diff options
Diffstat (limited to 'libavcodec/utvideodec.c')
-rw-r--r-- | libavcodec/utvideodec.c | 704 |
1 files changed, 296 insertions, 408 deletions
diff --git a/libavcodec/utvideodec.c b/libavcodec/utvideodec.c index 26b991c..c5f5534 100644 --- a/libavcodec/utvideodec.c +++ b/libavcodec/utvideodec.c @@ -2,20 +2,20 @@ * Ut Video decoder * Copyright (c) 2011 Konstantin Shishkov * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * Libav is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with Libav; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -27,16 +27,16 @@ #include <inttypes.h> #include <stdlib.h> -#include "libavutil/intreadwrite.h" +#define UNCHECKED_BITSTREAM_READER 1 +#include "libavutil/intreadwrite.h" #include "avcodec.h" -#include "bitstream.h" #include "bswapdsp.h" #include "bytestream.h" +#include "get_bits.h" #include "internal.h" #include "thread.h" #include "utvideo.h" -#include "vlc.h" static int build_huff10(const uint8_t *src, VLC *vlc, int *fsym) { @@ -75,8 +75,8 @@ static int build_huff10(const uint8_t *src, VLC *vlc, int *fsym) syms[i] = he[i].sym; code += 0x80000000u >> (he[i].len - 1); } - - return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 11), last + 1, +#define VLC_BITS 11 + return ff_init_vlc_sparse(vlc, VLC_BITS, last + 1, bits, sizeof(*bits), sizeof(*bits), codes, sizeof(*codes), sizeof(*codes), syms, sizeof(*syms), sizeof(*syms), 0); @@ -103,13 +103,14 @@ static int build_huff(const uint8_t *src, VLC *vlc, int *fsym) *fsym = he[0].sym; return 0; } - if (he[0].len > 32) - return -1; last = 255; while (he[last].len == 255 && last) last--; + if (he[last].len > 32) + return -1; + code = 1; for (i = last; i >= 0; i--) { codes[i] = code >> (32 - he[i].len); @@ -118,22 +119,22 @@ static int build_huff(const uint8_t *src, VLC *vlc, int *fsym) code += 0x80000000u >> (he[i].len - 1); } - return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 9), last + 1, + return ff_init_vlc_sparse(vlc, VLC_BITS, last + 1, bits, sizeof(*bits), sizeof(*bits), codes, sizeof(*codes), sizeof(*codes), syms, sizeof(*syms), sizeof(*syms), 0); } static int decode_plane10(UtvideoContext *c, int plane_no, - uint16_t *dst, int step, int stride, + uint16_t *dst, ptrdiff_t stride, int width, int height, const uint8_t *src, const uint8_t *huff, int use_pred) { - BitstreamContext bc; int i, j, slice, pix, ret; int sstart, send; VLC vlc; + GetBitContext gb; int prev, fsym; if ((ret = build_huff10(huff, &vlc, &fsym)) < 0) { @@ -151,7 +152,7 @@ static int decode_plane10(UtvideoContext *c, int plane_no, prev = 0x200; for (j = sstart; j < send; j++) { - for (i = 0; i < width * step; i += step) { + for (i = 0; i < width; i++) { pix = fsym; if (use_pred) { prev += pix; @@ -186,23 +187,16 @@ static int decode_plane10(UtvideoContext *c, int plane_no, goto fail; } - memcpy(c->slice_bits, src + slice_data_start + c->slices * 4, - slice_size); memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE); c->bdsp.bswap_buf((uint32_t *) c->slice_bits, - (uint32_t *) c->slice_bits, + (uint32_t *)(src + slice_data_start + c->slices * 4), (slice_data_end - slice_data_start + 3) >> 2); - bitstream_init8(&bc, c->slice_bits, slice_size); + init_get_bits(&gb, c->slice_bits, slice_size * 8); prev = 0x200; for (j = sstart; j < send; j++) { - for (i = 0; i < width * step; i += step) { - if (bitstream_bits_left(&bc) <= 0) { - av_log(c->avctx, AV_LOG_ERROR, - "Slice decoding ran out of bits\n"); - goto fail; - } - pix = bitstream_read_vlc(&bc, vlc.table, vlc.bits, 3); + for (i = 0; i < width; i++) { + pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3); if (pix < 0) { av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n"); goto fail; @@ -215,10 +209,15 @@ static int decode_plane10(UtvideoContext *c, int plane_no, dest[i] = pix; } dest += stride; + if (get_bits_left(&gb) < 0) { + av_log(c->avctx, AV_LOG_ERROR, + "Slice decoding ran out of bits\n"); + goto fail; + } } - if (bitstream_bits_left(&bc) > 32) + if (get_bits_left(&gb) > 32) av_log(c->avctx, AV_LOG_WARNING, - "%d bits left after decoding slice\n", bitstream_bits_left(&bc)); + "%d bits left after decoding slice\n", get_bits_left(&gb)); } ff_free_vlc(&vlc); @@ -229,7 +228,7 @@ fail: return AVERROR_INVALIDDATA; } -static int compute_cmask(int plane_no, int interlaced, int pix_fmt) +static int compute_cmask(int plane_no, int interlaced, enum AVPixelFormat pix_fmt) { const int is_luma = (pix_fmt == AV_PIX_FMT_YUV420P) && !plane_no; @@ -240,17 +239,64 @@ static int compute_cmask(int plane_no, int interlaced, int pix_fmt) } static int decode_plane(UtvideoContext *c, int plane_no, - uint8_t *dst, int step, ptrdiff_t stride, + uint8_t *dst, ptrdiff_t stride, int width, int height, const uint8_t *src, int use_pred) { int i, j, slice, pix; int sstart, send; VLC vlc; - BitstreamContext bc; - int prev, fsym; + GetBitContext gb; + int ret, prev, fsym; const int cmask = compute_cmask(plane_no, c->interlaced, c->avctx->pix_fmt); + if (c->pack) { + send = 0; + for (slice = 0; slice < c->slices; slice++) { + GetBitContext cbit, pbit; + uint8_t *dest, *p; + + ret = init_get_bits8(&cbit, c->control_stream[plane_no][slice], c->control_stream_size[plane_no][slice]); + if (ret < 0) + return ret; + + ret = init_get_bits8(&pbit, c->packed_stream[plane_no][slice], c->packed_stream_size[plane_no][slice]); + if (ret < 0) + return ret; + + sstart = send; + send = (height * (slice + 1) / c->slices) & cmask; + dest = dst + sstart * stride; + + if (3 * ((dst + send * stride - dest + 7)/8) > get_bits_left(&cbit)) + return AVERROR_INVALIDDATA; + + for (p = dest; p < dst + send * stride; p += 8) { + int bits = get_bits_le(&cbit, 3); + + if (bits == 0) { + *(uint64_t *) p = 0; + } else { + uint32_t sub = 0x80 >> (8 - (bits + 1)), add; + int k; + + if ((bits + 1) * 8 > get_bits_left(&pbit)) + return AVERROR_INVALIDDATA; + + for (k = 0; k < 8; k++) { + + p[k] = get_bits_le(&pbit, bits + 1); + add = (~p[k] & sub) << (8 - bits); + p[k] -= sub; + p[k] += add; + } + } + } + } + + return 0; + } + if (build_huff(src, &vlc, &fsym)) { av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n"); return AVERROR_INVALIDDATA; @@ -266,7 +312,7 @@ static int decode_plane(UtvideoContext *c, int plane_no, prev = 0x80; for (j = sstart; j < send; j++) { - for (i = 0; i < width * step; i += step) { + for (i = 0; i < width; i++) { pix = fsym; if (use_pred) { prev += pix; @@ -302,23 +348,16 @@ static int decode_plane(UtvideoContext *c, int plane_no, goto fail; } - memcpy(c->slice_bits, src + slice_data_start + c->slices * 4, - slice_size); memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE); c->bdsp.bswap_buf((uint32_t *) c->slice_bits, - (uint32_t *) c->slice_bits, + (uint32_t *)(src + slice_data_start + c->slices * 4), (slice_data_end - slice_data_start + 3) >> 2); - bitstream_init8(&bc, c->slice_bits, slice_size); + init_get_bits(&gb, c->slice_bits, slice_size * 8); prev = 0x80; for (j = sstart; j < send; j++) { - for (i = 0; i < width * step; i += step) { - if (bitstream_bits_left(&bc) <= 0) { - av_log(c->avctx, AV_LOG_ERROR, - "Slice decoding ran out of bits\n"); - goto fail; - } - pix = bitstream_read_vlc(&bc, vlc.table, vlc.bits, 4); + for (i = 0; i < width; i++) { + pix = get_vlc2(&gb, vlc.table, VLC_BITS, 3); if (pix < 0) { av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n"); goto fail; @@ -329,11 +368,16 @@ static int decode_plane(UtvideoContext *c, int plane_no, } dest[i] = pix; } + if (get_bits_left(&gb) < 0) { + av_log(c->avctx, AV_LOG_ERROR, + "Slice decoding ran out of bits\n"); + goto fail; + } dest += stride; } - if (bitstream_bits_left(&bc) > 32) + if (get_bits_left(&gb) > 32) av_log(c->avctx, AV_LOG_WARNING, - "%d bits left after decoding slice\n", bitstream_bits_left(&bc)); + "%d bits left after decoding slice\n", get_bits_left(&gb)); } ff_free_vlc(&vlc); @@ -344,49 +388,12 @@ fail: return AVERROR_INVALIDDATA; } -static void restore_rgb_planes(uint8_t *src, int step, ptrdiff_t stride, - int width, int height) -{ - int i, j; - uint8_t r, g, b; - - for (j = 0; j < height; j++) { - for (i = 0; i < width * step; i += step) { - r = src[i]; - g = src[i + 1]; - b = src[i + 2]; - src[i] = r + g - 0x80; - src[i + 2] = b + g - 0x80; - } - src += stride; - } -} +#undef A +#undef B +#undef C -static void restore_rgb_planes10(AVFrame *frame, int width, int height) -{ - uint16_t *src_r = (uint16_t *)frame->data[2]; - uint16_t *src_g = (uint16_t *)frame->data[0]; - uint16_t *src_b = (uint16_t *)frame->data[1]; - int r, g, b; - int i, j; - - for (j = 0; j < height; j++) { - for (i = 0; i < width; i++) { - r = src_r[i]; - g = src_g[i]; - b = src_b[i]; - src_r[i] = (r + g - 0x200) & 0x3FF; - src_b[i] = (b + g - 0x200) & 0x3FF; - } - src_r += frame->linesize[2] / 2; - src_g += frame->linesize[0] / 2; - src_b += frame->linesize[1] / 2; - } -} - -static void restore_median_planar(UtvideoContext *c, uint8_t *src, - ptrdiff_t stride, int width, int height, - int slices, int rmode) +static void restore_median_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride, + int width, int height, int slices, int rmode) { int i, j, slice; int A, B, C; @@ -405,7 +412,7 @@ static void restore_median_planar(UtvideoContext *c, uint8_t *src, // first line - left neighbour prediction bsrc[0] += 0x80; - c->hdspdec.add_hfyu_left_pred(bsrc, bsrc, width, 0); + c->llviddsp.add_left_pred(bsrc, bsrc, width, 0); bsrc += stride; if (slice_height <= 1) continue; @@ -413,16 +420,20 @@ static void restore_median_planar(UtvideoContext *c, uint8_t *src, C = bsrc[-stride]; bsrc[0] += C; A = bsrc[0]; - for (i = 1; i < width; i++) { + for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */ B = bsrc[i - stride]; bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); C = B; A = bsrc[i]; } + if (width > 16) + c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride + 16, + bsrc + 16, width - 16, &A, &B); + bsrc += stride; // the rest of lines use continuous median prediction for (j = 2; j < slice_height; j++) { - c->hdspdec.add_hfyu_median_pred(bsrc, bsrc - stride, + c->llviddsp.add_median_pred(bsrc, bsrc - stride, bsrc, width, &A, &B); bsrc += stride; } @@ -433,16 +444,15 @@ static void restore_median_planar(UtvideoContext *c, uint8_t *src, * so restoring function should take care of possible padding between * two parts of the same "line". */ -static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, - ptrdiff_t stride, int width, int height, - int slices, int rmode) +static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_t stride, + int width, int height, int slices, int rmode) { int i, j, slice; int A, B, C; uint8_t *bsrc; int slice_start, slice_height; const int cmask = ~(rmode ? 3 : 1); - const int stride2 = stride << 1; + const ptrdiff_t stride2 = stride << 1; for (slice = 0; slice < slices; slice++) { slice_start = ((slice * height) / slices) & cmask; @@ -456,8 +466,8 @@ static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, // first line - left neighbour prediction bsrc[0] += 0x80; - A = c->hdspdec.add_hfyu_left_pred(bsrc, bsrc, width, 0); - c->hdspdec.add_hfyu_left_pred(bsrc + stride, bsrc + stride, width, A); + A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0); + c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A); bsrc += stride2; if (slice_height <= 1) continue; @@ -465,154 +475,30 @@ static void restore_median_planar_il(UtvideoContext *c, uint8_t *src, C = bsrc[-stride2]; bsrc[0] += C; A = bsrc[0]; - for (i = 1; i < width; i++) { + for (i = 1; i < FFMIN(width, 16); i++) { /* scalar loop (DSP need align 16) */ B = bsrc[i - stride2]; bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); C = B; A = bsrc[i]; } - c->hdspdec.add_hfyu_median_pred(bsrc + stride, bsrc - stride, + if (width > 16) + c->llviddsp.add_median_pred(bsrc + 16, bsrc - stride2 + 16, + bsrc + 16, width - 16, &A, &B); + + c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride, bsrc + stride, width, &A, &B); bsrc += stride2; // the rest of lines use continuous median prediction for (j = 2; j < slice_height; j++) { - c->hdspdec.add_hfyu_median_pred(bsrc, bsrc - stride2, + c->llviddsp.add_median_pred(bsrc, bsrc - stride2, bsrc, width, &A, &B); - c->hdspdec.add_hfyu_median_pred(bsrc + stride, bsrc - stride, + c->llviddsp.add_median_pred(bsrc + stride, bsrc - stride, bsrc + stride, width, &A, &B); bsrc += stride2; } } } -static void restore_median_packed(uint8_t *src, int step, ptrdiff_t stride, - int width, int height, - int slices, int rmode) -{ - int i, j, slice; - int A, B, C; - uint8_t *bsrc; - int slice_start, slice_height; - const int cmask = ~rmode; - - for (slice = 0; slice < slices; slice++) { - slice_start = ((slice * height) / slices) & cmask; - slice_height = ((((slice + 1) * height) / slices) & cmask) - - slice_start; - if (!slice_height) - continue; - - bsrc = src + slice_start * stride; - - // first line - left neighbour prediction - bsrc[0] += 0x80; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - bsrc[i] += A; - A = bsrc[i]; - } - bsrc += stride; - if (slice_height == 1) - continue; - // second line - first element has top prediction, the rest uses median - C = bsrc[-stride]; - bsrc[0] += C; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - B = bsrc[i - stride]; - bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[i]; - } - bsrc += stride; - // the rest of lines use continuous median prediction - for (j = 2; j < slice_height; j++) { - for (i = 0; i < width * step; i += step) { - B = bsrc[i - stride]; - bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[i]; - } - bsrc += stride; - } - } -} - -/* UtVideo interlaced mode treats every two lines as a single one, - * so restoring function should take care of possible padding between - * two parts of the same "line". - */ -static void restore_median_packed_il(uint8_t *src, int step, ptrdiff_t stride, - int width, int height, - int slices, int rmode) -{ - int i, j, slice; - int A, B, C; - uint8_t *bsrc; - int slice_start, slice_height; - const int cmask = ~(rmode ? 3 : 1); - const ptrdiff_t stride2 = stride << 1; - - for (slice = 0; slice < slices; slice++) { - slice_start = ((slice * height) / slices) & cmask; - slice_height = ((((slice + 1) * height) / slices) & cmask) - - slice_start; - slice_height >>= 1; - if (!slice_height) - continue; - - bsrc = src + slice_start * stride; - - // first line - left neighbour prediction - bsrc[0] += 0x80; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - bsrc[i] += A; - A = bsrc[i]; - } - for (i = 0; i < width * step; i += step) { - bsrc[stride + i] += A; - A = bsrc[stride + i]; - } - bsrc += stride2; - if (slice_height == 1) - continue; - // second line - first element has top prediction, the rest uses median - C = bsrc[-stride2]; - bsrc[0] += C; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - B = bsrc[i - stride2]; - bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[i]; - } - for (i = 0; i < width * step; i += step) { - B = bsrc[i - stride]; - bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[stride + i]; - } - bsrc += stride2; - // the rest of lines use continuous median prediction - for (j = 2; j < slice_height; j++) { - for (i = 0; i < width * step; i += step) { - B = bsrc[i - stride2]; - bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[i]; - } - for (i = 0; i < width * step; i += step) { - B = bsrc[i - stride]; - bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C)); - C = B; - A = bsrc[i + stride]; - } - bsrc += stride2; - } - } -} - static void restore_gradient_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t stride, int width, int height, int slices, int rmode) { @@ -621,6 +507,7 @@ static void restore_gradient_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t s uint8_t *bsrc; int slice_start, slice_height; const int cmask = ~rmode; + int min_width = FFMIN(width, 32); for (slice = 0; slice < slices; slice++) { slice_start = ((slice * height) / slices) & cmask; @@ -633,19 +520,21 @@ static void restore_gradient_planar(UtvideoContext *c, uint8_t *src, ptrdiff_t s // first line - left neighbour prediction bsrc[0] += 0x80; - c->hdspdec.add_hfyu_left_pred(bsrc, bsrc, width, 0); + c->llviddsp.add_left_pred(bsrc, bsrc, width, 0); bsrc += stride; if (slice_height <= 1) continue; for (j = 1; j < slice_height; j++) { // second line - first element has top prediction, the rest uses gradient bsrc[0] = (bsrc[0] + bsrc[-stride]) & 0xFF; - for (i = 1; i < width; i++) { + for (i = 1; i < min_width; i++) { /* dsp need align 32 */ A = bsrc[i - stride]; B = bsrc[i - (stride + 1)]; C = bsrc[i - 1]; bsrc[i] = (A - B + C + bsrc[i]) & 0xFF; } + if (width > 32) + c->llviddsp.add_gradient_pred(bsrc + 32, stride, width - 32); bsrc += stride; } } @@ -660,6 +549,7 @@ static void restore_gradient_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_ int slice_start, slice_height; const int cmask = ~(rmode ? 3 : 1); const ptrdiff_t stride2 = stride << 1; + int min_width = FFMIN(width, 32); for (slice = 0; slice < slices; slice++) { slice_start = ((slice * height) / slices) & cmask; @@ -673,20 +563,23 @@ static void restore_gradient_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_ // first line - left neighbour prediction bsrc[0] += 0x80; - A = c->hdspdec.add_hfyu_left_pred(bsrc, bsrc, width, 0); - c->hdspdec.add_hfyu_left_pred(bsrc + stride, bsrc + stride, width, A); + A = c->llviddsp.add_left_pred(bsrc, bsrc, width, 0); + c->llviddsp.add_left_pred(bsrc + stride, bsrc + stride, width, A); bsrc += stride2; if (slice_height <= 1) continue; for (j = 1; j < slice_height; j++) { // second line - first element has top prediction, the rest uses gradient bsrc[0] = (bsrc[0] + bsrc[-stride2]) & 0xFF; - for (i = 1; i < width; i++) { + for (i = 1; i < min_width; i++) { /* dsp need align 32 */ A = bsrc[i - stride2]; B = bsrc[i - (stride2 + 1)]; C = bsrc[i - 1]; bsrc[i] = (A - B + C + bsrc[i]) & 0xFF; } + if (width > 32) + c->llviddsp.add_gradient_pred(bsrc + 32, stride2, width - 32); + A = bsrc[-stride]; B = bsrc[-(1 + stride + stride - width)]; C = bsrc[width - 1]; @@ -702,108 +595,6 @@ static void restore_gradient_planar_il(UtvideoContext *c, uint8_t *src, ptrdiff_ } } -static void restore_gradient_packed(uint8_t *src, int step, ptrdiff_t stride, - int width, int height, int slices, int rmode) -{ - int i, j, slice; - int A, B, C; - uint8_t *bsrc; - int slice_start, slice_height; - const int cmask = ~rmode; - - for (slice = 0; slice < slices; slice++) { - slice_start = ((slice * height) / slices) & cmask; - slice_height = ((((slice + 1) * height) / slices) & cmask) - - slice_start; - - if (!slice_height) - continue; - bsrc = src + slice_start * stride; - - // first line - left neighbour prediction - bsrc[0] += 0x80; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - bsrc[i] += A; - A = bsrc[i]; - } - bsrc += stride; - if (slice_height <= 1) - continue; - for (j = 1; j < slice_height; j++) { - // second line - first element has top prediction, the rest uses gradient - C = bsrc[-stride]; - bsrc[0] += C; - for (i = step; i < width * step; i += step) { - A = bsrc[i - stride]; - B = bsrc[i - (stride + step)]; - C = bsrc[i - step]; - bsrc[i] = (A - B + C + bsrc[i]) & 0xFF; - } - bsrc += stride; - } - } -} - -static void restore_gradient_packed_il(uint8_t *src, int step, ptrdiff_t stride, - int width, int height, int slices, int rmode) -{ - int i, j, slice; - int A, B, C; - uint8_t *bsrc; - int slice_start, slice_height; - const int cmask = ~(rmode ? 3 : 1); - const ptrdiff_t stride2 = stride << 1; - - for (slice = 0; slice < slices; slice++) { - slice_start = ((slice * height) / slices) & cmask; - slice_height = ((((slice + 1) * height) / slices) & cmask) - - slice_start; - slice_height >>= 1; - if (!slice_height) - continue; - - bsrc = src + slice_start * stride; - - // first line - left neighbour prediction - bsrc[0] += 0x80; - A = bsrc[0]; - for (i = step; i < width * step; i += step) { - bsrc[i] += A; - A = bsrc[i]; - } - for (i = 0; i < width * step; i += step) { - bsrc[stride + i] += A; - A = bsrc[stride + i]; - } - bsrc += stride2; - if (slice_height <= 1) - continue; - for (j = 1; j < slice_height; j++) { - // second line - first element has top prediction, the rest uses gradient - C = bsrc[-stride2]; - bsrc[0] += C; - for (i = step; i < width * step; i += step) { - A = bsrc[i - stride2]; - B = bsrc[i - (stride2 + step)]; - C = bsrc[i - step]; - bsrc[i] = (A - B + C + bsrc[i]) & 0xFF; - } - A = bsrc[-stride]; - B = bsrc[-(step + stride + stride - width * step)]; - C = bsrc[width * step - step]; - bsrc[stride] = (A - B + C + bsrc[stride]) & 0xFF; - for (i = step; i < width * step; i += step) { - A = bsrc[i - stride]; - B = bsrc[i - (step + stride)]; - C = bsrc[i - step + stride]; - bsrc[i + stride] = (A - B + C + bsrc[i + stride]) & 0xFF; - } - bsrc += stride2; - } - } -} - static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPacket *avpkt) { @@ -817,16 +608,63 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, GetByteContext gb; ThreadFrame frame = { .f = data }; - if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) { - av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); + if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) return ret; - } - - ff_thread_finish_setup(avctx); /* parse plane structure to get frame flags and validate slice offsets */ bytestream2_init(&gb, buf, buf_size); - if (c->pro) { + + if (c->pack) { + const uint8_t *packed_stream; + const uint8_t *control_stream; + GetByteContext pb; + uint32_t nb_cbs; + int left; + + c->frame_info = PRED_GRADIENT << 8; + + if (bytestream2_get_byte(&gb) != 1) + return AVERROR_INVALIDDATA; + bytestream2_skip(&gb, 3); + c->offset = bytestream2_get_le32(&gb); + + if (buf_size <= c->offset + 8LL) + return AVERROR_INVALIDDATA; + + bytestream2_init(&pb, buf + 8 + c->offset, buf_size - 8 - c->offset); + + nb_cbs = bytestream2_get_le32(&pb); + if (nb_cbs > c->offset) + return AVERROR_INVALIDDATA; + + packed_stream = buf + 8; + control_stream = packed_stream + (c->offset - nb_cbs); + left = control_stream - packed_stream; + + for (i = 0; i < c->planes; i++) { + for (j = 0; j < c->slices; j++) { + c->packed_stream[i][j] = packed_stream; + c->packed_stream_size[i][j] = bytestream2_get_le32(&pb); + if (c->packed_stream_size[i][j] > left) + return AVERROR_INVALIDDATA; + left -= c->packed_stream_size[i][j]; + packed_stream += c->packed_stream_size[i][j]; + } + } + + left = buf + buf_size - control_stream; + + for (i = 0; i < c->planes; i++) { + for (j = 0; j < c->slices; j++) { + c->control_stream[i][j] = control_stream; + c->control_stream_size[i][j] = bytestream2_get_le32(&pb); + if (c->control_stream_size[i][j] > left) + return AVERROR_INVALIDDATA; + left -= c->control_stream_size[i][j]; + control_stream += c->control_stream_size[i][j]; + } + } + } else if (c->pro) { if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) { av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n"); return AVERROR_INVALIDDATA; @@ -844,7 +682,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, for (j = 0; j < c->slices; j++) { slice_end = bytestream2_get_le32u(&gb); if (slice_end < 0 || slice_end < slice_start || - bytestream2_get_bytes_left(&gb) < slice_end) { + bytestream2_get_bytes_left(&gb) < slice_end + 1024LL) { av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n"); return AVERROR_INVALIDDATA; } @@ -893,56 +731,60 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, c->frame_pred = (c->frame_info >> 8) & 3; - av_fast_malloc(&c->slice_bits, &c->slice_bits_size, - max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE); + max_slice_size += 4*avctx->width; + + if (!c->pack) { + av_fast_malloc(&c->slice_bits, &c->slice_bits_size, + max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE); - if (!c->slice_bits) { - av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n"); - return AVERROR(ENOMEM); + if (!c->slice_bits) { + av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n"); + return AVERROR(ENOMEM); + } } switch (c->avctx->pix_fmt) { - case AV_PIX_FMT_RGB24: - case AV_PIX_FMT_RGBA: + case AV_PIX_FMT_GBRP: + case AV_PIX_FMT_GBRAP: for (i = 0; i < c->planes; i++) { - ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i], - c->planes, frame.f->linesize[0], avctx->width, + ret = decode_plane(c, i, frame.f->data[i], + frame.f->linesize[i], avctx->width, avctx->height, plane_start[i], c->frame_pred == PRED_LEFT); if (ret) return ret; if (c->frame_pred == PRED_MEDIAN) { if (!c->interlaced) { - restore_median_packed(frame.f->data[0] + ff_ut_rgb_order[i], - c->planes, frame.f->linesize[0], avctx->width, + restore_median_planar(c, frame.f->data[i], + frame.f->linesize[i], avctx->width, avctx->height, c->slices, 0); } else { - restore_median_packed_il(frame.f->data[0] + ff_ut_rgb_order[i], - c->planes, frame.f->linesize[0], + restore_median_planar_il(c, frame.f->data[i], + frame.f->linesize[i], avctx->width, avctx->height, c->slices, 0); } } else if (c->frame_pred == PRED_GRADIENT) { if (!c->interlaced) { - restore_gradient_packed(frame.f->data[0] + ff_ut_rgb_order[i], - c->planes, frame.f->linesize[0], - avctx->width, avctx->height, - c->slices, 0); + restore_gradient_planar(c, frame.f->data[i], + frame.f->linesize[i], avctx->width, + avctx->height, c->slices, 0); } else { - restore_gradient_packed_il(frame.f->data[0] + ff_ut_rgb_order[i], - c->planes, frame.f->linesize[0], - avctx->width, avctx->height, - c->slices, 0); + restore_gradient_planar_il(c, frame.f->data[i], + frame.f->linesize[i], + avctx->width, avctx->height, c->slices, + 0); } } } - restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0], - avctx->width, avctx->height); + c->utdsp.restore_rgb_planes(frame.f->data[2], frame.f->data[0], frame.f->data[1], + frame.f->linesize[2], frame.f->linesize[0], frame.f->linesize[1], + avctx->width, avctx->height); break; case AV_PIX_FMT_GBRAP10: case AV_PIX_FMT_GBRP10: for (i = 0; i < c->planes; i++) { - ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], 1, + ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2, avctx->width, avctx->height, plane_start[i], plane_start[i + 1] - 1024, @@ -950,11 +792,13 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, if (ret) return ret; } - restore_rgb_planes10(frame.f, avctx->width, avctx->height); + c->utdsp.restore_rgb_planes10((uint16_t *)frame.f->data[2], (uint16_t *)frame.f->data[0], (uint16_t *)frame.f->data[1], + frame.f->linesize[2] / 2, frame.f->linesize[0] / 2, frame.f->linesize[1] / 2, + avctx->width, avctx->height); break; case AV_PIX_FMT_YUV420P: for (i = 0; i < 3; i++) { - ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i], + ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i], avctx->width >> !!i, avctx->height >> !!i, plane_start[i], c->frame_pred == PRED_LEFT); if (ret) @@ -973,8 +817,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, } else if (c->frame_pred == PRED_GRADIENT) { if (!c->interlaced) { restore_gradient_planar(c, frame.f->data[i], frame.f->linesize[i], - avctx->width >> !!i, - avctx->height >> !!i, + avctx->width >> !!i, avctx->height >> !!i, c->slices, !i); } else { restore_gradient_planar_il(c, frame.f->data[i], frame.f->linesize[i], @@ -987,7 +830,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, break; case AV_PIX_FMT_YUV422P: for (i = 0; i < 3; i++) { - ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i], + ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i], avctx->width >> !!i, avctx->height, plane_start[i], c->frame_pred == PRED_LEFT); if (ret) @@ -1017,7 +860,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, break; case AV_PIX_FMT_YUV444P: for (i = 0; i < 3; i++) { - ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i], + ret = decode_plane(c, i, frame.f->data[i], frame.f->linesize[i], avctx->width, avctx->height, plane_start[i], c->frame_pred == PRED_LEFT); if (ret) @@ -1047,7 +890,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, break; case AV_PIX_FMT_YUV422P10: for (i = 0; i < 3; i++) { - ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], 1, frame.f->linesize[i] / 2, + ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], frame.f->linesize[i] / 2, avctx->width >> !!i, avctx->height, plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT); if (ret) @@ -1072,50 +915,20 @@ static av_cold int decode_init(AVCodecContext *avctx) c->avctx = avctx; + ff_utvideodsp_init(&c->utdsp); ff_bswapdsp_init(&c->bdsp); - ff_huffyuvdsp_init(&c->hdspdec); - - if (avctx->extradata_size >= 16) { - av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n", - avctx->extradata[3], avctx->extradata[2], - avctx->extradata[1], avctx->extradata[0]); - av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n", - AV_RB32(avctx->extradata + 4)); - c->frame_info_size = AV_RL32(avctx->extradata + 8); - c->flags = AV_RL32(avctx->extradata + 12); - - if (c->frame_info_size != 4) - avpriv_request_sample(avctx, "Frame info not 4 bytes"); - av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags); - c->slices = (c->flags >> 24) + 1; - c->compression = c->flags & 1; - c->interlaced = c->flags & 0x800; - } else if (avctx->extradata_size == 8) { - av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n", - avctx->extradata[3], avctx->extradata[2], - avctx->extradata[1], avctx->extradata[0]); - av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n", - AV_RB32(avctx->extradata + 4)); - c->interlaced = 0; - c->pro = 1; - c->frame_info_size = 4; - } else { - av_log(avctx, AV_LOG_ERROR, - "Insufficient extradata size %d, should be at least 16\n", - avctx->extradata_size); - return AVERROR_INVALIDDATA; - } + ff_llviddsp_init(&c->llviddsp); c->slice_bits_size = 0; switch (avctx->codec_tag) { case MKTAG('U', 'L', 'R', 'G'): c->planes = 3; - avctx->pix_fmt = AV_PIX_FMT_RGB24; + avctx->pix_fmt = AV_PIX_FMT_GBRP; break; case MKTAG('U', 'L', 'R', 'A'): c->planes = 4; - avctx->pix_fmt = AV_PIX_FMT_RGBA; + avctx->pix_fmt = AV_PIX_FMT_GBRAP; break; case MKTAG('U', 'L', 'Y', '0'): c->planes = 3; @@ -1159,12 +972,87 @@ static av_cold int decode_init(AVCodecContext *avctx) avctx->pix_fmt = AV_PIX_FMT_YUV444P; avctx->colorspace = AVCOL_SPC_BT709; break; + case MKTAG('U', 'M', 'Y', '2'): + c->planes = 3; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_YUV422P; + avctx->colorspace = AVCOL_SPC_BT470BG; + break; + case MKTAG('U', 'M', 'H', '2'): + c->planes = 3; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_YUV422P; + avctx->colorspace = AVCOL_SPC_BT709; + break; + case MKTAG('U', 'M', 'Y', '4'): + c->planes = 3; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_YUV444P; + avctx->colorspace = AVCOL_SPC_BT470BG; + break; + case MKTAG('U', 'M', 'H', '4'): + c->planes = 3; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_YUV444P; + avctx->colorspace = AVCOL_SPC_BT709; + break; + case MKTAG('U', 'M', 'R', 'G'): + c->planes = 3; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_GBRP; + break; + case MKTAG('U', 'M', 'R', 'A'): + c->planes = 4; + c->pack = 1; + avctx->pix_fmt = AV_PIX_FMT_GBRAP; + break; default: av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n", avctx->codec_tag); return AVERROR_INVALIDDATA; } + if (c->pack && avctx->extradata_size >= 16) { + av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n", + avctx->extradata[3], avctx->extradata[2], + avctx->extradata[1], avctx->extradata[0]); + av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n", + AV_RB32(avctx->extradata + 4)); + c->compression = avctx->extradata[8]; + if (c->compression != 2) + avpriv_request_sample(avctx, "Unknown compression type"); + c->slices = avctx->extradata[9] + 1; + } else if (avctx->extradata_size >= 16) { + av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n", + avctx->extradata[3], avctx->extradata[2], + avctx->extradata[1], avctx->extradata[0]); + av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n", + AV_RB32(avctx->extradata + 4)); + c->frame_info_size = AV_RL32(avctx->extradata + 8); + c->flags = AV_RL32(avctx->extradata + 12); + + if (c->frame_info_size != 4) + avpriv_request_sample(avctx, "Frame info not 4 bytes"); + av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags); + c->slices = (c->flags >> 24) + 1; + c->compression = c->flags & 1; + c->interlaced = c->flags & 0x800; + } else if (avctx->extradata_size == 8) { + av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n", + avctx->extradata[3], avctx->extradata[2], + avctx->extradata[1], avctx->extradata[0]); + av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n", + AV_RB32(avctx->extradata + 4)); + c->interlaced = 0; + c->pro = 1; + c->frame_info_size = 4; + } else { + av_log(avctx, AV_LOG_ERROR, + "Insufficient extradata size %d, should be at least 16\n", + avctx->extradata_size); + return AVERROR_INVALIDDATA; + } + return 0; } |