From: "Wang, Fei W" <fei.w.wang-at-intel.com@ffmpeg.org> To: "ffmpeg-devel@ffmpeg.org" <ffmpeg-devel@ffmpeg.org>, "Xiang, Haihao" <haihao.xiang@intel.com> Cc: "sw@jkqxz.net" <sw@jkqxz.net> Subject: Re: [FFmpeg-devel] [PATCH v1 4/4] vaapi_encode_h265: Query encoding block sizes and features Date: Mon, 21 Feb 2022 02:21:21 +0000 Message-ID: <1bcc3ab6d430f5b3831beb9748fdb791c0117bde.camel@intel.com> (raw) In-Reply-To: <4da316fe210ae6561a1a5036cc3a2fccb385253b.camel@intel.com> On Fri, 2022-02-18 at 04:38 +0000, Xiang, Haihao wrote: > On Fri, 2022-02-18 at 09:43 +0800, Fei Wang wrote: > > From: Mark Thompson <sw@jkqxz.net> > > > > Signed-off-by: Fei Wang <fei.w.wang@intel.com> > > --- > > libavcodec/vaapi_encode_h265.c | 107 > > +++++++++++++++++++++++++++++++-- > > 1 file changed, 102 insertions(+), 5 deletions(-) > > > > diff --git a/libavcodec/vaapi_encode_h265.c > > b/libavcodec/vaapi_encode_h265.c > > index 8319848e4a..e98502503d 100644 > > --- a/libavcodec/vaapi_encode_h265.c > > +++ b/libavcodec/vaapi_encode_h265.c > > @@ -56,6 +56,9 @@ typedef struct VAAPIEncodeH265Context { > > VAAPIEncodeContext common; > > > > // Encoder features. > > + uint32_t va_features; > > + // Block size info. > > + uint32_t va_bs; > > uint32_t ctu_size; > > uint32_t min_cb_size; > > > > @@ -427,9 +430,9 @@ static int > > vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx) > > vps->vps_max_latency_increase_plus1[i]; > > } > > > > - // These have to come from the capabilities of the > > encoder. We have no > > - // way to query them, so just hardcode parameters which work > > on the Intel > > - // driver. > > + // These values come from the capabilities of the first > > encoder > > + // implementation in the i965 driver on Intel Skylake. They > > may > > + // fail badly with other platforms or drivers. > > // CTB size from 8x8 to 32x32. > > sps->log2_min_luma_coding_block_size_minus3 = 0; > > sps->log2_diff_max_min_luma_coding_block_size = 2; > > @@ -447,6 +450,42 @@ static int > > vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx) > > > > sps->pcm_enabled_flag = 0; > > > > +// update sps setting according to queried result > > +#if VA_CHECK_VERSION(1, 13, 0) > > + if (priv->va_features) { > > + VAConfigAttribValEncHEVCFeatures features = { .value = > > priv- > > > va_features }; > > + > > + // Enable feature if get queried result is > > VA_FEATURE_SUPPORTED | > > VA_FEATURE_REQUIRED > > + sps->amp_enabled_flag = > > + !!features.bits.amp; > > + sps->sample_adaptive_offset_enabled_flag = > > + !!features.bits.sao; > > + sps->sps_temporal_mvp_enabled_flag = > > + !!features.bits.temporal_mvp; > > + sps->pcm_enabled_flag = > > + !!features.bits.pcm; > > + } > > + > > + if (priv->va_bs) { > > + VAConfigAttribValEncHEVCBlockSizes bs = { .value = priv- > > >va_bs }; > > + sps->log2_min_luma_coding_block_size_minus3 = > > + ff_ctz(priv->min_cb_size) - 3; > > + sps->log2_diff_max_min_luma_coding_block_size = > > + ff_ctz(priv->ctu_size) - ff_ctz(priv->min_cb_size); > > + > > + sps->log2_min_luma_transform_block_size_minus2 = > > + bs.bits.log2_min_luma_transform_block_size_minus2; > > + sps->log2_diff_max_min_luma_transform_block_size = > > + bs.bits.log2_max_luma_transform_block_size_minus2 - > > + bs.bits.log2_min_luma_transform_block_size_minus2; > > + > > + sps->max_transform_hierarchy_depth_inter = > > + bs.bits.max_max_transform_hierarchy_depth_inter; > > + sps->max_transform_hierarchy_depth_intra = > > + bs.bits.max_max_transform_hierarchy_depth_intra; > > + } > > +#endif > > + > > // STRPSs should ideally be here rather than defined > > individually in > > // each slice, but the structure isn't completely fixed so for > > now > > // don't bother. > > @@ -539,6 +578,23 @@ static int > > vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx) > > pps->cu_qp_delta_enabled_flag = (ctx->va_rc_mode != > > VA_RC_CQP); > > pps->diff_cu_qp_delta_depth = 0; > > > > +// update pps setting according to queried result > > +#if VA_CHECK_VERSION(1, 13, 0) > > + if (priv->va_features) { > > + VAConfigAttribValEncHEVCFeatures features = { .value = > > priv- > > > va_features }; > > + if (ctx->va_rc_mode != VA_RC_CQP) > > + pps->cu_qp_delta_enabled_flag = > > + !!features.bits.cu_qp_delta; > > Please fix the indentation Thanks Haihao to help review. Fixed in V2. Thanks Fei > > > + > > + pps->transform_skip_enabled_flag = > > + !!features.bits.transform_skip; > > + // set diff_cu_qp_delta_depth as its max value if > > cu_qp_delta > > enabled. Otherwise > > + // 0 will make cu_qp_delta invalid. > > + if (pps->cu_qp_delta_enabled_flag) > > + pps->diff_cu_qp_delta_depth = sps- > > > log2_diff_max_min_luma_coding_block_size; > > + } > > +#endif > > + > > if (ctx->tile_rows && ctx->tile_cols) { > > int uniform_spacing; > > > > @@ -640,8 +696,8 @@ static int > > vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx) > > > > .coded_buf = VA_INVALID_ID, > > > > - .collocated_ref_pic_index = 0xff, > > - > > + .collocated_ref_pic_index = sps- > > >sps_temporal_mvp_enabled_flag ? > > + 0 : 0xff, > > .last_picture = 0, > > > > .pic_init_qp = pps->init_qp_minus26 + 26, > > @@ -674,6 +730,8 @@ static int > > vaapi_encode_h265_init_sequence_params(AVCodecContext *avctx) > > .entropy_coding_sync_enabled_flag = pps- > > > entropy_coding_sync_enabled_flag, > > .loop_filter_across_tiles_enabled_flag = > > pps->loop_filter_across_tiles_enabled_flag, > > + .pps_loop_filter_across_slices_enabled_flag = > > + pps->pps_loop_filter_across_slices_enabled_flag, > > .scaling_list_data_present_flag = (sps- > > > sps_scaling_list_data_present_flag | > > pps- > > > pps_scaling_list_data_present_flag), > > .screen_content_flag = 0, > > @@ -1105,6 +1163,45 @@ static av_cold void > > vaapi_encode_h265_get_encoder_caps(AVCodecContext *avctx) > > VAAPIEncodeContext *ctx = avctx->priv_data; > > VAAPIEncodeH265Context *priv = avctx->priv_data; > > > > +#if VA_CHECK_VERSION(1, 13, 0) > > + { > > + VAConfigAttribValEncHEVCBlockSizes block_size; > > + VAConfigAttrib attr; > > + VAStatus vas; > > + > > + attr.type = VAConfigAttribEncHEVCFeatures; > > + vas = vaGetConfigAttributes(ctx->hwctx->display, ctx- > > >va_profile, > > + ctx->va_entrypoint, &attr, 1); > > + if (vas != VA_STATUS_SUCCESS) { > > + av_log(avctx, AV_LOG_WARNING, "Failed to query encoder > > " > > + "features, using guessed defaults.\n"); > > + } else if (attr.value == VA_ATTRIB_NOT_SUPPORTED) { > > + av_log(avctx, AV_LOG_WARNING, "Driver does not > > advertise " > > + "encoder features, using guessed defaults.\n"); > > + } else { > > + priv->va_features = attr.value; > > + } > > + > > + attr.type = VAConfigAttribEncHEVCBlockSizes; > > + vas = vaGetConfigAttributes(ctx->hwctx->display, ctx- > > >va_profile, > > + ctx->va_entrypoint, &attr, 1); > > + if (vas != VA_STATUS_SUCCESS) { > > + av_log(avctx, AV_LOG_WARNING, "Failed to query encoder > > " > > + "block size, using guessed defaults.\n"); > > + } else if (attr.value == VA_ATTRIB_NOT_SUPPORTED) { > > + av_log(avctx, AV_LOG_WARNING, "Driver does not > > advertise " > > + "encoder block size, using guessed > > defaults.\n"); > > + } else { > > + priv->va_bs = block_size.value = attr.value; > > + > > + priv->ctu_size = > > + 1 << > > block_size.bits.log2_max_coding_tree_block_size_minus3 + > > 3; > > + priv->min_cb_size = > > + 1 << > > block_size.bits.log2_min_luma_coding_block_size_minus3 + > > 3; > > + } > > + } > > +#endif > > + > > if (!priv->ctu_size) { > > priv->ctu_size = 32; > > priv->min_cb_size = 16; > > Tested with i965 and iHD drivers and it works well for me > > Thanks > Haihao > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
prev parent reply other threads:[~2022-02-21 2:21 UTC|newest] Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-02-18 1:42 [FFmpeg-devel] [PATCH v1 1/4] lavc/vaapi_encode_h265: Add P frame to GPB frame support for hevc_vaapi Fei Wang 2022-02-18 1:42 ` [FFmpeg-devel] [PATCH v1 2/4] vaapi_encode: Move block size calculation after entrypoint selection Fei Wang 2022-02-18 1:43 ` [FFmpeg-devel] [PATCH v1 3/4] vaapi_encode_h265: Explicitly set and correct some flags Fei Wang 2022-02-18 1:43 ` [FFmpeg-devel] [PATCH v1 4/4] vaapi_encode_h265: Query encoding block sizes and features Fei Wang 2022-02-18 4:38 ` Xiang, Haihao 2022-02-21 2:21 ` Wang, Fei W [this message]
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=1bcc3ab6d430f5b3831beb9748fdb791c0117bde.camel@intel.com \ --to=fei.w.wang-at-intel.com@ffmpeg.org \ --cc=ffmpeg-devel@ffmpeg.org \ --cc=haihao.xiang@intel.com \ --cc=sw@jkqxz.net \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git