From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH v2 01/27] avcodec/threadprogress: Add new API for frame-threaded progress
Date: Thu, 18 Apr 2024 22:40:19 +0200
Message-ID: <AS8P250MB074431726EC8F7F9E82625018F0E2@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM> (raw)
In-Reply-To: <GV1P250MB073755270DAD40B30BE913B78F002@GV1P250MB0737.EURP250.PROD.OUTLOOK.COM>
Andreas Rheinhardt:
> The API is similar to the ThreadFrame API, with the exception
> that it no longer has an included AVFrame and that it has its
> own mutexes and condition variables which makes it more independent
> of pthread_frame.c. One can wait on anything via a ThreadProgress.
> One just has to ensure that the lifetime of the object containing
> the ThreadProgress is long enough. This will typically be solved
> by putting a ThreadProgress in a refcounted structure that is
> shared between threads.
>
> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
> ---
> Now including its own mutex and condition variable. Hopefully
> no system has constraints on the number of mutexes and condition
> variables it supports.
>
> libavcodec/threadprogress.c | 73 +++++++++++++++++++++++++++++
> libavcodec/threadprogress.h | 91 +++++++++++++++++++++++++++++++++++++
> 2 files changed, 164 insertions(+)
> create mode 100644 libavcodec/threadprogress.c
> create mode 100644 libavcodec/threadprogress.h
>
> diff --git a/libavcodec/threadprogress.c b/libavcodec/threadprogress.c
> new file mode 100644
> index 0000000000..dd2a4d2336
> --- /dev/null
> +++ b/libavcodec/threadprogress.c
> @@ -0,0 +1,73 @@
> +/*
> + * Copyright (c) 2022 Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
> + *
> + * This file is part of FFmpeg.
> + *
> + * 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.
> + *
> + * 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 FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + */
> +
> +#include <limits.h>
> +#include <stdatomic.h>
> +
> +#include "pthread_internal.h"
> +#include "threadprogress.h"
> +#include "libavutil/attributes.h"
> +
> +DEFINE_OFFSET_ARRAY(ThreadProgress, thread_progress, init,
> + (offsetof(ThreadProgress, progress_mutex)),
> + (offsetof(ThreadProgress, progress_cond)));
> +
> +av_cold int ff_thread_progress_init(ThreadProgress *pro, int init_mode)
> +{
> + atomic_init(&pro->progress, init_mode ? -1 : INT_MAX);
> + if (!init_mode) {
> + pro->init = 0;
> + return 0;
> + }
> + return ff_pthread_init(pro, thread_progress_offsets);
> +}
> +
> +av_cold void ff_thread_progress_destroy(ThreadProgress *pro)
> +{
> + ff_pthread_free(pro, thread_progress_offsets);
> +}
> +
> +void ff_thread_progress_report(ThreadProgress *pro, int n)
> +{
> + if (atomic_load_explicit(&pro->progress, memory_order_relaxed) >= n)
> + return;
> +
> + atomic_store_explicit(&pro->progress, n, memory_order_release);
> +
> + pthread_mutex_lock(&pro->progress_mutex);
> + pthread_cond_broadcast(&pro->progress_cond);
> + pthread_mutex_unlock(&pro->progress_mutex);
> +}
> +
> +void ff_thread_progress_await(const ThreadProgress *pro_c, int n)
> +{
> + /* Casting const away here is safe, because we only read from progress
> + * and will leave pro_c in the same state upon leaving the function
> + * as it had at the beginning. */
> + ThreadProgress *pro = (ThreadProgress*)pro_c;
> +
> + if (atomic_load_explicit(&pro->progress, memory_order_acquire) >= n)
> + return;
> +
> + pthread_mutex_lock(&pro->progress_mutex);
> + while (atomic_load_explicit(&pro->progress, memory_order_relaxed) < n)
> + pthread_cond_wait(&pro->progress_cond, &pro->progress_mutex);
> + pthread_mutex_unlock(&pro->progress_mutex);
> +}
> diff --git a/libavcodec/threadprogress.h b/libavcodec/threadprogress.h
> new file mode 100644
> index 0000000000..786802cbf1
> --- /dev/null
> +++ b/libavcodec/threadprogress.h
> @@ -0,0 +1,91 @@
> +/*
> + * Copyright (c) 2022 Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
> + *
> + * This file is part of FFmpeg.
> + *
> + * 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.
> + *
> + * 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 FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + */
> +
> +#ifndef AVCODEC_THREADPROGRESS_H
> +#define AVCODEC_THREADPROGRESS_H
> +
> +/**
> + * ThreadProgress is an API to easily notify other threads about progress
> + * of any kind as long as it can be packaged into an int and is consistent
> + * with the natural ordering of integers.
> + *
> + * Each initialized ThreadProgress can be in one of two modes: No-op mode
> + * or ordinary mode. In the former mode, ff_thread_report_progress() and
> + * ff_thread_await_progress() are no-ops to simply support usecases like
> + * non-frame-threading. Only in the latter case perform these functions
> + * what their name already implies.
> + */
> +
> +#include <limits.h>
> +#include <stdatomic.h>
> +#include "libavutil/thread.h"
> +
> +/**
> + * This struct should be treated as opaque by users.
> + */
> +typedef struct ThreadProgress {
> + atomic_int progress;
> + unsigned init;
> + AVMutex progress_mutex;
> + AVCond progress_cond;
> +} ThreadProgress;
> +
> +/**
> + * Initialize a ThreadProgress.
> + *
> + * @param init_mode If zero, the ThreadProgress will be initialized so that
> + * to be in no-op mode as described above. Otherwise
> + */
> +int ff_thread_progress_init(ThreadProgress *pro, int init_mode);
> +
> +/**
> + * Destroy a ThreadProgress. Can be called on a ThreadProgress that
> + * has never been initialized provided that the ThreadProgress struct
> + * has been initially zeroed. Must be called even if ff_thread_progress_init()
> + * failed.
> + */
> +void ff_thread_progress_destroy(ThreadProgress *pro);
> +
> +/**
> + * Reset the ThreadProgress's progress progress counter; must only be called
> + * if it is not in use in any way (e.g. no thread may wait on it via
> + * ff_thread_progress_await()).
> + */
> +static inline void ff_thread_progress_reset(ThreadProgress *pro)
> +{
> + atomic_init(&pro->progress, pro->init ? -1 : INT_MAX);
> +}
> +
> +/**
> + * This function is a no-op in no-op mode; otherwise it notifies
> + * other threads that a certain level of progress has been reached.
> + * Later calls with lower values of progress have no effect.
> + */
> +void ff_thread_progress_report(ThreadProgress *pro, int progress);
> +
> +/**
> + * This function is a no-op in no-op mode; otherwise it waits
> + * until other threads have reached a certain level of progress:
> + * This function will return after another thread has called
> + * ff_thread_progress_report() with the same or higher value for progress.
> + */
> +void ff_thread_progress_await(const ThreadProgress *pro, int progress);
> +
> +#endif /* AVCODEC_THREADPROGRESS_H */
Will apply this patchset tomorrow unless there are objections.
- Andreas
_______________________________________________
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:[~2024-04-18 20:40 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-04-08 19:51 Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 02/27] avcodec/decode: Add new ProgressFrame API Andreas Rheinhardt
2024-04-09 1:40 ` Michael Niedermayer
2024-04-09 6:35 ` Andreas Rheinhardt
2024-04-10 7:01 ` Anton Khirnov
2024-04-10 7:09 ` Andreas Rheinhardt
2024-04-10 7:59 ` Anton Khirnov
2024-04-10 8:02 ` Andreas Rheinhardt
2024-04-10 8:06 ` Anton Khirnov
2024-04-10 8:09 ` Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 03/27] avcodec/mimic: Switch to ProgressFrames Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 04/27] avcodec/vp3: " Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 05/27] avcodec/vp9: " Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 06/27] avcodec/vp9: Fix race when attaching side-data for show-existing frame Andreas Rheinhardt
2024-04-10 7:06 ` Anton Khirnov
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 07/27] avcodec/vp9: Reduce wait times Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 08/27] avcodec/vp9: Simplify replacing VP9Frame Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 09/27] avcodec/vp9: Replace atomic_store() by atomic_init() Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 10/27] avcodec/pthread_frame: Add API to share RefStruct refs just once Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 11/27] avcodec/wavpack: Use ThreadProgress API Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 12/27] avcodec/wavpack: Move initializing DSD data to a better place Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 13/27] avcodec/wavpack: Only reset DSD context upon parameter change Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 14/27] avcodec/wavpack: Optimize always-false comparison away Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 15/27] avcodec/wavpack: Move transient variable from context to stack Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 16/27] avcodec/vp8: Convert to ProgressFrame API Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 17/27] avcodec/vp8: Mark flushing functions as av_cold Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 18/27] avcodec/codec_internal: Remove FF_CODEC_CAP_ALLOCATE_PROGRESS Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 19/27] avcodec/hevcdec: Switch to ProgressFrames Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 20/27] avcodec/pngdec: " Andreas Rheinhardt
2024-04-08 20:13 ` [FFmpeg-devel] [PATCH v2 21/27] avcodec/ffv1dec: " Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 22/27] avcodec/qsv: Use RefStruct API for memory id (mids) array Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 23/27] avcodec/rkmppdec: Fix double-free on error Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 24/27] avcodec/rkmppdec: Check av_buffer_ref() Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 25/27] avcodec/rkmppdec: Use RefStruct API for references to decoder itself Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 26/27] avcodec/rkmppdec: Allocate AVDRMFrameDescriptor and frame ctx jointly Andreas Rheinhardt
2024-04-08 20:14 ` [FFmpeg-devel] [PATCH v2 27/27] avcodec/v4l2_(m2m|buffers): Use RefStruct API for context references Andreas Rheinhardt
2024-04-09 6:33 ` [FFmpeg-devel] [PATCH v3 01/27] avcodec/threadprogress: Add new API for frame-threaded progress Andreas Rheinhardt
2024-04-10 6:38 ` [FFmpeg-devel] [PATCH v2 " Anton Khirnov
2024-04-18 20:40 ` Andreas Rheinhardt [this message]
2024-04-18 21:30 ` epirat07
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=AS8P250MB074431726EC8F7F9E82625018F0E2@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM \
--to=andreas.rheinhardt@outlook.com \
--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