt(uint8_t *dst, const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) + +fpel_func(put, 4, mmx); +fpel_func(put, 8, mmx); +fpel_func(put, 16, sse); +fpel_func(put, 32, sse); +fpel_func(put, 64, sse); +fpel_func(avg, 4, sse); +fpel_func(avg, 8, sse); +fpel_func(avg, 16, sse2); +fpel_func(avg, 32, sse2); +fpel_func(avg, 64, sse2); +#undef fpel_func + +#define mc_func(avg, sz, dir, opt) \ +void \ +ff_ ## avg ## _8tap_1d_ ## dir ## _ ## sz ## _ ## opt(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, \ + const int8_t (*filter)[16]) + +#define mc_funcs(sz) \ + mc_func(put, sz, h, ssse3); \ + mc_func(avg, sz, h, ssse3); \ + mc_func(put, sz, v, ssse3); \ + mc_func(avg, sz, v, ssse3) + +mc_funcs(4); +mc_funcs(8); + +#undef mc_funcs +#undef mc_func + +#define mc_rep_func(avg, sz, hsz, dir, opt) \ +static av_always_inline void \ +ff_ ## avg ## _8tap_1d_ ## dir ## _ ## sz ## _ ## opt(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, \ + const int8_t (*filter)[16]) \ +{ \ + ff_ ## avg ## _8tap_1d_ ## dir ## _ ## hsz ## _ ## opt(dst, src, \ + dst_stride, \ + src_stride, \ + h, \ + filter); \ + ff_ ## avg ## _8tap_1d_ ## dir ## _ ## hsz ## _ ## opt(dst + hsz, \ + src + hsz, \ + dst_stride, \ + src_stride, \ + h, filter); \ +} + +#define mc_rep_funcs(sz, hsz) \ + mc_rep_func(put, sz, hsz, h, ssse3); \ + mc_rep_func(avg, sz, hsz, h, ssse3); \ + mc_rep_func(put, sz, hsz, v, ssse3); \ + mc_rep_func(avg, sz, hsz, v, ssse3) + +mc_rep_funcs(16, 8); +mc_rep_funcs(32, 16); +mc_rep_funcs(64, 32); + +#undef mc_rep_funcs +#undef mc_rep_func + +extern const int8_t ff_filters_ssse3[3][15][4][16]; + +#define filter_8tap_2d_fn(op, sz, f, fname) \ +static void \ +op ## _8tap_ ## fname ## _ ## sz ## hv_ssse3(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, int my) \ +{ \ + LOCAL_ALIGNED_16(uint8_t, temp, [71 * 64]); \ + ff_put_8tap_1d_h_ ## sz ## _ssse3(temp, src - 3 * src_stride, \ + 64, src_stride, \ + h + 7, \ + ff_filters_ssse3[f][mx - 1]); \ + ff_ ## op ## _8tap_1d_v_ ## sz ## _ssse3(dst, temp + 3 * 64, \ + dst_stride, 64, \ + h, \ + ff_filters_ssse3[f][my - 1]); \ +} + +#define filters_8tap_2d_fn(op, sz) \ + filter_8tap_2d_fn(op, sz, FILTER_8TAP_REGULAR, regular) \ + filter_8tap_2d_fn(op, sz, FILTER_8TAP_SHARP, sharp) \ + filter_8tap_2d_fn(op, sz, FILTER_8TAP_SMOOTH, smooth) + +#define filters_8tap_2d_fn2(op) \ + filters_8tap_2d_fn(op, 64) \ + filters_8tap_2d_fn(op, 32) \ + filters_8tap_2d_fn(op, 16) \ + filters_8tap_2d_fn(op, 8) \ + filters_8tap_2d_fn(op, 4) + +filters_8tap_2d_fn2(put) +filters_8tap_2d_fn2(avg) + +#undef filters_8tap_2d_fn2 +#undef filters_8tap_2d_fn +#undef filter_8tap_2d_fn + +#define filter_8tap_1d_fn(op, sz, f, fname, dir, dvar) \ +static void \ +op ## _8tap_ ## fname ## _ ## sz ## dir ## _ssse3(uint8_t *dst, \ + const uint8_t *src, \ + ptrdiff_t dst_stride, \ + ptrdiff_t src_stride, \ + int h, int mx, \ + int my) \ +{ \ + ff_ ## op ## _8tap_1d_ ## dir ## _ ## sz ## _ssse3(dst, src, \ + dst_stride, \ + src_stride, h, \ + ff_filters_ssse3[f][dvar - 1]); \ +} + +#define filters_8tap_1d_fn(op, sz, dir, dvar) \ + filter_8tap_1d_fn(op, sz, FILTER_8TAP_REGULAR, regular, dir, dvar) \ + filter_8tap_1d_fn(op, sz, FILTER_8TAP_SHARP, sharp, dir, dvar) \ + filter_8tap_1d_fn(op, sz, FILTER_8TAP_SMOOTH, smooth, dir, dvar) + +#define filters_8tap_1d_fn2(op, sz) \ + filters_8tap_1d_fn(op, sz, h, mx) \ + filters_8tap_1d_fn(op, sz, v, my) + +#define filters_8tap_1d_fn3(op) \ + filters_8tap_1d_fn2(op, 64) \ + filters_8tap_1d_fn2(op, 32) \ + filters_8tap_1d_fn2(op, 16) \ + filters_8tap_1d_fn2(op, 8) \ + filters_8tap_1d_fn2(op, 4) + +filters_8tap_1d_fn3(put) +filters_8tap_1d_fn3(avg) + +#undef filters_8tap_1d_fn +#undef filters_8tap_1d_fn2 +#undef filters_8tap_1d_fn3 +#undef filter_8tap_1d_fn + +#endif /* HAVE_YASM */ + +av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp) +{ +#if HAVE_YASM + int cpu_flags = av_get_cpu_flags(); + +#define init_fpel(idx1, idx2, sz, type, opt) \ + dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \ + dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \ + dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][0][0] = \ + dsp->mc[idx1][FILTER_BILINEAR ][idx2][0][0] = ff_ ## type ## sz ## _ ## opt + + +#define init_subpel1(idx1, idx2, idxh, idxv, sz, dir, type, opt) \ + dsp->mc[idx1][FILTER_8TAP_SMOOTH][idx2][idxh][idxv] = type ## _8tap_smooth_ ## sz ## dir ## _ ## opt; \ + dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][idxh][idxv] = type ## _8tap_regular_ ## sz ## dir ## _ ## opt; \ + dsp->mc[idx1][FILTER_8TAP_SHARP][idx2][idxh][idxv] = type ## _8tap_sharp_ ## sz ## dir ## _ ## opt + +#define init_subpel2(idx, idxh, idxv, dir, type, opt) \ + init_subpel1(0, idx, idxh, idxv, 64, dir, type, opt); \ + init_subpel1(1, idx, idxh, idxv, 32, dir, type, opt); \ + init_subpel1(2, idx, idxh, idxv, 16, dir, type, opt); \ + init_subpel1(3, idx, idxh, idxv, 8, dir, type, opt); \ + init_subpel1(4, idx, idxh, idxv, 4, dir, type, opt) + +#define init_subpel3(idx, type, opt) \ + init_subpel2(idx, 1, 1, hv, type, opt); \ + init_subpel2(idx, 0, 1, v, type, opt); \ + init_subpel2(idx, 1, 0, h, type, opt) + + if (EXTERNAL_MMX(cpu_flags)) { + init_fpel(4, 0, 4, put, mmx); + init_fpel(3, 0, 8, put, mmx); + } + + if (EXTERNAL_SSE(cpu_flags)) { + init_fpel(2, 0, 16, put, sse); + init_fpel(1, 0, 32, put, sse); + init_fpel(0, 0, 64, put, sse); + init_fpel(4, 1, 4, avg, sse); + init_fpel(3, 1, 8, avg, sse); + } + + if (EXTERNAL_SSE2(cpu_flags)) { + init_fpel(2, 1, 16, avg, sse2); + init_fpel(1, 1, 32, avg, sse2); + init_fpel(0, 1, 64, avg, sse2); + } + + if (EXTERNAL_SSSE3(cpu_flags)) { + init_subpel3(0, put, ssse3); + init_subpel3(1, avg, ssse3); + } + +#undef init_fpel +#undef init_subpel1 +#undef init_subpel2 +#undef init_subpel3 + +#endif /* HAVE_YASM */ +} -- cgit v1.2.3