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

371 lines
12KB

  1. /*
  2. * Copyright (c) 2018 Ronald S. Bultje <rsbultje gmail com>
  3. * Copyright (c) 2018 James Almer <jamrial gmail com>
  4. *
  5. * This file is part of FFmpeg.
  6. *
  7. * FFmpeg is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU Lesser General Public
  9. * License as published by the Free Software Foundation; either
  10. * version 2.1 of the License, or (at your option) any later version.
  11. *
  12. * FFmpeg is distributed in the hope that it will be useful,
  13. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  15. * Lesser General Public License for more details.
  16. *
  17. * You should have received a copy of the GNU Lesser General Public
  18. * License along with FFmpeg; if not, write to the Free Software
  19. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20. */
  21. #include <dav1d/dav1d.h>
  22. #include "libavutil/avassert.h"
  23. #include "libavutil/mastering_display_metadata.h"
  24. #include "libavutil/imgutils.h"
  25. #include "libavutil/opt.h"
  26. #include "avcodec.h"
  27. #include "decode.h"
  28. #include "internal.h"
  29. typedef struct Libdav1dContext {
  30. AVClass *class;
  31. Dav1dContext *c;
  32. AVBufferPool *pool;
  33. int pool_size;
  34. Dav1dData data;
  35. int tile_threads;
  36. int frame_threads;
  37. int apply_grain;
  38. } Libdav1dContext;
  39. static const enum AVPixelFormat pix_fmt[][3] = {
  40. [DAV1D_PIXEL_LAYOUT_I400] = { AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12 },
  41. [DAV1D_PIXEL_LAYOUT_I420] = { AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV420P12 },
  42. [DAV1D_PIXEL_LAYOUT_I422] = { AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV422P12 },
  43. [DAV1D_PIXEL_LAYOUT_I444] = { AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV444P12 },
  44. };
  45. static const enum AVPixelFormat pix_fmt_rgb[3] = {
  46. AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12,
  47. };
  48. static void libdav1d_log_callback(void *opaque, const char *fmt, va_list vl)
  49. {
  50. AVCodecContext *c = opaque;
  51. av_vlog(c, AV_LOG_ERROR, fmt, vl);
  52. }
  53. static int libdav1d_picture_allocator(Dav1dPicture *p, void *cookie)
  54. {
  55. Libdav1dContext *dav1d = cookie;
  56. enum AVPixelFormat format = pix_fmt[p->p.layout][p->seq_hdr->hbd];
  57. int ret, linesize[4], h = FFALIGN(p->p.h, 128);
  58. uint8_t *aligned_ptr, *data[4];
  59. AVBufferRef *buf;
  60. ret = av_image_fill_arrays(data, linesize, NULL, format, FFALIGN(p->p.w, 128),
  61. h, DAV1D_PICTURE_ALIGNMENT);
  62. if (ret < 0)
  63. return ret;
  64. if (ret != dav1d->pool_size) {
  65. av_buffer_pool_uninit(&dav1d->pool);
  66. // Use twice the amount of required padding bytes for aligned_ptr below.
  67. dav1d->pool = av_buffer_pool_init(ret + DAV1D_PICTURE_ALIGNMENT * 2, NULL);
  68. if (!dav1d->pool) {
  69. dav1d->pool_size = 0;
  70. return AVERROR(ENOMEM);
  71. }
  72. dav1d->pool_size = ret;
  73. }
  74. buf = av_buffer_pool_get(dav1d->pool);
  75. if (!buf)
  76. return AVERROR(ENOMEM);
  77. // libdav1d requires DAV1D_PICTURE_ALIGNMENT aligned buffers, which av_malloc()
  78. // doesn't guarantee for example when AVX is disabled at configure time.
  79. // Use the extra DAV1D_PICTURE_ALIGNMENT padding bytes in the buffer to align it
  80. // if required.
  81. aligned_ptr = (uint8_t *)FFALIGN((uintptr_t)buf->data, DAV1D_PICTURE_ALIGNMENT);
  82. ret = av_image_fill_pointers(data, format, h, aligned_ptr, linesize);
  83. if (ret < 0) {
  84. av_buffer_unref(&buf);
  85. return ret;
  86. }
  87. p->data[0] = data[0];
  88. p->data[1] = data[1];
  89. p->data[2] = data[2];
  90. p->stride[0] = linesize[0];
  91. p->stride[1] = linesize[1];
  92. p->allocator_data = buf;
  93. return 0;
  94. }
  95. static void libdav1d_picture_release(Dav1dPicture *p, void *cookie)
  96. {
  97. AVBufferRef *buf = p->allocator_data;
  98. av_buffer_unref(&buf);
  99. }
  100. static av_cold int libdav1d_init(AVCodecContext *c)
  101. {
  102. Libdav1dContext *dav1d = c->priv_data;
  103. Dav1dSettings s;
  104. int threads = (c->thread_count ? c->thread_count : av_cpu_count()) * 3 / 2;
  105. int res;
  106. av_log(c, AV_LOG_INFO, "libdav1d %s\n", dav1d_version());
  107. dav1d_default_settings(&s);
  108. s.logger.cookie = c;
  109. s.logger.callback = libdav1d_log_callback;
  110. s.allocator.cookie = dav1d;
  111. s.allocator.alloc_picture_callback = libdav1d_picture_allocator;
  112. s.allocator.release_picture_callback = libdav1d_picture_release;
  113. s.frame_size_limit = c->max_pixels;
  114. s.apply_grain = dav1d->apply_grain;
  115. s.n_tile_threads = dav1d->tile_threads
  116. ? dav1d->tile_threads
  117. : FFMIN(floor(sqrt(threads)), DAV1D_MAX_TILE_THREADS);
  118. s.n_frame_threads = dav1d->frame_threads
  119. ? dav1d->frame_threads
  120. : FFMIN(ceil(threads / s.n_tile_threads), DAV1D_MAX_FRAME_THREADS);
  121. av_log(c, AV_LOG_DEBUG, "Using %d frame threads, %d tile threads\n",
  122. s.n_frame_threads, s.n_tile_threads);
  123. res = dav1d_open(&dav1d->c, &s);
  124. if (res < 0)
  125. return AVERROR(ENOMEM);
  126. return 0;
  127. }
  128. static void libdav1d_flush(AVCodecContext *c)
  129. {
  130. Libdav1dContext *dav1d = c->priv_data;
  131. dav1d_data_unref(&dav1d->data);
  132. dav1d_flush(dav1d->c);
  133. }
  134. static void libdav1d_data_free(const uint8_t *data, void *opaque) {
  135. AVBufferRef *buf = opaque;
  136. av_buffer_unref(&buf);
  137. }
  138. static int libdav1d_receive_frame(AVCodecContext *c, AVFrame *frame)
  139. {
  140. Libdav1dContext *dav1d = c->priv_data;
  141. Dav1dData *data = &dav1d->data;
  142. Dav1dPicture pic = { 0 }, *p = &pic;
  143. int res;
  144. if (!data->sz) {
  145. AVPacket pkt = { 0 };
  146. res = ff_decode_get_packet(c, &pkt);
  147. if (res < 0 && res != AVERROR_EOF)
  148. return res;
  149. if (pkt.size) {
  150. res = dav1d_data_wrap(data, pkt.data, pkt.size, libdav1d_data_free, pkt.buf);
  151. if (res < 0) {
  152. av_packet_unref(&pkt);
  153. return res;
  154. }
  155. data->m.timestamp = pkt.pts;
  156. data->m.offset = pkt.pos;
  157. data->m.duration = pkt.duration;
  158. pkt.buf = NULL;
  159. av_packet_unref(&pkt);
  160. }
  161. }
  162. res = dav1d_send_data(dav1d->c, data);
  163. if (res < 0) {
  164. if (res == AVERROR(EINVAL))
  165. res = AVERROR_INVALIDDATA;
  166. if (res != AVERROR(EAGAIN))
  167. return res;
  168. }
  169. res = dav1d_get_picture(dav1d->c, p);
  170. if (res < 0) {
  171. if (res == AVERROR(EINVAL))
  172. res = AVERROR_INVALIDDATA;
  173. else if (res == AVERROR(EAGAIN) && c->internal->draining)
  174. res = AVERROR_EOF;
  175. return res;
  176. }
  177. av_assert0(p->data[0] && p->allocator_data);
  178. // This requires the custom allocator above
  179. frame->buf[0] = av_buffer_ref(p->allocator_data);
  180. if (!frame->buf[0]) {
  181. dav1d_picture_unref(p);
  182. return AVERROR(ENOMEM);
  183. }
  184. frame->data[0] = p->data[0];
  185. frame->data[1] = p->data[1];
  186. frame->data[2] = p->data[2];
  187. frame->linesize[0] = p->stride[0];
  188. frame->linesize[1] = p->stride[1];
  189. frame->linesize[2] = p->stride[1];
  190. c->profile = p->seq_hdr->profile;
  191. c->level = ((p->seq_hdr->operating_points[0].major_level - 2) << 2)
  192. | p->seq_hdr->operating_points[0].minor_level;
  193. frame->width = p->p.w;
  194. frame->height = p->p.h;
  195. if (c->width != p->p.w || c->height != p->p.h) {
  196. res = ff_set_dimensions(c, p->p.w, p->p.h);
  197. if (res < 0)
  198. goto fail;
  199. }
  200. switch (p->seq_hdr->chr) {
  201. case DAV1D_CHR_VERTICAL:
  202. frame->chroma_location = c->chroma_sample_location = AVCHROMA_LOC_LEFT;
  203. break;
  204. case DAV1D_CHR_COLOCATED:
  205. frame->chroma_location = c->chroma_sample_location = AVCHROMA_LOC_TOPLEFT;
  206. break;
  207. }
  208. frame->colorspace = c->colorspace = (enum AVColorSpace) p->seq_hdr->mtrx;
  209. frame->color_primaries = c->color_primaries = (enum AVColorPrimaries) p->seq_hdr->pri;
  210. frame->color_trc = c->color_trc = (enum AVColorTransferCharacteristic) p->seq_hdr->trc;
  211. frame->color_range = c->color_range = p->seq_hdr->color_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
  212. if (p->p.layout == DAV1D_PIXEL_LAYOUT_I444 &&
  213. p->seq_hdr->mtrx == DAV1D_MC_IDENTITY &&
  214. p->seq_hdr->pri == DAV1D_COLOR_PRI_BT709 &&
  215. p->seq_hdr->trc == DAV1D_TRC_SRGB)
  216. frame->format = c->pix_fmt = pix_fmt_rgb[p->seq_hdr->hbd];
  217. else
  218. frame->format = c->pix_fmt = pix_fmt[p->p.layout][p->seq_hdr->hbd];
  219. // match timestamps and packet size
  220. frame->pts = frame->best_effort_timestamp = p->m.timestamp;
  221. #if FF_API_PKT_PTS
  222. FF_DISABLE_DEPRECATION_WARNINGS
  223. frame->pkt_pts = p->m.timestamp;
  224. FF_ENABLE_DEPRECATION_WARNINGS
  225. #endif
  226. frame->pkt_dts = p->m.timestamp;
  227. frame->pkt_pos = p->m.offset;
  228. frame->pkt_size = p->m.size;
  229. frame->pkt_duration = p->m.duration;
  230. frame->key_frame = p->frame_hdr->frame_type == DAV1D_FRAME_TYPE_KEY;
  231. switch (p->frame_hdr->frame_type) {
  232. case DAV1D_FRAME_TYPE_KEY:
  233. case DAV1D_FRAME_TYPE_INTRA:
  234. frame->pict_type = AV_PICTURE_TYPE_I;
  235. break;
  236. case DAV1D_FRAME_TYPE_INTER:
  237. frame->pict_type = AV_PICTURE_TYPE_P;
  238. break;
  239. case DAV1D_FRAME_TYPE_SWITCH:
  240. frame->pict_type = AV_PICTURE_TYPE_SP;
  241. break;
  242. default:
  243. res = AVERROR_INVALIDDATA;
  244. goto fail;
  245. }
  246. if (p->mastering_display) {
  247. AVMasteringDisplayMetadata *mastering = av_mastering_display_metadata_create_side_data(frame);
  248. if (!mastering) {
  249. res = AVERROR(ENOMEM);
  250. goto fail;
  251. }
  252. for (int i = 0; i < 3; i++) {
  253. mastering->display_primaries[i][0] = av_make_q(p->mastering_display->primaries[i][0], 1 << 16);
  254. mastering->display_primaries[i][1] = av_make_q(p->mastering_display->primaries[i][1], 1 << 16);
  255. }
  256. mastering->white_point[0] = av_make_q(p->mastering_display->white_point[0], 1 << 16);
  257. mastering->white_point[1] = av_make_q(p->mastering_display->white_point[1], 1 << 16);
  258. mastering->max_luminance = av_make_q(p->mastering_display->max_luminance, 1 << 8);
  259. mastering->min_luminance = av_make_q(p->mastering_display->min_luminance, 1 << 14);
  260. mastering->has_primaries = 1;
  261. mastering->has_luminance = 1;
  262. }
  263. if (p->content_light) {
  264. AVContentLightMetadata *light = av_content_light_metadata_create_side_data(frame);
  265. if (!light) {
  266. res = AVERROR(ENOMEM);
  267. goto fail;
  268. }
  269. light->MaxCLL = p->content_light->max_content_light_level;
  270. light->MaxFALL = p->content_light->max_frame_average_light_level;
  271. }
  272. res = 0;
  273. fail:
  274. dav1d_picture_unref(p);
  275. if (res < 0)
  276. av_frame_unref(frame);
  277. return res;
  278. }
  279. static av_cold int libdav1d_close(AVCodecContext *c)
  280. {
  281. Libdav1dContext *dav1d = c->priv_data;
  282. av_buffer_pool_uninit(&dav1d->pool);
  283. dav1d_data_unref(&dav1d->data);
  284. dav1d_close(&dav1d->c);
  285. return 0;
  286. }
  287. #define OFFSET(x) offsetof(Libdav1dContext, x)
  288. #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM
  289. static const AVOption libdav1d_options[] = {
  290. { "tilethreads", "Tile threads", OFFSET(tile_threads), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, DAV1D_MAX_TILE_THREADS, VD },
  291. { "framethreads", "Frame threads", OFFSET(frame_threads), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, DAV1D_MAX_FRAME_THREADS, VD },
  292. { "filmgrain", "Apply Film Grain", OFFSET(apply_grain), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, VD },
  293. { NULL }
  294. };
  295. static const AVClass libdav1d_class = {
  296. .class_name = "libdav1d decoder",
  297. .item_name = av_default_item_name,
  298. .option = libdav1d_options,
  299. .version = LIBAVUTIL_VERSION_INT,
  300. };
  301. AVCodec ff_libdav1d_decoder = {
  302. .name = "libdav1d",
  303. .long_name = NULL_IF_CONFIG_SMALL("dav1d AV1 decoder by VideoLAN"),
  304. .type = AVMEDIA_TYPE_VIDEO,
  305. .id = AV_CODEC_ID_AV1,
  306. .priv_data_size = sizeof(Libdav1dContext),
  307. .init = libdav1d_init,
  308. .close = libdav1d_close,
  309. .flush = libdav1d_flush,
  310. .receive_frame = libdav1d_receive_frame,
  311. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  312. .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_SETS_PKT_DTS,
  313. .priv_class = &libdav1d_class,
  314. .wrapper_name = "libdav1d",
  315. };