Some measurements: cpu: Intel(R) Xeon(R) Silver 4210R CPU @ 2.40GHz gpu: Quadro RTX 4000
GPU H264 DECODING ================= NOTE: - first column is max floating average over past 10s - second column is overall maximum during measurement (at least 1 min) - measure for custom app using new ffmpeg api, a very few ms could maybe be trimmed off by more efficient implementation - 1x LOW DELAY OFF is there for comparison 1x LOW DELAY ON --------------- h264_1920_1080_420_8b_120p_bcount:0_crf:14.ts 14ms 26ms h264_1920_1080_420_8b_120p_bcount:2_crf:14.ts 28ms 69ms h264_1920_1080_420_8b_120p_bcount:4_crf:14.ts 28ms 78ms h264_1920_1080_420_8b_30p_bcount:0_crf:14.ts 41ms 53ms h264_1920_1080_420_8b_30p_bcount:2_crf:14.ts 105ms 244ms h264_1920_1080_420_8b_30p_bcount:4_crf:14.ts 105ms 280ms h264_1920_1080_420_8b_60p_bcount:0_crf:14.ts 23ms 36ms h264_1920_1080_420_8b_60p_bcount:2_crf:14.ts 53ms 108ms h264_1920_1080_420_8b_60p_bcount:4_crf:14.ts 54ms 129ms h264_2880_1620_420_8b_60p_bcount:0_crf:14.ts 27ms 41ms h264_2880_1620_420_8b_60p_bcount:2_crf:14.ts 55ms 111ms h264_2880_1620_420_8b_60p_bcount:4_crf:14.ts 55ms 131ms h264_2880_1620_420_8b_120p_bcount:0_crf:14.ts 20ms 46ms h264_2880_1620_420_8b_120p_bcount:2_crf:14.ts 30ms 70ms h264_2880_1620_420_8b_120p_bcount:4_crf:14.ts 31ms 79ms 1x LOW DELAY OFF ---------------- h264_1920_1080_420_8b_120p_bcount:0_crf:14.ts 36ms 48ms h264_1920_1080_420_8b_120p_bcount:2_crf:14.ts 44ms 85ms h264_1920_1080_420_8b_120p_bcount:4_crf:14.ts 53ms 103ms h264_1920_1080_420_8b_30p_bcount:0_crf:14.ts 137ms 151ms h264_1920_1080_420_8b_30p_bcount:2_crf:14.ts 172ms 309ms h264_1920_1080_420_8b_30p_bcount:4_crf:14.ts 206ms 380ms h264_1920_1080_420_8b_60p_bcount:0_crf:14.ts 69ms 82ms h264_1920_1080_420_8b_60p_bcount:2_crf:14.ts 86ms 162ms h264_1920_1080_420_8b_60p_bcount:4_crf:14.ts 104ms 210ms h264_2880_1620_420_8b_60p_bcount:0_crf:14.ts 71ms 86ms h264_2880_1620_420_8b_60p_bcount:2_crf:14.ts 88ms 161ms h264_2880_1620_420_8b_60p_bcount:4_crf:14.ts 105ms 212ms h264_2880_1620_420_8b_120p_bcount:0_crf:14.ts 38ms 52ms h264_2880_1620_420_8b_120p_bcount:2_crf:14.ts 46ms 85ms h264_2880_1620_420_8b_120p_bcount:4_crf:14.ts 54ms 106ms 3x LOW DELAY ON --------------- (nvdec: 40% load) h264_1920_1080_420_8b_120p_bcount:0_crf:14.ts 14ms 31ms h264_1920_1080_420_8b_120p_bcount:2_crf:14.ts 27ms 67ms h264_1920_1080_420_8b_120p_bcount:4_crf:14.ts 28ms 77ms (nvdec: 24% load) h264_1920_1080_420_8b_30p_bcount:0_crf:14.ts 41ms 58ms h264_1920_1080_420_8b_30p_bcount:2_crf:14.ts 105ms 246ms h264_1920_1080_420_8b_30p_bcount:4_crf:14.ts 105ms 277ms (nvdec: 35% load) h264_1920_1080_420_8b_60p_bcount:0_crf:14.ts 23ms 35ms h264_1920_1080_420_8b_60p_bcount:2_crf:14.ts 53ms 109ms h264_1920_1080_420_8b_60p_bcount:4_crf:14.ts 53ms 130ms (nvdec: 36% load) h264_2880_1620_420_8b_60p_bcount:0_crf:14.ts 26ms 44ms h264_2880_1620_420_8b_60p_bcount:2_crf:14.ts 54ms 112ms h264_2880_1620_420_8b_60p_bcount:4_crf:14.ts 55ms 133ms (nvdec: 62% load) h264_2880_1620_420_8b_120p_bcount:0_crf:14.ts 18ms 43ms h264_2880_1620_420_8b_120p_bcount:2_crf:14.ts 30ms 71ms h264_2880_1620_420_8b_120p_bcount:4_crf:14.ts 31ms 80ms 7x LOW DELAY ON --------------- (nvdec: 70% load) h264_1920_1080_420_8b_120p_bcount:0_crf:14.ts 15ms 40ms h264_1920_1080_420_8b_120p_bcount:2_crf:14.ts 28ms 69ms h264_1920_1080_420_8b_120p_bcount:4_crf:14.ts 29ms 78ms 11x LOW DELAY ON ---------------- (nvdec: 65% load) h264_1920_1080_420_8b_60p_bcount:0_crf:14.ts 23ms 45ms h264_1920_1080_420_8b_60p_bcount:2_crf:14.ts 53ms 108ms h264_1920_1080_420_8b_60p_bcount:4_crf:14.ts 54ms 131ms 6x LOW DELAY ON --------------- (nvdec: 70% load) h264_2880_1620_420_8b_60p_bcount:0_crf:14.ts 28ms 53ms h264_2880_1620_420_8b_60p_bcount:2_crf:14.ts 55ms 112ms h264_2880_1620_420_8b_60p_bcount:4_crf:14.ts 55ms 134ms 4x LOW DELAY ON --------------- (nvdec: 78% load) h264_2880_1620_420_8b_120p_bcount:0_crf:14.ts 24ms 58ms h264_2880_1620_420_8b_120p_bcount:2_crf:14.ts 31ms 70ms h264_2880_1620_420_8b_120p_bcount:4_crf:14.ts 33ms 79ms Sample bitrates for bcount:0 inputs: h264_1920_1080_420_8b_120p_bcount:0_crf:14.ts: 89890 kb/s h264_1920_1080_420_8b_30p_bcount:0_crf:14.ts: 66513 kb/s h264_1920_1080_420_8b_60p_bcount:0_crf:14.ts: 85399 kb/s h264_2880_1620_420_8b_60p_bcount:0_crf:14.ts: 138562 kb/s h264_2880_1620_420_8b_120p_bcount:0_crf:14.ts: 149715 kb/s Best regards Michal Novotny On Mon, 22 Feb 2021 at 18:06, clime <cli...@gmail.com> wrote: > > From: Michal Novotny <michal.novo...@comprimato.com> > > * zero is recommended value in Nvidia coding samples for low latency use-cases > > Signed-off-by: Michal Novotny <michal.novo...@comprimato.com> > --- > libavcodec/cuviddec.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/libavcodec/cuviddec.c b/libavcodec/cuviddec.c > index 49775b5a09..ec57afdefe 100644 > --- a/libavcodec/cuviddec.c > +++ b/libavcodec/cuviddec.c > @@ -999,7 +999,7 @@ static av_cold int cuvid_decode_init(AVCodecContext > *avctx) > } > > ctx->cuparseinfo.ulMaxNumDecodeSurfaces = ctx->nb_surfaces; > - ctx->cuparseinfo.ulMaxDisplayDelay = 4; > + ctx->cuparseinfo.ulMaxDisplayDelay = (avctx->flags & > AV_CODEC_FLAG_LOW_DELAY) ? 0 : 4; > ctx->cuparseinfo.pUserData = avctx; > ctx->cuparseinfo.pfnSequenceCallback = cuvid_handle_video_sequence; > ctx->cuparseinfo.pfnDecodePicture = cuvid_handle_picture_decode; > -- > 2.30.1 > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
