From: "Tomas Härdin" <tjoppen@acc.umu.se> To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org> Subject: [FFmpeg-devel] [PATCH 13/13] lavc/jpeg2000dec: Component-level threading of write_frame() Date: Tue, 14 Jun 2022 16:47:23 +0200 Message-ID: <7b9cfcb8fe330310e5d7558d67d424d626fe6719.camel@acc.umu.se> (raw) In-Reply-To: <10ec51ef44325c2de6d5de7b994a9b6c8eb5e3a2.camel@acc.umu.se> [-- Attachment #1: Type: text/plain, Size: 212 bytes --] Don't have access to the full machine to test this with 96 threads. On 2/3rds of an AMD EPYC 7R32 (-threads 64) it runs at 50 fps. Specifically the decoder uses 59.2 seconds to decode a 60.0 second clip. /Tomas [-- Attachment #2: 0013-lavc-jpeg2000dec-Component-level-threading-of-write_.patch --] [-- Type: text/x-patch, Size: 5575 bytes --] From 19fc2413dc2bafff577c68830cde48e08138771e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Tomas=20H=C3=A4rdin?= <git@haerdin.se> Date: Tue, 14 Jun 2022 15:45:32 +0200 Subject: [PATCH 13/13] lavc/jpeg2000dec: Component-level threading of write_frame() Split off MCT and don't bother with it unless the picture actually uses MCT. --- libavcodec/jpeg2000dec.c | 35 ++++++++++++++++++++++++----------- 1 file changed, 24 insertions(+), 11 deletions(-) diff --git a/libavcodec/jpeg2000dec.c b/libavcodec/jpeg2000dec.c index 18ebe5219d..8eaeda1c66 100644 --- a/libavcodec/jpeg2000dec.c +++ b/libavcodec/jpeg2000dec.c @@ -156,6 +156,7 @@ typedef struct Jpeg2000DecoderContext { // used for idwt slicing int reslevel, dir, slices; int have_dwt97_int; // 1 if any coding style is FF_DWT97_INT + int have_mct; } Jpeg2000DecoderContext; /* get_bits functions for JPEG2000 packet bitstream @@ -600,6 +601,9 @@ static int get_cod(Jpeg2000DecoderContext *s, Jpeg2000CodingStyle *c, return AVERROR_INVALIDDATA; } + if (tmp.mct) + s->have_mct = 1; + if ((ret = get_cox(s, &tmp)) < 0) return ret; tmp.init = 1; @@ -2073,16 +2077,14 @@ static int jpeg2000_dwt97_int_postshift(AVCodecContext *avctx, void *td, #define WRITE_FRAME(D, PIXEL) \ static inline void write_frame_ ## D(Jpeg2000DecoderContext * s, Jpeg2000Tile * tile, \ - AVFrame * picture, int precision) \ + AVFrame * picture, int precision, int compno) \ { \ const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(s->avctx->pix_fmt); \ int planar = !!(pixdesc->flags & AV_PIX_FMT_FLAG_PLANAR); \ int pixelsize = planar ? 1 : pixdesc->nb_components; \ \ - int compno; \ int x, y; \ \ - for (compno = 0; compno < s->ncomponents; compno++) { \ Jpeg2000Component *comp = tile->comp + compno; \ Jpeg2000CodingStyle *codsty = tile->codsty + compno; \ PIXEL *line; \ @@ -2129,8 +2131,6 @@ static int jpeg2000_dwt97_int_postshift(AVCodecContext *avctx, void *td, } \ line += picture->linesize[plane] / sizeof(PIXEL); \ } \ - } \ - \ } WRITE_FRAME(8, uint8_t) @@ -2138,26 +2138,36 @@ WRITE_FRAME(16, uint16_t) #undef WRITE_FRAME -static int jpeg2000_mct_write_frame(AVCodecContext *avctx, void *td, - int jobnr, int threadnr) +static int jpeg2000_mct(AVCodecContext *avctx, void *td, + int jobnr, int threadnr) { Jpeg2000DecoderContext *s = avctx->priv_data; - AVFrame *picture = td; Jpeg2000Tile *tile = s->tile + jobnr; /* inverse MCT transformation */ if (tile->codsty[0].mct) mct_decode(s, tile); + return 0; +} + +static int jpeg2000_write_frame(AVCodecContext *avctx, void *td, + int jobnr, int threadnr) +{ + Jpeg2000DecoderContext *s = avctx->priv_data; + AVFrame *picture = td; + Jpeg2000Tile *tile = s->tile + jobnr / s->ncomponents; + int compno = jobnr % s->ncomponents; + if (s->precision <= 8) { - write_frame_8(s, tile, picture, 8); + write_frame_8(s, tile, picture, 8, compno); } else { int precision = picture->format == AV_PIX_FMT_XYZ12 || picture->format == AV_PIX_FMT_RGB48 || picture->format == AV_PIX_FMT_RGBA64 || picture->format == AV_PIX_FMT_GRAY16 ? 16 : s->precision; - write_frame_16(s, tile, picture, precision); + write_frame_16(s, tile, picture, precision, compno); } return 0; @@ -2695,7 +2705,10 @@ static int jpeg2000_decode_frame(AVCodecContext *avctx, AVFrame *picture, if (s->have_dwt97_int) avctx->execute2(avctx, jpeg2000_dwt97_int_postshift, NULL, NULL, s->numXtiles * s->numYtiles * s->ncomponents * s->slices); - avctx->execute2(avctx, jpeg2000_mct_write_frame, picture, NULL, s->numXtiles * s->numYtiles); + if (s->have_mct) + avctx->execute2(avctx, jpeg2000_mct, NULL, NULL, s->numXtiles * s->numYtiles); + + avctx->execute2(avctx, jpeg2000_write_frame, picture, NULL, s->numXtiles * s->numYtiles * s->ncomponents); jpeg2000_dec_cleanup(s, 0); -- 2.30.2 [-- Attachment #3: Type: text/plain, Size: 251 bytes --] _______________________________________________ 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".
next prev parent reply other threads:[~2022-06-14 14:47 UTC|newest] Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-06-14 14:39 [FFmpeg-devel] [PATCH 01/13] lavc/jpeg2000dec: Finer granularity threading Tomas Härdin 2022-06-14 14:39 ` [FFmpeg-devel] [PATCH 02/13] lavc/jpeg2000dec: Reindent Tomas Härdin 2022-06-14 14:40 ` [FFmpeg-devel] [PATCH 03/13] lavc/jpeg2000dwt: Implement sliced transforms Tomas Härdin 2022-06-14 14:40 ` [FFmpeg-devel] [PATCH 04/13] lavc/jpeg2000dec: Implement IDWT slicing Tomas Härdin 2022-06-14 14:41 ` [FFmpeg-devel] [PATCH 05/13] lavc/jpeg2000dec: Thread init_tile() Tomas Härdin 2022-06-14 21:11 ` Michael Niedermayer 2022-06-15 13:11 ` Tomas Härdin 2022-06-15 21:05 ` Michael Niedermayer 2022-06-16 9:28 ` Tomas Härdin 2022-06-14 14:42 ` [FFmpeg-devel] [PATCH 06/13] lavu/mem: Add ff_fast_recalloc() Tomas Härdin 2022-06-14 20:26 ` Michael Niedermayer 2022-06-15 9:59 ` Tomas Härdin 2022-06-15 12:15 ` James Almer 2022-06-16 12:44 ` Tomas Härdin 2022-06-18 14:57 ` Anton Khirnov 2022-06-21 8:04 ` Tomas Härdin 2022-06-14 14:42 ` [FFmpeg-devel] [PATCH 07/13] lavc/jpeg2000*: Use ff_fast_recalloc() to eliminate lots of allocations Tomas Härdin 2022-06-14 15:23 ` Andreas Rheinhardt 2022-06-15 10:03 ` Tomas Härdin 2022-06-14 14:43 ` [FFmpeg-devel] [PATCH 08/13] lavc/jpeg2000: Switch Jpeg2000TgtNode to int32_t parent Tomas Härdin 2022-06-14 14:43 ` [FFmpeg-devel] [PATCH 09/13] lavc/jpeg2000: Speed up ff_jpeg2000_tag_tree_init() using stereotypes for sizes <= 4x4 Tomas Härdin 2022-06-18 15:00 ` Anton Khirnov 2022-06-21 7:57 ` Tomas Härdin 2022-06-14 14:43 ` [FFmpeg-devel] [PATCH 10/13] lavc/jpeg2000: Reindent Tomas Härdin 2022-06-14 14:44 ` [FFmpeg-devel] [PATCH 11/13] lavc/jpeg2000: Minimize calls to av_codec_is_encoder() Tomas Härdin 2022-06-14 15:04 ` Andreas Rheinhardt 2022-06-15 10:20 ` Tomas Härdin 2022-06-14 14:44 ` [FFmpeg-devel] [PATCH 12/13] lavc/jpeg2000dec: Use coarser slicing for initial reslevels Tomas Härdin 2022-06-14 14:47 ` Tomas Härdin [this message] 2022-06-18 14:50 ` [FFmpeg-devel] [PATCH 01/13] lavc/jpeg2000dec: Finer granularity threading Anton Khirnov 2022-06-24 8:19 ` Tomas Härdin
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=7b9cfcb8fe330310e5d7558d67d424d626fe6719.camel@acc.umu.se \ --to=tjoppen@acc.umu.se \ --cc=ffmpeg-devel@ffmpeg.org \ /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