Compare commits

...

69 Commits

Author SHA1 Message Date
dec05eba
953953eac7 m 2024-11-05 01:45:15 +01:00
dec05eba
10fe4f6c51 4.2.5 2024-11-05 01:09:25 +01:00
dec05eba
aac9b9cde7 Fix incorrect bitrate calculation for constant bitrate 2024-11-05 01:08:56 +01:00
dec05eba
258f690a89 Update issues link 2024-11-01 09:18:27 +01:00
dec05eba
ea139ab9bb Mention opensuse 2024-10-28 23:05:56 +01:00
dec05eba
05c74e5a77 Mention the new UI 2024-10-27 23:11:11 +01:00
dec05eba
07ce374201 Fix cursor position when scaling output with multiple monitors on monitor capture on x11 2024-10-26 23:55:41 +02:00
dec05eba
112640282d Add option to change output resolution (-s) 2024-10-26 20:58:00 +02:00
dec05eba
5ffa725367 Add unofficial sources list 2024-10-24 23:31:33 +02:00
dec05eba
a03463082f Revert "Debug output temp"
This reverts commit 86c57c85ac.
2024-10-24 15:26:47 +02:00
dec05eba
86c57c85ac Debug output temp 2024-10-24 15:11:47 +02:00
dec05eba
1aae305a36 Revert "Temp debug output"
This reverts commit 919c87768e.
2024-10-24 15:08:43 +02:00
dec05eba
919c87768e Temp debug output 2024-10-24 14:55:39 +02:00
dec05eba
ff77c64309 Mesa && AMD check 2024-10-23 22:01:03 +02:00
dec05eba
293ec193ba Support old amd gpus that need radeon driver 2024-10-23 21:26:05 +02:00
dec05eba
422f214283 4.2.4 2024-10-22 18:47:23 +02:00
dec05eba
768f1bd61e Fix shadow variable 2024-10-22 18:45:18 +02:00
dec05eba
7428f9ffee Allow recording virtual audio devices that have forward slash (/) in the name 2024-10-22 18:39:01 +02:00
dec05eba
10165977f6 4.2.3 2024-10-19 00:48:53 +02:00
dec05eba
62509a4b60 nixos fix: look for gsr-kms-server in PATH if not found in same directory as gpu-screen-recorder 2024-10-19 00:48:04 +02:00
dec05eba
eb602b9b53 4.2.2: AMD IS SHIT 2024-10-16 01:07:58 +02:00
dec05eba
19d77b8a96 Enable low latency encoding mode on amd (only available on amd since mesa version mesa 24.1.4) 2024-10-16 00:35:18 +02:00
dec05eba
601219febb README kdenlive 2024-10-14 14:22:37 +02:00
dec05eba
3e199f241a 4.2.1 2024-10-12 12:20:23 +02:00
dec05eba
a5895b775d appendum 2024-10-12 11:58:56 +02:00
dec05eba
2c4b077f43 Better frametimes when recording is too slow 2024-10-12 11:49:31 +02:00
dec05eba
1e6866236d Fix build 2024-10-12 01:47:11 +02:00
dec05eba
551f58fca1 Disable kms client check for now 2024-10-12 01:40:12 +02:00
dec05eba
f260934667 Version 4.2.0 2024-10-12 01:25:35 +02:00
dec05eba
5172a460a9 Update TODO 2024-10-12 01:16:21 +02:00
dec05eba
9248663666 Better frame timing when frame encoding time overflows 2024-10-12 01:12:15 +02:00
dec05eba
fff68b8a18 More frame timing 2024-10-12 00:59:40 +02:00
dec05eba
037f5b3e5b Add back better frame timing for smoother video 2024-10-11 23:52:54 +02:00
dec05eba
935a7b892e Rename video encoder cuda to video encoder nvenc 2024-10-09 18:32:26 +02:00
dec05eba
0a8d176ab6 Usage 2024-10-07 23:28:20 +02:00
dec05eba
35c98148f3 usage 2024-10-07 23:26:31 +02:00
dec05eba
5ba8c6933e m 2024-10-07 22:31:24 +02:00
dec05eba
5d7933b0c9 Replace -bitrate with -q for constant bitrate (-bm cbr) 2024-10-07 22:30:11 +02:00
dec05eba
67bb47a355 More constant bitrate 2024-10-07 18:19:53 +02:00
dec05eba
ef6d265c85 Make -ab and -vb in kbps 2024-10-07 01:09:25 +02:00
dec05eba
d17bc3912d Add constant bitrate mode (-bm cbr with and -vb option) 2024-10-07 00:56:24 +02:00
dec05eba
28ba2e8f3e gsr-kms-server 'security': only allow gpu-screen-recorder to get framebuffer 2024-10-06 22:26:47 +02:00
dec05eba
29a21ffff0 Move some TODO from README to TODO file 2024-10-06 02:25:04 +02:00
dec05eba
a11c65f7bf Update TODO 2024-10-06 02:24:08 +02:00
dec05eba
793d4839f5 Support hevc vulkan, fix vulkan encoding for all resolutions, disable vulkan encoding for now until drivers and ffmpeg work properly 2024-10-03 12:42:17 +02:00
dec05eba
c74c89e028 Proper error handling for fd cleanup in vaapi query 2024-10-02 18:16:53 +02:00
dec05eba
fdc4bfe0d2 no vulkan prototypes 2024-10-02 17:53:46 +02:00
dec05eba
ecf4876654 vulkan-headers as dependency 2024-10-02 17:50:13 +02:00
dec05eba
07a843747c Comment out vulkan includes for now 2024-10-01 23:10:20 +02:00
dec05eba
4f709f0fa6 Rename codec query cuda to codec query nvenc 2024-10-01 18:30:11 +02:00
dec05eba
720a4e2919 Support gpus that only support low power encoding mode (Intel Arc A380) 2024-10-01 18:28:06 +02:00
dec05eba
0eb9ce003d Cleanup rc 2024-09-29 04:56:25 +02:00
dec05eba
ebc8c69bac Add vulkan video encoding option (-k h264_vulkan). WIP, not fully hardware accelerated yet 2024-09-27 03:03:09 +02:00
dec05eba
412cf0dbae Revert "Temp fix for nvfbc"
This reverts commit 8ab1150fb8.
2024-09-26 18:15:15 +02:00
dec05eba
8ab1150fb8 Temp fix for nvfbc 2024-09-26 17:53:08 +02:00
dec05eba
ddbb2f63b7 Revert 2024-09-26 17:49:30 +02:00
dec05eba
51ee40f85d Revert "Better glx config selection"
This reverts commit 69c3102465.
2024-09-26 17:47:43 +02:00
dec05eba
77fea88fe6 debug 2024-09-26 17:37:00 +02:00
dec05eba
69c3102465 Better glx config selection 2024-09-26 17:35:16 +02:00
dec05eba
2a87561d34 nvfbc fix test 2024-09-26 16:56:55 +02:00
dec05eba
52bd102a26 Revert "temp debug output"
This reverts commit 97998f1bab.
2024-09-26 16:56:36 +02:00
dec05eba
97998f1bab temp debug output 2024-09-26 16:41:55 +02:00
dec05eba
1aaa26d87e Move codec query from encoder to separate file 2024-09-26 16:08:26 +02:00
dec05eba
0b20a46e58 Name capture/encoder variable self 2024-09-26 02:36:31 +02:00
dec05eba
aa2fa1e17e Add template for vulkan video encoding 2024-09-26 02:23:56 +02:00
dec05eba
2b60c02181 m 2024-09-24 22:46:31 +02:00
dec05eba
1b4ccce170 minor readme 2024-09-24 22:45:33 +02:00
dec05eba
3bcc7aa439 Update about link in readme 2024-09-24 22:44:09 +02:00
dec05eba
5863fda414 Comments 2024-09-22 19:58:54 +02:00
42 changed files with 2326 additions and 1215 deletions

View File

@@ -7,6 +7,8 @@ similar to shadowplay on windows. This is the fastest screen recording tool for
This screen recorder can be used for recording your desktop offline, for live streaming and for nvidia shadowplay-like instant replay,
where only the last few minutes are saved.
This is a cli-only tool, if you want an UI for this check out [GPU Screen Recorder GTK](https://git.dec05eba.com/gpu-screen-recorder-gtk/) or if you prefer a ShadowPlay-like UI then check out [GPU Screen Recorder UI](https://git.dec05eba.com/gpu-screen-recorder-ui/).
Supported video codecs:
* H264 (default)
* HEVC (Optionally with HDR)
@@ -46,14 +48,22 @@ This should work fine on AMD/Intel X11 or Wayland. On Nvidia X11 G-SYNC only wor
For example it can cause your computer to freeze when recording certain games.
# Installation
If you are running an Arch Linux based distro, then you can find gpu screen recorder on aur under the name gpu-screen-recorder-git (`yay -S gpu-screen-recorder-git`).\
If you are running an Arch Linux based distro then you can find gpu screen recorder on aur under the name gpu-screen-recorder-git (`yay -S gpu-screen-recorder-git`).\
If you are running another distro then you can run `sudo ./install.sh`, but you need to manually install the dependencies, as described below.\
You can also install gpu screen recorder ([the gtk gui version](https://git.dec05eba.com/gpu-screen-recorder-gtk/)) from [flathub](https://flathub.org/apps/details/com.dec05eba.gpu_screen_recorder), which is the easiest method
to install GPU Screen Recorder on non-arch based distros.\
The only official ways to install GPU Screen Recorder is either from source, AUR or flathub. If you install GPU Screen Recorder from somewhere else and have an issue then try installing it
from one of the official sources before reporting it as an issue.
If you install GPU Screen Recorder flatpak, which is the gtk gui version then you can still run GPU Screen Recorder command line by using the flatpak command option, for example `flatpak run --command=gpu-screen-recorder com.dec05eba.gpu_screen_recorder -w screen -f 60 -o video.mp4`. Note that if you want to record your monitor on AMD/Intel then you need to install the flatpak system-wide (like so: `flatpak install flathub --system com.dec05eba.gpu_screen_recorder`).
## Unofficial install methods
The only official ways to install GPU Screen Recorder is either from source, AUR or flathub. Other sources may be out of date and missing features or may not work correctly.\
If you install GPU Screen Recorder from somewhere else and have an issue then try installing it from one of the official sources before reporting it as an issue.\
If you still prefer to install GPU Screen Recorder with a package manager instead of from source or as a flatpak then you may be able to find a package for your distro.\
Here are some known unofficial packages:
* Debian/Ubuntu: [Pacstall](https://pacstall.dev/packages/gpu-screen-recorder)
* Nix: [NixOS wiki](https://wiki.nixos.org/wiki/Gpu-screen-recorder)
* openSUSE: [openSUSE software repository](https://software.opensuse.org/package/gpu-screen-recorder)
* Gentoo: [Guru](https://github.com/gentoo/guru/blob/master/media-video/gpu-screen-recorder/gpu-screen-recorder-9999.ebuild)
# Dependencies
GPU Screen Recorder uses meson build system so you need to install `meson` to build GPU Screen Recorder.
@@ -61,6 +71,7 @@ GPU Screen Recorder uses meson build system so you need to install `meson` to bu
These are the dependencies needed to build GPU Screen Recorder:
* libglvnd (which provides libgl, libglx and libegl)
* vulkan-headers
* ffmpeg (libavcodec, libavformat, libavutil, libswresample, libavfilter)
* x11 (libx11, libxcomposite, libxrandr, libxfixes, libxdamage)
* libpulse
@@ -93,7 +104,9 @@ When compiling GPU Screen Recorder with portal support (`-Dportal=true`, which i
* libpipewire (and libspa which is usually part of libpipewire)
# How to use
Run `gpu-screen-recorder --help` to see all options and also examples.
Run `gpu-screen-recorder --help` to see all options and also examples.\
There is also a gui for the gpu screen recorder called [GPU Screen Recorder GTK](https://git.dec05eba.com/gpu-screen-recorder-gtk/).\
There is also a new alternative UI for GPU Screen Recorder in the style of ShadowPlay called [GPU Screen Recorder UI](https://git.dec05eba.com/gpu-screen-recorder-ui/).
## Recording
Here is an example of how to record your monitor and the default audio output: `gpu-screen-recorder -w screen -f 60 -a default_output -o ~/Videos/test_video.mp4`.
Yyou can stop and save the recording with `Ctrl+C` or by running `killall -SIGINT gpu-screen-recorder`.
@@ -111,15 +124,14 @@ The replay buffer is stored in ram (as encoded video), so don't use a too large
To save a video in replay mode, you need to send signal SIGUSR1 to gpu screen recorder. You can do this by running `killall -SIGUSR1 gpu-screen-recorder`.\
To stop recording send SIGINT to gpu screen recorder. You can do this by running `killall -SIGINT gpu-screen-recorder` or pressing `Ctrl-C` in the terminal that runs gpu screen recorder. When recording a regular non-replay video this will also save the video.\
To pause/unpause recording send SIGUSR2 to gpu screen recorder. You can do this by running `killall -SIGUSR2 gpu-screen-recorder`. This is only applicable and useful when recording (not streaming nor replay).\
## Finding audio device name
You can find the default output audio device (headset, speakers (in other words, desktop audio)) with the command `pactl get-default-sink`. Add `monitor` to the end of that to use that as an audio input in gpu screen recorder.\
You can find the default input audio device (microphone) with the command `pactl get-default-source`. This input should not have `monitor` added to the end when used in gpu screen recorder.\
Example of recording both desktop audio and microphone: `gpu-screen-recorder -w screen -f 60 -a "$(pactl get-default-sink).monitor" -a "$(pactl get-default-source)" -o ~/Videos/test_video.mp4`.\
A name (that is visible to pipewire) can be given to an audio input device by prefixing the audio input with `<name>/`, for example `dummy/$(pactl get-default-sink).monitor`.\
Note that if you use multiple audio inputs then they are each recorded into separate audio tracks in the video file. If you want to merge multiple audio inputs into one audio track then separate the audio inputs by "|" in one -a argument,
for example `-a "$(pactl get-default-sink).monitor|$(pactl get-default-source)"`.
## Audio device name
To record the default output device (desktop audio) you can use the `default_output` option, for example `-a default_output`.\
To record the default input device (microphone) you can use the `default_input` option, for example `-a default_input`.\
To list all available audio devices run `gpu-screen-recorder --list-audio-devices`. The name to use with GPU Screen Recorder will be on the left side and the human readable name is on the right side.\
To record multiple audio devices to multiple audio tracks specify the `-a` option multiple times, for example `-a default_output -a default_input`.\
To record multiple audio devices into one audio track (merged) specify the `-a` option once split with `|` for each audio device, for example `-a "default_output|default_input"`.\
In wireplumber the name of the audio will be in the format `gsr-<audio_device>`, but you can change that name by prefixing the audio device with a name and then a forward slash, for example: `-a "name/default_output"`.
There is also a gui for the gpu screen recorder called [gpu-screen-recorder-gtk](https://git.dec05eba.com/gpu-screen-recorder-gtk/).
## Simple way to run replay without gui
Run the script `scripts/start-replay.sh` to start replay and then `scripts/save-replay.sh` to save a replay and `scripts/stop-replay.sh` to stop the replay. The videos are saved to `$HOME/Videos`.
You can use these scripts to start replay at system startup if you add `scripts/start-replay.sh` to startup (this can be done differently depending on your desktop environment / window manager) and then go into
@@ -137,9 +149,11 @@ You have to reboot your computer after installing GPU Screen Recorder for the fi
Look at the [scripts](https://git.dec05eba.com/gpu-screen-recorder/tree/scripts) directory for script examples. For example if you want to automatically save a recording/replay into a folder with the same name as the game you are recording.
# Reporting bugs
Issues are reported on this Github page: [https://github.com/dec05eba/gpu-screen-recorder-issues/issues](https://github.com/dec05eba/gpu-screen-recorder-issues/issues)
Issues are reported on this Github page: [https://github.com/dec05eba/gpu-screen-recorder-issues](https://github.com/dec05eba/gpu-screen-recorder-issues).
# Contributing patches
See [https://git.dec05eba.com/?p=about](https://git.dec05eba.com/?p=about)
See [https://git.dec05eba.com/?p=about](https://git.dec05eba.com/?p=about) for contribution steps.
# Donations
See [https://git.dec05eba.com/?p=about](https://git.dec05eba.com/?p=about) for donation options.
# Demo
[![Click here to watch a demo video on youtube](https://img.youtube.com/vi/n5tm0g01n6A/0.jpg)](https://www.youtube.com/watch?v=n5tm0g01n6A)
@@ -166,13 +180,5 @@ Try recording to an SSD and make sure it's not using NTFS file system. Also reco
You have to either record in hdr mode (-k `hevc_hdr` or -k `av1_hdr` option) to record a HDR video or record with desktop portal option (`-w portal`) to turn the HDR recording into SDR.
## GPU Screen Recorder records night light when recording in HDR mode
You can record with desktop portal option (`-w portal`) instead which ignores night light, if you are ok with recording without HDR.
# Donations
If you want to donate you can donate via bitcoin or monero.
* Bitcoin: bc1qqvuqnwrdyppf707ge27fqz2n9y9gu7lf5ypyuf
* Monero: 4An9kp2qW1C9Gah7ewv4JzcNFQ5TAX7ineGCqXWK6vQnhsGGcRpNgcn8r9EC3tMcgY7vqCKs3nSRXhejMHBaGvFdN2egYet
# TODO
* Dynamically change bitrate/resolution to match desired fps. This would be helpful when streaming for example, where the encode output speed also depends on upload speed to the streaming service.
* Implement opengl injection to capture texture. This fixes VRR without having to use NvFBC direct capture.
* Always use direct capture with NvFBC once the capture issue in mpv fullscreen has been resolved (maybe detect if direct capture fails in nvfbc and switch to non-direct recording. NvFBC says if direct capture fails).
## Kdenlive says that the video is not usable for editing because it has variable frame rate
To fix this you can either record the video in .mkv format or constant frame rate (-fm cfr).

39
TODO
View File

@@ -74,11 +74,9 @@ Make it possible to select which /dev/dri/card* to use, but that requires opengl
Test if p2 state can be worked around by using pure nvenc api and overwriting cuInit/cuCtxCreate* to not do anything. Cuda might be loaded when using nvenc but it might not be used, with certain record options? (such as h264 p5).
nvenc uses cuda when using b frames and rgb->yuv conversion, so convert the image ourselves instead.-
Mesa doesn't support global headers (AV_CODEC_FLAG_GLOBAL_HEADER) with h264... which also breaks mkv since mkv requires global header. Right now gpu screen recorder will forcefully set video codec to hevc when h264 is requested for mkv files.
Drop frames if live streaming cant keep up with target fps, or dynamically change resolution/quality.
Support low power option (does it even work with vaapi in ffmpeg??). Would be very useful for steam deck.
Support low power option.
Instead of sending a big list of drm data back to kms client, send the monitor we want to record to kms server and the server should respond with only the matching monitor, and cursor.
@@ -103,7 +101,7 @@ Investigate if there is a way to do gpu->gpu copy directly without touching syst
Go back to using pure vaapi without opengl for video encoding? rotation (transpose) can be done if its done after (rgb to yuv) color conversion.
Implement scaling and use lanczos resampling for better quality. Lanczos resampling can also be used for YUV chroma for better color quality on small text.
Use lanczos resampling for better scaling quality. Lanczos resampling can also be used for YUV chroma for better color quality on small text.
Flac is disabled because the frame sizes are too large which causes big audio/video desync.
@@ -152,15 +150,38 @@ Allow prime-run on x11 if monitor capture and the prime gpu is not nvidia.
Enable 2-pass encoding.
Add cbr option.
Restart replay/update video resolution if monitor resolution changes.
Fix pure vaapi copy on intel.
ffmpeg supports vulkan encoding now (h264!). Doesn't work on amd yet because mesa is missing VK_KHR_video_maintenance1, see https://gitlab.freedesktop.org/mesa/mesa/-/issues/11857. Test on nvidia!
Test vaapi low latency mode (setenv("AMD_DEBUG", "lowlatencyenc", true);), added in mesa 24.1.4, released on july 17, 2024. Note that this forces gpu power usage to max at all times, even when recording at 2 fps.
Use nvidia low latency options for better encoding times.
Test ideal async_depth value. Increasing async_depth also increased gpu memory usage a lot (from 100mb to 500mb when moving from async_depth 2 to 16) at 4k resolution. Setting it to 8 increases it by 200mb which might be ok.
Replace -encoder cpu with -k h264_software?
Change vp8/vp9 quality options, right now the file size is too large (for vp9 at least at very_high quality).
Support recording while in replay mode. This will be needed when enabling replay on system startup with systemd service and wanting to record a video besides that.
The harder and more bloat solution for this would be to make an IPC.
The simple solution would be to use SIGUSR2 for starting/stopping recording since SIGUSR2 is unused for replays. That would mean SIGUSR2 for pausing recording would be ignored.
It also means that the video will be created in the same directory as the replay (or have option to specify another location for that) but the filename would have to be generated automatically.
To rename the file you would have to use -sc to rename it with a script, or add an option to provide a template for the name.
Dynamically change bitrate/resolution to match desired fps. This would be helpful when streaming for example, where the encode output speed also depends on upload speed to the streaming service.
Implement opengl injection to capture texture. This fixes VRR without having to use NvFBC direct capture and also allows perfect frame timing.
Always use direct capture with NvFBC once the capture issue in mpv fullscreen has been resolved (maybe detect if direct capture fails in nvfbc and switch to non-direct recording. NvFBC says if direct capture fails).
Support ROI (AV_FRAME_DATA_REGIONS_OF_INTEREST).
Default to hevc if capture size is larger than 4096 in width or height.
Set low latency mode on vulkan encoding.
Support pipewire audio capture which also allows capturing audio from a single application. This can also be done with pulseaudio by creating a virtual sink:
pactl load-module module-combine-sink sink_name=gsr2 slaves=$(pactl get-default-sink) sink_properties=device.description="gsr"
pactl move-sink-input 2944 gsr2 # 2944 comes from 'pactl list sink-inputs'
and then record gsr2.monitor.
Or use pa_stream_set_monitor_stream, which also takes the sink-input as input. However need to track when the sink disconnects to mute and then reconnect again.
Support recording/replay/livestreaming at the same time by allowing commands to be run on an existing gpu screen recorder instance.

View File

@@ -6,6 +6,7 @@ EnvironmentFile=-%h/.config/gpu-screen-recorder.env
Environment=WINDOW=screen
Environment=CONTAINER=mp4
Environment=QUALITY=very_high
Environment=BITRATE_MODE=qp
Environment=CODEC=auto
Environment=AUDIO_CODEC=opus
Environment=AUDIO_DEVICE=default_output
@@ -18,11 +19,12 @@ Environment=COLOR_RANGE=limited
Environment=KEYINT=2
Environment=ENCODER=gpu
Environment=RESTORE_PORTAL_SESSION=yes
Environment=OUTPUT_RESOLUTION=0x0
Environment=ADDITIONAL_ARGS=
ExecStart=gpu-screen-recorder -v no -w "${WINDOW}" -c "${CONTAINER}" -q "${QUALITY}" -k "${CODEC}" -ac "${AUDIO_CODEC}" -a "${AUDIO_DEVICE}" -a "${SECONDARY_AUDIO_DEVICE}" -f "${FRAMERATE}" -r "${REPLAYDURATION}" -o "${OUTPUTDIR}" -df "${MAKEFOLDERS}" $ADDITIONAL_ARGS -cr "${COLOR_RANGE}" -keyint "${KEYINT}" -restore-portal-session "${RESTORE_PORTAL_SESSION}" -encoder "${ENCODER}"
ExecStart=gpu-screen-recorder -v no -w "${WINDOW}" -s "${OUTPUT_RESOLUTION}" -c "${CONTAINER}" -q "${QUALITY}" -k "${CODEC}" -ac "${AUDIO_CODEC}" -a "${AUDIO_DEVICE}" -a "${SECONDARY_AUDIO_DEVICE}" -f "${FRAMERATE}" -r "${REPLAYDURATION}" -o "${OUTPUTDIR}" -df "${MAKEFOLDERS}" $ADDITIONAL_ARGS -cr "${COLOR_RANGE}" -keyint "${KEYINT}" -restore-portal-session "${RESTORE_PORTAL_SESSION}" -encoder "${ENCODER}" -bm "${BITRATE_MODE}"
KillSignal=SIGINT
Restart=on-failure
RestartSec=5s
[Install]
WantedBy=default.target
WantedBy=default.target

View File

@@ -11,6 +11,7 @@ typedef struct {
bool hdr;
bool record_cursor;
int fps;
vec2i output_resolution;
} gsr_capture_kms_params;
gsr_capture* gsr_capture_kms_create(const gsr_capture_kms_params *params);

View File

@@ -15,6 +15,7 @@ typedef struct {
gsr_color_range color_range;
bool record_cursor;
bool use_software_video_encoder;
vec2i output_resolution;
} gsr_capture_nvfbc_params;
gsr_capture* gsr_capture_nvfbc_create(const gsr_capture_nvfbc_params *params);

View File

@@ -11,6 +11,7 @@ typedef struct {
bool restore_portal_session;
/* If this is set to NULL then this defaults to $XDG_CONFIG_HOME/gpu-screen-recorder/restore_token ($XDG_CONFIG_HOME defaults to $HOME/.config) */
const char *portal_session_token_filepath;
vec2i output_resolution;
} gsr_capture_portal_params;
gsr_capture* gsr_capture_portal_create(const gsr_capture_portal_params *params);

View File

@@ -8,10 +8,10 @@ typedef struct {
gsr_egl *egl;
unsigned long window;
bool follow_focused; /* If this is set then |window| is ignored */
vec2i region_size; /* This is currently only used with |follow_focused| */
gsr_color_range color_range;
bool record_cursor;
gsr_color_depth color_depth;
vec2i output_resolution;
} gsr_capture_xcomposite_params;
gsr_capture* gsr_capture_xcomposite_create(const gsr_capture_xcomposite_params *params);

View File

@@ -0,0 +1,23 @@
#ifndef GSR_CODEC_QUERY_H
#define GSR_CODEC_QUERY_H
#include <stdbool.h>
typedef struct {
bool supported;
bool low_power;
} gsr_supported_video_codec;
typedef struct {
gsr_supported_video_codec h264;
gsr_supported_video_codec hevc;
gsr_supported_video_codec hevc_hdr;
gsr_supported_video_codec hevc_10bit;
gsr_supported_video_codec av1;
gsr_supported_video_codec av1_hdr;
gsr_supported_video_codec av1_10bit;
gsr_supported_video_codec vp8;
gsr_supported_video_codec vp9;
} gsr_supported_video_codecs;
#endif /* GSR_CODEC_QUERY_H */

View File

@@ -0,0 +1,8 @@
#ifndef GSR_CODEC_QUERY_NVENC_H
#define GSR_CODEC_QUERY_NVENC_H
#include "codec_query.h"
bool gsr_get_supported_video_codecs_nvenc(gsr_supported_video_codecs *video_codecs, bool cleanup);
#endif /* GSR_CODEC_QUERY_NVENC_H */

View File

@@ -0,0 +1,8 @@
#ifndef GSR_CODEC_QUERY_VAAPI_H
#define GSR_CODEC_QUERY_VAAPI_H
#include "codec_query.h"
bool gsr_get_supported_video_codecs_vaapi(gsr_supported_video_codecs *video_codecs, const char *card_path, bool cleanup);
#endif /* GSR_CODEC_QUERY_VAAPI_H */

View File

@@ -0,0 +1,8 @@
#ifndef GSR_CODEC_QUERY_VULKAN_H
#define GSR_CODEC_QUERY_VULKAN_H
#include "codec_query.h"
bool gsr_get_supported_video_codecs_vulkan(gsr_supported_video_codecs *video_codecs, const char *card_path, bool cleanup);
#endif /* GSR_CODEC_QUERY_VULKAN_H */

View File

@@ -313,6 +313,9 @@ struct gsr_egl {
void (*glUniform2f)(int location, float v0, float v1);
void (*glDebugMessageCallback)(GLDEBUGPROC callback, const void *userParam);
void (*glScissor)(int x, int y, int width, int height);
void (*glReadPixels)(int x, int y, int width, int height, unsigned int format, unsigned int type, void *pixels);
void* (*glMapBuffer)(unsigned int target, unsigned int access);
unsigned char (*glUnmapBuffer)(unsigned int target);
};
bool gsr_egl_load(gsr_egl *self, Display *dpy, bool wayland, bool is_monitor_capture);

View File

@@ -1,16 +0,0 @@
#ifndef GSR_ENCODER_VIDEO_CUDA_H
#define GSR_ENCODER_VIDEO_CUDA_H
#include "video.h"
typedef struct gsr_egl gsr_egl;
typedef struct {
gsr_egl *egl;
bool overclock;
gsr_color_depth color_depth;
} gsr_video_encoder_cuda_params;
gsr_video_encoder* gsr_video_encoder_cuda_create(const gsr_video_encoder_cuda_params *params);
#endif /* GSR_ENCODER_VIDEO_CUDA_H */

View File

@@ -0,0 +1,16 @@
#ifndef GSR_ENCODER_VIDEO_NVENC_H
#define GSR_ENCODER_VIDEO_NVENC_H
#include "video.h"
typedef struct gsr_egl gsr_egl;
typedef struct {
gsr_egl *egl;
bool overclock;
gsr_color_depth color_depth;
} gsr_video_encoder_nvenc_params;
gsr_video_encoder* gsr_video_encoder_nvenc_create(const gsr_video_encoder_nvenc_params *params);
#endif /* GSR_ENCODER_VIDEO_NVENC_H */

View File

@@ -8,22 +8,9 @@ typedef struct gsr_video_encoder gsr_video_encoder;
typedef struct AVCodecContext AVCodecContext;
typedef struct AVFrame AVFrame;
typedef struct {
bool h264;
bool hevc;
bool hevc_hdr;
bool hevc_10bit;
bool av1;
bool av1_hdr;
bool av1_10bit;
bool vp8;
bool vp9;
} gsr_supported_video_codecs;
struct gsr_video_encoder {
gsr_supported_video_codecs (*get_supported_codecs)(gsr_video_encoder *encoder, bool cleanup);
bool (*start)(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame);
void (*copy_textures_to_frame)(gsr_video_encoder *encoder, AVFrame *frame); /* Can be NULL */
void (*copy_textures_to_frame)(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion); /* Can be NULL */
/* |textures| should be able to fit 2 elements */
void (*get_textures)(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color);
void (*destroy)(gsr_video_encoder *encoder, AVCodecContext *video_codec_context);
@@ -32,9 +19,8 @@ struct gsr_video_encoder {
bool started;
};
gsr_supported_video_codecs gsr_video_encoder_get_supported_codecs(gsr_video_encoder *encoder, bool cleanup);
bool gsr_video_encoder_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame);
void gsr_video_encoder_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame);
void gsr_video_encoder_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion);
void gsr_video_encoder_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color);
void gsr_video_encoder_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context);

View File

@@ -0,0 +1,15 @@
#ifndef GSR_ENCODER_VIDEO_VULKAN_H
#define GSR_ENCODER_VIDEO_VULKAN_H
#include "video.h"
typedef struct gsr_egl gsr_egl;
typedef struct {
gsr_egl *egl;
gsr_color_depth color_depth;
} gsr_video_encoder_vulkan_params;
gsr_video_encoder* gsr_video_encoder_vulkan_create(const gsr_video_encoder_vulkan_params *params);
#endif /* GSR_ENCODER_VIDEO_VULKAN_H */

View File

@@ -51,4 +51,6 @@ bool video_codec_context_is_vaapi(AVCodecContext *video_codec_context);
bool vaapi_copy_drm_planes_to_video_surface(AVCodecContext *video_codec_context, AVFrame *video_frame, vec2i source_pos, vec2i source_size, vec2i dest_pos, vec2i dest_size, uint32_t format, vec2i size, const int *fds, const uint32_t *offsets, const uint32_t *pitches, const uint64_t *modifiers, int num_planes);
bool vaapi_copy_egl_image_to_video_surface(gsr_egl *egl, EGLImage image, vec2i source_pos, vec2i source_size, vec2i dest_pos, vec2i dest_size, AVCodecContext *video_codec_context, AVFrame *video_frame);
vec2i scale_keep_aspect_ratio(vec2i from, vec2i to);
#endif /* GSR_UTILS_H */

View File

@@ -9,4 +9,8 @@ typedef struct {
float x, y;
} vec2f;
typedef struct {
double x, y;
} vec2d;
#endif /* VEC2_H */

View File

@@ -146,13 +146,41 @@ static bool create_socket_path(char *output_path, size_t output_path_size) {
return true;
}
static void string_copy(char *dst, const char *src, int len) {
int src_len = strlen(src);
int min_len = src_len;
if(len - 1 < min_len)
min_len = len - 1;
memcpy(dst, src, min_len);
dst[min_len] = '\0';
static bool readlink_realpath(const char *filepath, char *buffer) {
char symlinked_path[PATH_MAX];
ssize_t bytes_written = readlink(filepath, symlinked_path, sizeof(symlinked_path) - 1);
if(bytes_written == -1 && errno == EINVAL) {
/* Not a symlink */
snprintf(symlinked_path, sizeof(symlinked_path), "%s", filepath);
} else if(bytes_written == -1) {
return false;
} else {
symlinked_path[bytes_written] = '\0';
}
if(!realpath(symlinked_path, buffer))
return false;
return true;
}
static bool strcat_safe(char *str, int size, const char *str_to_add) {
const int str_len = strlen(str);
const int str_to_add_len = strlen(str_to_add);
if(str_len + str_to_add_len + 1 >= size)
return false;
memcpy(str + str_len, str_to_add, str_to_add_len);
str[str_len + str_to_add_len] = '\0';
return true;
}
static void file_get_directory(char *filepath) {
char *end = strrchr(filepath, '/');
if(end == NULL)
filepath[0] = '\0';
else
*end = '\0';
}
static bool find_program_in_path(const char *program_name, char *filepath, int filepath_len) {
@@ -206,10 +234,26 @@ int gsr_kms_client_init(gsr_kms_client *self, const char *card_path) {
}
char server_filepath[PATH_MAX];
if(!find_program_in_path("gsr-kms-server", server_filepath, sizeof(server_filepath))) {
fprintf(stderr, "gsr error: gsr_kms_client_init: gsr-kms-server is not installed\n");
if(!readlink_realpath("/proc/self/exe", server_filepath)) {
fprintf(stderr, "gsr error: gsr_kms_client_init: failed to resolve /proc/self/exe\n");
return -1;
}
file_get_directory(server_filepath);
if(!strcat_safe(server_filepath, sizeof(server_filepath), "/gsr-kms-server")) {
fprintf(stderr, "gsr error: gsr_kms_client_init: gsr-kms-server path too long\n");
return -1;
}
if(access(server_filepath, F_OK) != 0) {
fprintf(stderr, "gsr info: gsr_kms_client_init: gsr-kms-server is not installed in the same directory as gpu-screen-recorder (%s not found), looking for gsr-kms-server in PATH instead\n", server_filepath);
if(!find_program_in_path("gsr-kms-server", server_filepath, sizeof(server_filepath)) || access(server_filepath, F_OK) != 0) {
fprintf(stderr, "gsr error: gsr_kms_client_init: gsr-kms-server was not found in PATH. Please install gpu-screen-recorder properly\n");
return -1;
}
}
fprintf(stderr, "gsr info: gsr_kms_client_init: setting up connection to %s\n", server_filepath);
const bool inside_flatpak = getenv("FLATPAK_ID") != NULL;
const char *home = getenv("HOME");
@@ -251,7 +295,7 @@ int gsr_kms_client_init(gsr_kms_client *self, const char *card_path) {
}
local_addr.sun_family = AF_UNIX;
string_copy(local_addr.sun_path, self->initial_socket_path, sizeof(local_addr.sun_path));
snprintf(local_addr.sun_path, sizeof(local_addr.sun_path), "%s", (const char*)self->initial_socket_path);
const mode_t prev_mask = umask(0000);
const int bind_res = bind(self->initial_socket_fd, (struct sockaddr*)&local_addr, sizeof(local_addr.sun_family) + strlen(local_addr.sun_path));

View File

@@ -1,3 +1,7 @@
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include "../kms_shared.h"
#include <stdio.h>
@@ -6,6 +10,7 @@
#include <stdlib.h>
#include <unistd.h>
#include <limits.h>
#include <fcntl.h>
#include <sys/socket.h>
#include <sys/un.h>
@@ -413,14 +418,80 @@ static double clock_get_monotonic_seconds(void) {
return (double)ts.tv_sec + (double)ts.tv_nsec * 0.000000001;
}
static void string_copy(char *dst, const char *src, int len) {
int src_len = strlen(src);
int min_len = src_len;
if(len - 1 < min_len)
min_len = len - 1;
memcpy(dst, src, min_len);
dst[min_len] = '\0';
}
// static bool readlink_realpath(const char *filepath, char *buffer) {
// char symlinked_path[PATH_MAX];
// ssize_t bytes_written = readlink(filepath, symlinked_path, sizeof(symlinked_path) - 1);
// if(bytes_written == -1 && errno == EINVAL) {
// /* Not a symlink */
// snprintf(symlinked_path, sizeof(symlinked_path), "%s", filepath);
// } else if(bytes_written == -1) {
// return false;
// } else {
// symlinked_path[bytes_written] = '\0';
// }
// if(!realpath(symlinked_path, buffer))
// return false;
// return true;
// }
// static void file_get_directory(char *filepath) {
// char *end = strrchr(filepath, '/');
// if(end == NULL)
// filepath[0] = '\0';
// else
// *end = '\0';
// }
// static bool string_ends_with(const char *str, const char *ends_with) {
// const int len = strlen(str);
// const int ends_with_len = strlen(ends_with);
// return len >= ends_with_len && memcmp(str + len - ends_with_len, ends_with, ends_with_len) == 0;
// }
// This is not foolproof, but the assumption is that gsr-kms-server and gpu-screen-recorder are installed in the same directory
// in a location that only the root user can write to (usually /usr/bin or /usr/local/bin) and if the client runs from that location
// and is called gpu-screen-recorder then gsr-kms-server can only be used by a malicious program if the malicious program
// had root access, to modify that program install directory.
// static bool is_remote_peer_program_gpu_screen_recorder(int socket_fd) {
// // TODO: Use SO_PEERPIDFD on kernel >= 6.5 to avoid a race condition in the /proc/<pid> check
// struct ucred cred;
// socklen_t ucred_len = sizeof(cred);
// if(getsockopt(socket_fd, SOL_SOCKET, SO_PEERCRED, &cred, &ucred_len) == -1) {
// fprintf(stderr, "kms server error: failed to get peer credentials, error: %s\n", strerror(errno));
// return false;
// }
// char self_directory[PATH_MAX];
// if(!readlink_realpath("/proc/self/exe", self_directory)) {
// fprintf(stderr, "kms server error: failed to resolve /proc/self/exe\n");
// return false;
// }
// file_get_directory(self_directory);
// char peer_directory[PATH_MAX];
// char peer_exe_path[PATH_MAX];
// snprintf(peer_exe_path, sizeof(peer_exe_path), "/proc/%d/exe", (int)cred.pid);
// if(!readlink_realpath(peer_exe_path, peer_directory)) {
// fprintf(stderr, "kms server error: failed to resolve /proc/self/exe\n");
// return false;
// }
// if(!string_ends_with(peer_directory, "/gpu-screen-recorder")) {
// fprintf(stderr, "kms server error: only gpu-screen-recorder can use gsr-kms-server. client program location is %s\n", peer_directory);
// return false;
// }
// file_get_directory(peer_directory);
// if(strcmp(self_directory, peer_directory) != 0) {
// fprintf(stderr, "kms server error: the client program is in directory %s but only programs in %s can run gsr-kms-server\n", peer_directory, self_directory);
// return false;
// }
// return true;
// }
int main(int argc, char **argv) {
int res = 0;
@@ -478,7 +549,7 @@ int main(int argc, char **argv) {
while(clock_get_monotonic_seconds() - start_time < connect_timeout_sec) {
struct sockaddr_un remote_addr = {0};
remote_addr.sun_family = AF_UNIX;
string_copy(remote_addr.sun_path, domain_socket_path, sizeof(remote_addr.sun_path));
snprintf(remote_addr.sun_path, sizeof(remote_addr.sun_path), "%s", domain_socket_path);
// TODO: Check if parent disconnected
if(connect(socket_fd, (struct sockaddr*)&remote_addr, sizeof(remote_addr.sun_family) + strlen(remote_addr.sun_path)) == -1) {
if(errno == ECONNREFUSED || errno == ENOENT) {
@@ -505,6 +576,11 @@ int main(int argc, char **argv) {
goto done;
}
// if(!is_remote_peer_program_gpu_screen_recorder(socket_fd)) {
// res = 3;
// goto done;
// }
for(;;) {
gsr_kms_request request;
request.version = 0;

View File

@@ -1,4 +1,4 @@
project('gpu-screen-recorder', ['c', 'cpp'], version : '4.1.11', default_options : ['warning_level=2'])
project('gpu-screen-recorder', ['c', 'cpp'], version : '4.2.5', default_options : ['warning_level=2'])
add_project_arguments('-Wshadow', language : ['c', 'cpp'])
if get_option('buildtype') == 'debug'
@@ -14,9 +14,13 @@ src = [
'src/capture/xcomposite.c',
'src/capture/kms.c',
'src/encoder/video/video.c',
'src/encoder/video/cuda.c',
'src/encoder/video/nvenc.c',
'src/encoder/video/vaapi.c',
'src/encoder/video/vulkan.c',
'src/encoder/video/software.c',
'src/codec_query/nvenc.c',
'src/codec_query/vaapi.c',
'src/codec_query/vulkan.c',
'src/egl.c',
'src/cuda.c',
'src/xnvctrl.c',

View File

@@ -1,11 +1,11 @@
[package]
name = "gpu-screen-recorder"
type = "executable"
version = "4.1.11"
version = "4.2.5"
platforms = ["posix"]
[config]
ignore_dirs = ["kms/server", "build"]
ignore_dirs = ["kms/server", "build", "debug-build"]
#error_on_warning = "true"
[define]
@@ -32,3 +32,4 @@ wayland-client = ">=1"
dbus-1 = ">=1"
libpipewire-0.3 = ">=1"
libspa-0.2 = ">=0"
#vulkan = ">=1"

View File

@@ -1,6 +1,6 @@
#!/bin/sh
window=$(xdotool selectwindow)
window_name=$(xdotool getwindowclassname "$window" || xdotool getwindowname "$window" || echo "Game")
window_name=$(xdotool getwindowname "$window" || xdotool getwindowclassname "$window" || echo "Game")
window_name="$(echo "$window_name" | tr '/\\' '_')"
gpu-screen-recorder -w "$window" -f 60 -a default_output -o "$HOME/Videos/recording/$window_name/$(date +"Video_%Y-%m-%d_%H-%M-%S.mp4")"

View File

@@ -4,7 +4,7 @@
# gpu-screen-recorder -w screen -f 60 -a default_output -r 60 -sc scripts/record-save-application-name.sh -c mp4 -o "$HOME/Videos"
window=$(xdotool getwindowfocus)
window_name=$(xdotool getwindowclassname "$window" || xdotool getwindowname "$window" || echo "Game")
window_name=$(xdotool getwindowname "$window" || xdotool getwindowclassname "$window" || echo "Game")
window_name="$(echo "$window_name" | tr '/\\' '_')"
video_dir="$HOME/Videos/Replays/$window_name"

View File

@@ -1,6 +1,6 @@
#!/bin/sh
window=$(xdotool selectwindow)
window_name=$(xdotool getwindowclassname "$window" || xdotool getwindowname "$window" || echo "Game")
window_name=$(xdotool getwindowname "$window" || xdotool getwindowclassname "$window" || echo "Game")
window_name="$(echo "$window_name" | tr '/\\' '_')"
gpu-screen-recorder -w "$window" -f 60 -c mkv -a default_output -r 60 -o "$HOME/Videos/Replays/$window_name"

View File

@@ -214,8 +214,15 @@ static int gsr_capture_kms_start(gsr_capture *cap, AVCodecContext *video_codec_c
/* Disable vsync */
self->params.egl->eglSwapInterval(self->params.egl->egl_display, 0);
video_codec_context->width = FFALIGN(self->capture_size.x, 2);
video_codec_context->height = FFALIGN(self->capture_size.y, 2);
if(self->params.output_resolution.x == 0 && self->params.output_resolution.y == 0) {
self->params.output_resolution = self->capture_size;
video_codec_context->width = FFALIGN(self->capture_size.x, 2);
video_codec_context->height = FFALIGN(self->capture_size.y, 2);
} else {
self->params.output_resolution = scale_keep_aspect_ratio(self->capture_size, self->params.output_resolution);
video_codec_context->width = FFALIGN(self->params.output_resolution.x, 2);
video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
}
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
@@ -429,7 +436,12 @@ static gsr_kms_response_item* find_cursor_drm_if_on_monitor(gsr_capture_kms *sel
return cursor_drm_fd;
}
static void render_drm_cursor(gsr_capture_kms *self, gsr_color_conversion *color_conversion, const gsr_kms_response_item *cursor_drm_fd, vec2i target_pos, float texture_rotation) {
static void render_drm_cursor(gsr_capture_kms *self, gsr_color_conversion *color_conversion, const gsr_kms_response_item *cursor_drm_fd, vec2i target_pos, float texture_rotation, vec2i output_size) {
const vec2d scale = {
self->capture_size.x == 0 ? 0 : (double)output_size.x / (double)self->capture_size.x,
self->capture_size.y == 0 ? 0 : (double)output_size.y / (double)self->capture_size.y
};
const bool cursor_texture_id_is_external = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_NVIDIA;
const vec2i cursor_size = {cursor_drm_fd->width, cursor_drm_fd->height};
@@ -458,6 +470,9 @@ static void render_drm_cursor(gsr_capture_kms *self, gsr_color_conversion *color
break;
}
cursor_pos.x *= scale.x;
cursor_pos.y *= scale.y;
cursor_pos.x += target_pos.x;
cursor_pos.y += target_pos.y;
@@ -487,32 +502,37 @@ static void render_drm_cursor(gsr_capture_kms *self, gsr_color_conversion *color
self->params.egl->eglDestroyImage(self->params.egl->egl_display, cursor_image);
self->params.egl->glEnable(GL_SCISSOR_TEST);
self->params.egl->glScissor(target_pos.x, target_pos.y, self->capture_size.x, self->capture_size.y);
self->params.egl->glScissor(target_pos.x, target_pos.y, output_size.x, output_size.y);
gsr_color_conversion_draw(color_conversion, self->cursor_texture_id,
cursor_pos, cursor_size,
cursor_pos, (vec2i){cursor_size.x * scale.x, cursor_size.y * scale.y},
(vec2i){0, 0}, cursor_size,
texture_rotation, cursor_texture_id_is_external);
self->params.egl->glDisable(GL_SCISSOR_TEST);
}
static void render_x11_cursor(gsr_capture_kms *self, gsr_color_conversion *color_conversion, vec2i capture_pos, vec2i target_pos) {
static void render_x11_cursor(gsr_capture_kms *self, gsr_color_conversion *color_conversion, vec2i capture_pos, vec2i target_pos, vec2i output_size) {
if(!self->x11_cursor.visible)
return;
const vec2d scale = {
self->capture_size.x == 0 ? 0 : (double)output_size.x / (double)self->capture_size.x,
self->capture_size.y == 0 ? 0 : (double)output_size.y / (double)self->capture_size.y
};
gsr_cursor_tick(&self->x11_cursor, DefaultRootWindow(self->params.egl->x11.dpy));
const vec2i cursor_pos = {
target_pos.x + self->x11_cursor.position.x - self->x11_cursor.hotspot.x - capture_pos.x,
target_pos.y + self->x11_cursor.position.y - self->x11_cursor.hotspot.y - capture_pos.y
target_pos.x + (self->x11_cursor.position.x - self->x11_cursor.hotspot.x - capture_pos.x) * scale.x,
target_pos.y + (self->x11_cursor.position.y - self->x11_cursor.hotspot.y - capture_pos.y) * scale.y
};
self->params.egl->glEnable(GL_SCISSOR_TEST);
self->params.egl->glScissor(target_pos.x, target_pos.y, self->capture_size.x, self->capture_size.y);
self->params.egl->glScissor(target_pos.x, target_pos.y, output_size.x, output_size.y);
gsr_color_conversion_draw(color_conversion, self->x11_cursor.texture_id,
cursor_pos, self->x11_cursor.size,
cursor_pos, (vec2i){self->x11_cursor.size.x * scale.x, self->x11_cursor.size.y * scale.y},
(vec2i){0, 0}, self->x11_cursor.size,
0.0f, false);
@@ -562,8 +582,12 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
" If you are experience performance problems in the video then record a single window on X11 or use portal capture option instead\n");
}
const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;
vec2i output_size = is_scaled ? self->params.output_resolution : self->capture_size;
output_size = scale_keep_aspect_ratio(self->capture_size, output_size);
const float texture_rotation = monitor_rotation_to_radians(self->monitor_rotation);
const vec2i target_pos = { max_int(0, frame->width / 2 - self->capture_size.x / 2), max_int(0, frame->height / 2 - self->capture_size.y / 2) };
const vec2i target_pos = { max_int(0, frame->width / 2 - output_size.x / 2), max_int(0, frame->height / 2 - output_size.y / 2) };
self->capture_size = rotate_capture_size_if_rotated(self, (vec2i){ drm_fd->src_w, drm_fd->src_h });
gsr_capture_kms_update_capture_size_change(self, color_conversion, target_pos, drm_fd);
@@ -586,7 +610,7 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
pitches[i] = drm_fd->dma_buf[i].pitch;
modifiers[i] = drm_fd->modifier;
}
if(!vaapi_copy_drm_planes_to_video_surface(self->video_codec_context, frame, (vec2i){capture_pos.x, capture_pos.y}, self->capture_size, target_pos, self->capture_size, drm_fd->pixel_format, (vec2i){drm_fd->width, drm_fd->height}, fds, offsets, pitches, modifiers, drm_fd->num_dma_bufs)) {
if(!vaapi_copy_drm_planes_to_video_surface(self->video_codec_context, frame, (vec2i){capture_pos.x, capture_pos.y}, self->capture_size, target_pos, output_size, drm_fd->pixel_format, (vec2i){drm_fd->width, drm_fd->height}, fds, offsets, pitches, modifiers, drm_fd->num_dma_bufs)) {
fprintf(stderr, "gsr error: gsr_capture_kms_capture: vaapi_copy_drm_planes_to_video_surface failed, falling back to opengl copy. Please report this as an issue at https://github.com/dec05eba/gpu-screen-recorder-issues\n");
self->fast_path_failed = true;
}
@@ -602,7 +626,7 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
}
gsr_color_conversion_draw(color_conversion, self->external_texture_fallback ? self->external_input_texture_id : self->input_texture_id,
target_pos, self->capture_size,
target_pos, output_size,
capture_pos, self->capture_size,
texture_rotation, self->external_texture_fallback);
}
@@ -613,9 +637,9 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
// the cursor plane is not available when the cursor is on the monitor controlled by the nvidia device.
if(self->is_x11) {
const vec2i cursor_monitor_offset = self->capture_pos;
render_x11_cursor(self, color_conversion, cursor_monitor_offset, target_pos);
render_x11_cursor(self, color_conversion, cursor_monitor_offset, target_pos, output_size);
} else if(cursor_drm_fd) {
render_drm_cursor(self, color_conversion, cursor_drm_fd, target_pos, texture_rotation);
render_drm_cursor(self, color_conversion, cursor_drm_fd, target_pos, texture_rotation, output_size);
}
}

View File

@@ -102,7 +102,7 @@ static void set_func_ptr(void **dst, void *src) {
}
static bool gsr_capture_nvfbc_load_library(gsr_capture *cap) {
gsr_capture_nvfbc *cap_nvfbc = cap->priv;
gsr_capture_nvfbc *self = cap->priv;
dlerror(); /* clear */
void *lib = dlopen("libnvidia-fbc.so.1", RTLD_LAZY);
@@ -111,23 +111,23 @@ static bool gsr_capture_nvfbc_load_library(gsr_capture *cap) {
return false;
}
set_func_ptr((void**)&cap_nvfbc->nv_fbc_create_instance, dlsym(lib, "NvFBCCreateInstance"));
if(!cap_nvfbc->nv_fbc_create_instance) {
set_func_ptr((void**)&self->nv_fbc_create_instance, dlsym(lib, "NvFBCCreateInstance"));
if(!self->nv_fbc_create_instance) {
fprintf(stderr, "gsr error: unable to resolve symbol 'NvFBCCreateInstance'\n");
dlclose(lib);
return false;
}
memset(&cap_nvfbc->nv_fbc_function_list, 0, sizeof(cap_nvfbc->nv_fbc_function_list));
cap_nvfbc->nv_fbc_function_list.dwVersion = NVFBC_VERSION;
NVFBCSTATUS status = cap_nvfbc->nv_fbc_create_instance(&cap_nvfbc->nv_fbc_function_list);
memset(&self->nv_fbc_function_list, 0, sizeof(self->nv_fbc_function_list));
self->nv_fbc_function_list.dwVersion = NVFBC_VERSION;
NVFBCSTATUS status = self->nv_fbc_create_instance(&self->nv_fbc_function_list);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: failed to create NvFBC instance (status: %d)\n", status);
dlclose(lib);
return false;
}
cap_nvfbc->library = lib;
self->library = lib;
return true;
}
@@ -153,64 +153,64 @@ static void set_vertical_sync_enabled(gsr_egl *egl, int enabled) {
fprintf(stderr, "gsr warning: setting vertical sync failed\n");
}
static void gsr_capture_nvfbc_destroy_session(gsr_capture_nvfbc *cap_nvfbc) {
if(cap_nvfbc->fbc_handle_created && cap_nvfbc->capture_session_created) {
static void gsr_capture_nvfbc_destroy_session(gsr_capture_nvfbc *self) {
if(self->fbc_handle_created && self->capture_session_created) {
NVFBC_DESTROY_CAPTURE_SESSION_PARAMS destroy_capture_params;
memset(&destroy_capture_params, 0, sizeof(destroy_capture_params));
destroy_capture_params.dwVersion = NVFBC_DESTROY_CAPTURE_SESSION_PARAMS_VER;
cap_nvfbc->nv_fbc_function_list.nvFBCDestroyCaptureSession(cap_nvfbc->nv_fbc_handle, &destroy_capture_params);
cap_nvfbc->capture_session_created = false;
self->nv_fbc_function_list.nvFBCDestroyCaptureSession(self->nv_fbc_handle, &destroy_capture_params);
self->capture_session_created = false;
}
}
static void gsr_capture_nvfbc_destroy_handle(gsr_capture_nvfbc *cap_nvfbc) {
if(cap_nvfbc->fbc_handle_created) {
static void gsr_capture_nvfbc_destroy_handle(gsr_capture_nvfbc *self) {
if(self->fbc_handle_created) {
NVFBC_DESTROY_HANDLE_PARAMS destroy_params;
memset(&destroy_params, 0, sizeof(destroy_params));
destroy_params.dwVersion = NVFBC_DESTROY_HANDLE_PARAMS_VER;
cap_nvfbc->nv_fbc_function_list.nvFBCDestroyHandle(cap_nvfbc->nv_fbc_handle, &destroy_params);
cap_nvfbc->fbc_handle_created = false;
cap_nvfbc->nv_fbc_handle = 0;
self->nv_fbc_function_list.nvFBCDestroyHandle(self->nv_fbc_handle, &destroy_params);
self->fbc_handle_created = false;
self->nv_fbc_handle = 0;
}
}
static void gsr_capture_nvfbc_destroy_session_and_handle(gsr_capture_nvfbc *cap_nvfbc) {
gsr_capture_nvfbc_destroy_session(cap_nvfbc);
gsr_capture_nvfbc_destroy_handle(cap_nvfbc);
static void gsr_capture_nvfbc_destroy_session_and_handle(gsr_capture_nvfbc *self) {
gsr_capture_nvfbc_destroy_session(self);
gsr_capture_nvfbc_destroy_handle(self);
}
static int gsr_capture_nvfbc_setup_handle(gsr_capture_nvfbc *cap_nvfbc) {
static int gsr_capture_nvfbc_setup_handle(gsr_capture_nvfbc *self) {
NVFBCSTATUS status;
NVFBC_CREATE_HANDLE_PARAMS create_params;
memset(&create_params, 0, sizeof(create_params));
create_params.dwVersion = NVFBC_CREATE_HANDLE_PARAMS_VER;
create_params.bExternallyManagedContext = NVFBC_TRUE;
create_params.glxCtx = cap_nvfbc->params.egl->glx_context;
create_params.glxFBConfig = cap_nvfbc->params.egl->glx_fb_config;
create_params.glxCtx = self->params.egl->glx_context;
create_params.glxFBConfig = self->params.egl->glx_fb_config;
status = cap_nvfbc->nv_fbc_function_list.nvFBCCreateHandle(&cap_nvfbc->nv_fbc_handle, &create_params);
status = self->nv_fbc_function_list.nvFBCCreateHandle(&self->nv_fbc_handle, &create_params);
if(status != NVFBC_SUCCESS) {
// Reverse engineering for interoperability
const uint8_t enable_key[] = { 0xac, 0x10, 0xc9, 0x2e, 0xa5, 0xe6, 0x87, 0x4f, 0x8f, 0x4b, 0xf4, 0x61, 0xf8, 0x56, 0x27, 0xe9 };
create_params.privateData = enable_key;
create_params.privateDataSize = 16;
status = cap_nvfbc->nv_fbc_function_list.nvFBCCreateHandle(&cap_nvfbc->nv_fbc_handle, &create_params);
status = self->nv_fbc_function_list.nvFBCCreateHandle(&self->nv_fbc_handle, &create_params);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", cap_nvfbc->nv_fbc_function_list.nvFBCGetLastErrorStr(cap_nvfbc->nv_fbc_handle));
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", self->nv_fbc_function_list.nvFBCGetLastErrorStr(self->nv_fbc_handle));
goto error_cleanup;
}
}
cap_nvfbc->fbc_handle_created = true;
self->fbc_handle_created = true;
NVFBC_GET_STATUS_PARAMS status_params;
memset(&status_params, 0, sizeof(status_params));
status_params.dwVersion = NVFBC_GET_STATUS_PARAMS_VER;
status = cap_nvfbc->nv_fbc_function_list.nvFBCGetStatus(cap_nvfbc->nv_fbc_handle, &status_params);
status = self->nv_fbc_function_list.nvFBCGetStatus(self->nv_fbc_handle, &status_params);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", cap_nvfbc->nv_fbc_function_list.nvFBCGetLastErrorStr(cap_nvfbc->nv_fbc_handle));
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", self->nv_fbc_function_list.nvFBCGetLastErrorStr(self->nv_fbc_handle));
goto error_cleanup;
}
@@ -219,10 +219,10 @@ static int gsr_capture_nvfbc_setup_handle(gsr_capture_nvfbc *cap_nvfbc) {
goto error_cleanup;
}
cap_nvfbc->tracking_width = XWidthOfScreen(DefaultScreenOfDisplay(cap_nvfbc->params.egl->x11.dpy));
cap_nvfbc->tracking_height = XHeightOfScreen(DefaultScreenOfDisplay(cap_nvfbc->params.egl->x11.dpy));
cap_nvfbc->tracking_type = strcmp(cap_nvfbc->params.display_to_capture, "screen") == 0 ? NVFBC_TRACKING_SCREEN : NVFBC_TRACKING_OUTPUT;
if(cap_nvfbc->tracking_type == NVFBC_TRACKING_OUTPUT) {
self->tracking_width = XWidthOfScreen(DefaultScreenOfDisplay(self->params.egl->x11.dpy));
self->tracking_height = XHeightOfScreen(DefaultScreenOfDisplay(self->params.egl->x11.dpy));
self->tracking_type = strcmp(self->params.display_to_capture, "screen") == 0 ? NVFBC_TRACKING_SCREEN : NVFBC_TRACKING_OUTPUT;
if(self->tracking_type == NVFBC_TRACKING_OUTPUT) {
if(!status_params.bXRandRAvailable) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: the xrandr extension is not available\n");
goto error_cleanup;
@@ -233,93 +233,98 @@ static int gsr_capture_nvfbc_setup_handle(gsr_capture_nvfbc *cap_nvfbc) {
goto error_cleanup;
}
cap_nvfbc->output_id = get_output_id_from_display_name(status_params.outputs, status_params.dwOutputNum, cap_nvfbc->params.display_to_capture, &cap_nvfbc->tracking_width, &cap_nvfbc->tracking_height);
if(cap_nvfbc->output_id == 0) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: display '%s' not found\n", cap_nvfbc->params.display_to_capture);
self->output_id = get_output_id_from_display_name(status_params.outputs, status_params.dwOutputNum, self->params.display_to_capture, &self->tracking_width, &self->tracking_height);
if(self->output_id == 0) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: display '%s' not found\n", self->params.display_to_capture);
goto error_cleanup;
}
}
if(!self->capture_region) {
self->width = self->tracking_width;
self->height = self->tracking_height;
}
return 0;
error_cleanup:
gsr_capture_nvfbc_destroy_session_and_handle(cap_nvfbc);
gsr_capture_nvfbc_destroy_session_and_handle(self);
return -1;
}
static int gsr_capture_nvfbc_setup_session(gsr_capture_nvfbc *cap_nvfbc) {
static int gsr_capture_nvfbc_setup_session(gsr_capture_nvfbc *self) {
NVFBC_CREATE_CAPTURE_SESSION_PARAMS create_capture_params;
memset(&create_capture_params, 0, sizeof(create_capture_params));
create_capture_params.dwVersion = NVFBC_CREATE_CAPTURE_SESSION_PARAMS_VER;
create_capture_params.eCaptureType = NVFBC_CAPTURE_TO_GL;
create_capture_params.bWithCursor = (!cap_nvfbc->params.direct_capture || cap_nvfbc->supports_direct_cursor) ? NVFBC_TRUE : NVFBC_FALSE;
if(!cap_nvfbc->params.record_cursor)
create_capture_params.bWithCursor = (!self->params.direct_capture || self->supports_direct_cursor) ? NVFBC_TRUE : NVFBC_FALSE;
if(!self->params.record_cursor)
create_capture_params.bWithCursor = false;
if(cap_nvfbc->capture_region)
create_capture_params.captureBox = (NVFBC_BOX){ cap_nvfbc->x, cap_nvfbc->y, cap_nvfbc->width, cap_nvfbc->height };
create_capture_params.eTrackingType = cap_nvfbc->tracking_type;
create_capture_params.dwSamplingRateMs = (uint32_t)ceilf(1000.0f / (float)cap_nvfbc->params.fps);
create_capture_params.bAllowDirectCapture = cap_nvfbc->params.direct_capture ? NVFBC_TRUE : NVFBC_FALSE;
create_capture_params.bPushModel = cap_nvfbc->params.direct_capture ? NVFBC_TRUE : NVFBC_FALSE;
if(self->capture_region)
create_capture_params.captureBox = (NVFBC_BOX){ self->x, self->y, self->width, self->height };
create_capture_params.eTrackingType = self->tracking_type;
create_capture_params.dwSamplingRateMs = (uint32_t)ceilf(1000.0f / (float)self->params.fps);
create_capture_params.bAllowDirectCapture = self->params.direct_capture ? NVFBC_TRUE : NVFBC_FALSE;
create_capture_params.bPushModel = self->params.direct_capture ? NVFBC_TRUE : NVFBC_FALSE;
create_capture_params.bDisableAutoModesetRecovery = true;
if(cap_nvfbc->tracking_type == NVFBC_TRACKING_OUTPUT)
create_capture_params.dwOutputId = cap_nvfbc->output_id;
if(self->tracking_type == NVFBC_TRACKING_OUTPUT)
create_capture_params.dwOutputId = self->output_id;
NVFBCSTATUS status = cap_nvfbc->nv_fbc_function_list.nvFBCCreateCaptureSession(cap_nvfbc->nv_fbc_handle, &create_capture_params);
NVFBCSTATUS status = self->nv_fbc_function_list.nvFBCCreateCaptureSession(self->nv_fbc_handle, &create_capture_params);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", cap_nvfbc->nv_fbc_function_list.nvFBCGetLastErrorStr(cap_nvfbc->nv_fbc_handle));
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", self->nv_fbc_function_list.nvFBCGetLastErrorStr(self->nv_fbc_handle));
return -1;
}
cap_nvfbc->capture_session_created = true;
self->capture_session_created = true;
memset(&cap_nvfbc->setup_params, 0, sizeof(cap_nvfbc->setup_params));
cap_nvfbc->setup_params.dwVersion = NVFBC_TOGL_SETUP_PARAMS_VER;
cap_nvfbc->setup_params.eBufferFormat = NVFBC_BUFFER_FORMAT_BGRA;
memset(&self->setup_params, 0, sizeof(self->setup_params));
self->setup_params.dwVersion = NVFBC_TOGL_SETUP_PARAMS_VER;
self->setup_params.eBufferFormat = NVFBC_BUFFER_FORMAT_BGRA;
status = cap_nvfbc->nv_fbc_function_list.nvFBCToGLSetUp(cap_nvfbc->nv_fbc_handle, &cap_nvfbc->setup_params);
status = self->nv_fbc_function_list.nvFBCToGLSetUp(self->nv_fbc_handle, &self->setup_params);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", cap_nvfbc->nv_fbc_function_list.nvFBCGetLastErrorStr(cap_nvfbc->nv_fbc_handle));
gsr_capture_nvfbc_destroy_session(cap_nvfbc);
fprintf(stderr, "gsr error: gsr_capture_nvfbc_start failed: %s\n", self->nv_fbc_function_list.nvFBCGetLastErrorStr(self->nv_fbc_handle));
gsr_capture_nvfbc_destroy_session(self);
return -1;
}
return 0;
}
static void gsr_capture_nvfbc_stop(gsr_capture_nvfbc *cap_nvfbc) {
gsr_capture_nvfbc_destroy_session_and_handle(cap_nvfbc);
if(cap_nvfbc->library) {
dlclose(cap_nvfbc->library);
cap_nvfbc->library = NULL;
static void gsr_capture_nvfbc_stop(gsr_capture_nvfbc *self) {
gsr_capture_nvfbc_destroy_session_and_handle(self);
if(self->library) {
dlclose(self->library);
self->library = NULL;
}
if(cap_nvfbc->params.display_to_capture) {
free((void*)cap_nvfbc->params.display_to_capture);
cap_nvfbc->params.display_to_capture = NULL;
if(self->params.display_to_capture) {
free((void*)self->params.display_to_capture);
self->params.display_to_capture = NULL;
}
}
static int gsr_capture_nvfbc_start(gsr_capture *cap, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_capture_nvfbc *cap_nvfbc = cap->priv;
gsr_capture_nvfbc *self = cap->priv;
if(!gsr_capture_nvfbc_load_library(cap))
return -1;
cap_nvfbc->x = max_int(cap_nvfbc->params.pos.x, 0);
cap_nvfbc->y = max_int(cap_nvfbc->params.pos.y, 0);
cap_nvfbc->width = max_int(cap_nvfbc->params.size.x, 0);
cap_nvfbc->height = max_int(cap_nvfbc->params.size.y, 0);
self->x = max_int(self->params.pos.x, 0);
self->y = max_int(self->params.pos.y, 0);
self->width = max_int(self->params.size.x, 0);
self->height = max_int(self->params.size.y, 0);
cap_nvfbc->capture_region = (cap_nvfbc->x > 0 || cap_nvfbc->y > 0 || cap_nvfbc->width > 0 || cap_nvfbc->height > 0);
self->capture_region = (self->x > 0 || self->y > 0 || self->width > 0 || self->height > 0);
cap_nvfbc->supports_direct_cursor = false;
self->supports_direct_cursor = false;
int driver_major_version = 0;
int driver_minor_version = 0;
if(cap_nvfbc->params.direct_capture && get_driver_version(&driver_major_version, &driver_minor_version)) {
if(self->params.direct_capture && get_driver_version(&driver_major_version, &driver_minor_version)) {
fprintf(stderr, "Info: detected nvidia version: %d.%d\n", driver_major_version, driver_minor_version);
// TODO:
if(version_at_least(driver_major_version, driver_minor_version, 515, 57) && version_less_than(driver_major_version, driver_minor_version, 520, 56)) {
cap_nvfbc->params.direct_capture = false;
self->params.direct_capture = false;
fprintf(stderr, "Warning: \"screen-direct\" has temporary been disabled as it causes stuttering with driver versions >= 515.57 and < 520.56. Please update your driver if possible. Capturing \"screen\" instead.\n");
}
@@ -328,68 +333,83 @@ static int gsr_capture_nvfbc_start(gsr_capture *cap, AVCodecContext *video_codec
/*
if(direct_capture) {
if(version_at_least(driver_major_version, driver_minor_version, 515, 57))
cap_nvfbc->supports_direct_cursor = true;
self->supports_direct_cursor = true;
else
fprintf(stderr, "Info: capturing \"screen-direct\" but driver version appears to be less than 515.57. Disabling capture of cursor. Please update your driver if you want to capture your cursor or record \"screen\" instead.\n");
}
*/
}
if(gsr_capture_nvfbc_setup_handle(cap_nvfbc) != 0) {
if(gsr_capture_nvfbc_setup_handle(self) != 0) {
goto error_cleanup;
}
if(gsr_capture_nvfbc_setup_session(cap_nvfbc) != 0) {
if(gsr_capture_nvfbc_setup_session(self) != 0) {
goto error_cleanup;
}
if(cap_nvfbc->capture_region) {
video_codec_context->width = FFALIGN(cap_nvfbc->width, 2);
video_codec_context->height = FFALIGN(cap_nvfbc->height, 2);
if(self->capture_region) {
video_codec_context->width = FFALIGN(self->width, 2);
video_codec_context->height = FFALIGN(self->height, 2);
} else {
video_codec_context->width = FFALIGN(cap_nvfbc->tracking_width, 2);
video_codec_context->height = FFALIGN(cap_nvfbc->tracking_height, 2);
video_codec_context->width = FFALIGN(self->tracking_width, 2);
video_codec_context->height = FFALIGN(self->tracking_height, 2);
}
if(self->params.output_resolution.x == 0 && self->params.output_resolution.y == 0) {
self->params.output_resolution = (vec2i){video_codec_context->width, video_codec_context->height};
} else {
self->params.output_resolution = scale_keep_aspect_ratio((vec2i){video_codec_context->width, video_codec_context->height}, self->params.output_resolution);
video_codec_context->width = FFALIGN(self->params.output_resolution.x, 2);
video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
}
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
/* Disable vsync */
set_vertical_sync_enabled(cap_nvfbc->params.egl, 0);
set_vertical_sync_enabled(self->params.egl, 0);
return 0;
error_cleanup:
gsr_capture_nvfbc_stop(cap_nvfbc);
gsr_capture_nvfbc_stop(self);
return -1;
}
static int gsr_capture_nvfbc_capture(gsr_capture *cap, AVFrame *frame, gsr_color_conversion *color_conversion) {
gsr_capture_nvfbc *cap_nvfbc = cap->priv;
gsr_capture_nvfbc *self = cap->priv;
const double nvfbc_recreate_retry_time_seconds = 1.0;
if(cap_nvfbc->nvfbc_needs_recreate) {
if(self->nvfbc_needs_recreate) {
const double now = clock_get_monotonic_seconds();
if(now - cap_nvfbc->nvfbc_dead_start >= nvfbc_recreate_retry_time_seconds) {
cap_nvfbc->nvfbc_dead_start = now;
gsr_capture_nvfbc_destroy_session_and_handle(cap_nvfbc);
if(now - self->nvfbc_dead_start >= nvfbc_recreate_retry_time_seconds) {
self->nvfbc_dead_start = now;
gsr_capture_nvfbc_destroy_session_and_handle(self);
if(gsr_capture_nvfbc_setup_handle(cap_nvfbc) != 0) {
if(gsr_capture_nvfbc_setup_handle(self) != 0) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_capture failed to recreate nvfbc handle, trying again in %f second(s)\n", nvfbc_recreate_retry_time_seconds);
return -1;
}
if(gsr_capture_nvfbc_setup_session(cap_nvfbc) != 0) {
if(gsr_capture_nvfbc_setup_session(self) != 0) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_capture failed to recreate nvfbc session, trying again in %f second(s)\n", nvfbc_recreate_retry_time_seconds);
return -1;
}
cap_nvfbc->nvfbc_needs_recreate = false;
self->nvfbc_needs_recreate = false;
} else {
return 0;
}
}
const vec2i frame_size = (vec2i){self->width, self->height};
const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;
vec2i output_size = is_scaled ? self->params.output_resolution : frame_size;
output_size = scale_keep_aspect_ratio(frame_size, output_size);
const vec2i target_pos = { max_int(0, frame->width / 2 - output_size.x / 2), max_int(0, frame->height / 2 - output_size.y / 2) };
NVFBC_FRAME_GRAB_INFO frame_info;
memset(&frame_info, 0, sizeof(frame_info));
@@ -400,24 +420,24 @@ static int gsr_capture_nvfbc_capture(gsr_capture *cap, AVFrame *frame, gsr_color
grab_params.pFrameGrabInfo = &frame_info;
grab_params.dwTimeoutMs = 0;
NVFBCSTATUS status = cap_nvfbc->nv_fbc_function_list.nvFBCToGLGrabFrame(cap_nvfbc->nv_fbc_handle, &grab_params);
NVFBCSTATUS status = self->nv_fbc_function_list.nvFBCToGLGrabFrame(self->nv_fbc_handle, &grab_params);
if(status != NVFBC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_capture_nvfbc_capture failed: %s (%d), recreating session after %f second(s)\n", cap_nvfbc->nv_fbc_function_list.nvFBCGetLastErrorStr(cap_nvfbc->nv_fbc_handle), status, nvfbc_recreate_retry_time_seconds);
cap_nvfbc->nvfbc_needs_recreate = true;
cap_nvfbc->nvfbc_dead_start = clock_get_monotonic_seconds();
fprintf(stderr, "gsr error: gsr_capture_nvfbc_capture failed: %s (%d), recreating session after %f second(s)\n", self->nv_fbc_function_list.nvFBCGetLastErrorStr(self->nv_fbc_handle), status, nvfbc_recreate_retry_time_seconds);
self->nvfbc_needs_recreate = true;
self->nvfbc_dead_start = clock_get_monotonic_seconds();
return 0;
}
cap_nvfbc->params.egl->glFlush();
cap_nvfbc->params.egl->glFinish();
self->params.egl->glFlush();
self->params.egl->glFinish();
gsr_color_conversion_draw(color_conversion, cap_nvfbc->setup_params.dwTextures[grab_params.dwTextureIndex],
(vec2i){0, 0}, (vec2i){frame->width, frame->height},
(vec2i){0, 0}, (vec2i){frame->width, frame->height},
gsr_color_conversion_draw(color_conversion, self->setup_params.dwTextures[grab_params.dwTextureIndex],
target_pos, (vec2i){output_size.x, output_size.y},
(vec2i){0, 0}, frame_size,
0.0f, false);
cap_nvfbc->params.egl->glFlush();
cap_nvfbc->params.egl->glFinish();
self->params.egl->glFlush();
self->params.egl->glFinish();
return 0;
}
@@ -429,8 +449,9 @@ static gsr_source_color gsr_capture_nvfbc_get_source_color(gsr_capture *cap) {
static void gsr_capture_nvfbc_destroy(gsr_capture *cap, AVCodecContext *video_codec_context) {
(void)video_codec_context;
gsr_capture_nvfbc *cap_nvfbc = cap->priv;
gsr_capture_nvfbc_stop(cap_nvfbc);
gsr_capture_nvfbc *self = cap->priv;
gsr_capture_nvfbc_stop(self);
free(cap->priv);
free(cap);
}
@@ -465,7 +486,7 @@ gsr_capture* gsr_capture_nvfbc_create(const gsr_capture_nvfbc_params *params) {
cap_nvfbc->params = *params;
cap_nvfbc->params.display_to_capture = display_to_capture;
cap_nvfbc->params.fps = max_int(cap_nvfbc->params.fps, 1);
*cap = (gsr_capture) {
.start = gsr_capture_nvfbc_start,
.tick = NULL,

View File

@@ -300,8 +300,15 @@ static int gsr_capture_portal_start(gsr_capture *cap, AVCodecContext *video_code
/* Disable vsync */
self->params.egl->eglSwapInterval(self->params.egl->egl_display, 0);
video_codec_context->width = FFALIGN(self->capture_size.x, 2);
video_codec_context->height = FFALIGN(self->capture_size.y, 2);
if(self->params.output_resolution.x == 0 && self->params.output_resolution.y == 0) {
self->params.output_resolution = self->capture_size;
video_codec_context->width = FFALIGN(self->capture_size.x, 2);
video_codec_context->height = FFALIGN(self->capture_size.y, 2);
} else {
self->params.output_resolution = scale_keep_aspect_ratio(self->capture_size, self->params.output_resolution);
video_codec_context->width = FFALIGN(self->params.output_resolution.x, 2);
video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
}
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
@@ -334,8 +341,12 @@ static int gsr_capture_portal_capture(gsr_capture *cap, AVFrame *frame, gsr_colo
} else {
return 0;
}
const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;
vec2i output_size = is_scaled ? self->params.output_resolution : self->capture_size;
output_size = scale_keep_aspect_ratio(self->capture_size, output_size);
const vec2i target_pos = { max_int(0, frame->width / 2 - self->capture_size.x / 2), max_int(0, frame->height / 2 - self->capture_size.y / 2) };
const vec2i target_pos = { max_int(0, frame->width / 2 - output_size.x / 2), max_int(0, frame->height / 2 - output_size.y / 2) };
self->params.egl->glFlush();
self->params.egl->glFinish();
@@ -354,7 +365,7 @@ static int gsr_capture_portal_capture(gsr_capture *cap, AVFrame *frame, gsr_colo
pitches[i] = self->dmabuf_data[i].stride;
modifiers[i] = pipewire_modifiers;
}
if(!vaapi_copy_drm_planes_to_video_surface(self->video_codec_context, frame, (vec2i){region.x, region.y}, self->capture_size, target_pos, self->capture_size, pipewire_fourcc, self->capture_size, fds, offsets, pitches, modifiers, self->num_dmabuf_data)) {
if(!vaapi_copy_drm_planes_to_video_surface(self->video_codec_context, frame, (vec2i){region.x, region.y}, self->capture_size, target_pos, output_size, pipewire_fourcc, self->capture_size, fds, offsets, pitches, modifiers, self->num_dmabuf_data)) {
fprintf(stderr, "gsr error: gsr_capture_portal_capture: vaapi_copy_drm_planes_to_video_surface failed, falling back to opengl copy. Please report this as an issue at https://github.com/dec05eba/gpu-screen-recorder-issues\n");
self->fast_path_failed = true;
}
@@ -364,21 +375,26 @@ static int gsr_capture_portal_capture(gsr_capture *cap, AVFrame *frame, gsr_colo
if(self->fast_path_failed) {
gsr_color_conversion_draw(color_conversion, using_external_image ? self->texture_map.external_texture_id : self->texture_map.texture_id,
target_pos, self->capture_size,
target_pos, output_size,
(vec2i){region.x, region.y}, self->capture_size,
0.0f, using_external_image);
}
if(self->params.record_cursor) {
if(self->params.record_cursor && self->texture_map.cursor_texture_id > 0 && cursor_region.width > 0) {
const vec2d scale = {
self->capture_size.x == 0 ? 0 : (double)output_size.x / (double)self->capture_size.x,
self->capture_size.y == 0 ? 0 : (double)output_size.y / (double)self->capture_size.y
};
const vec2i cursor_pos = {
target_pos.x + cursor_region.x,
target_pos.y + cursor_region.y
target_pos.x + (cursor_region.x * scale.x),
target_pos.y + (cursor_region.y * scale.y)
};
self->params.egl->glEnable(GL_SCISSOR_TEST);
self->params.egl->glScissor(target_pos.x, target_pos.y, self->capture_size.x, self->capture_size.y);
self->params.egl->glScissor(target_pos.x, target_pos.y, output_size.x, output_size.y);
gsr_color_conversion_draw(color_conversion, self->texture_map.cursor_texture_id,
(vec2i){cursor_pos.x, cursor_pos.y}, (vec2i){cursor_region.width, cursor_region.height},
(vec2i){cursor_pos.x, cursor_pos.y}, (vec2i){cursor_region.width * scale.x, cursor_region.height * scale.y},
(vec2i){0, 0}, (vec2i){cursor_region.width, cursor_region.height},
0.0f, false);
self->params.egl->glDisable(GL_SCISSOR_TEST);
@@ -414,9 +430,9 @@ static void gsr_capture_portal_clear_damage(gsr_capture *cap) {
static void gsr_capture_portal_destroy(gsr_capture *cap, AVCodecContext *video_codec_context) {
(void)video_codec_context;
gsr_capture_portal *cap_portal = cap->priv;
gsr_capture_portal *self = cap->priv;
if(cap->priv) {
gsr_capture_portal_stop(cap_portal);
gsr_capture_portal_stop(self);
free(cap->priv);
cap->priv = NULL;
}

View File

@@ -113,13 +113,14 @@ static int gsr_capture_xcomposite_start(gsr_capture *cap, AVCodecContext *video_
self->params.egl->glGetTexLevelParameteriv(GL_TEXTURE_2D, 0, GL_TEXTURE_HEIGHT, &self->texture_size.y);
self->params.egl->glBindTexture(GL_TEXTURE_2D, 0);
vec2i video_size = self->texture_size;
if(self->params.region_size.x > 0 && self->params.region_size.y > 0)
video_size = self->params.region_size;
video_codec_context->width = FFALIGN(video_size.x, 2);
video_codec_context->height = FFALIGN(video_size.y, 2);
if(self->params.output_resolution.x == 0 && self->params.output_resolution.y == 0) {
self->params.output_resolution = self->texture_size;
video_codec_context->width = FFALIGN(self->texture_size.x, 2);
video_codec_context->height = FFALIGN(self->texture_size.y, 2);
} else {
video_codec_context->width = FFALIGN(self->params.output_resolution.x, 2);
video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
}
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
@@ -257,14 +258,18 @@ static int gsr_capture_xcomposite_capture(gsr_capture *cap, AVFrame *frame, gsr_
gsr_color_conversion_clear(color_conversion);
}
const vec2i target_pos = { max_int(0, frame->width / 2 - self->texture_size.x / 2), max_int(0, frame->height / 2 - self->texture_size.y / 2) };
const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;
vec2i output_size = is_scaled ? self->params.output_resolution : self->texture_size;
output_size = scale_keep_aspect_ratio(self->texture_size, output_size);
const vec2i target_pos = { max_int(0, frame->width / 2 - output_size.x / 2), max_int(0, frame->height / 2 - output_size.y / 2) };
self->params.egl->glFlush();
self->params.egl->glFinish();
/* Fast opengl free path */
if(!self->fast_path_failed && video_codec_context_is_vaapi(self->video_codec_context) && self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD) {
if(!vaapi_copy_egl_image_to_video_surface(self->params.egl, self->window_texture.image, (vec2i){0, 0}, self->texture_size, target_pos, self->texture_size, self->video_codec_context, frame)) {
if(!vaapi_copy_egl_image_to_video_surface(self->params.egl, self->window_texture.image, (vec2i){0, 0}, self->texture_size, target_pos, output_size, self->video_codec_context, frame)) {
fprintf(stderr, "gsr error: gsr_capture_xcomposite_capture: vaapi_copy_egl_image_to_video_surface failed, falling back to opengl copy. Please report this as an issue at https://github.com/dec05eba/gpu-screen-recorder-issues\n");
self->fast_path_failed = true;
}
@@ -274,24 +279,29 @@ static int gsr_capture_xcomposite_capture(gsr_capture *cap, AVFrame *frame, gsr_
if(self->fast_path_failed) {
gsr_color_conversion_draw(color_conversion, window_texture_get_opengl_texture_id(&self->window_texture),
target_pos, self->texture_size,
target_pos, output_size,
(vec2i){0, 0}, self->texture_size,
0.0f, false);
}
if(self->params.record_cursor && self->cursor.visible) {
const vec2d scale = {
self->texture_size.x == 0 ? 0 : (double)output_size.x / (double)self->texture_size.x,
self->texture_size.y == 0 ? 0 : (double)output_size.y / (double)self->texture_size.y
};
gsr_cursor_tick(&self->cursor, self->window);
const vec2i cursor_pos = {
target_pos.x + self->cursor.position.x - self->cursor.hotspot.x,
target_pos.y + self->cursor.position.y - self->cursor.hotspot.y
target_pos.x + (self->cursor.position.x - self->cursor.hotspot.x) * scale.x,
target_pos.y + (self->cursor.position.y - self->cursor.hotspot.y) * scale.y
};
self->params.egl->glEnable(GL_SCISSOR_TEST);
self->params.egl->glScissor(target_pos.x, target_pos.y, self->texture_size.x, self->texture_size.y);
self->params.egl->glScissor(target_pos.x, target_pos.y, output_size.x, output_size.y);
gsr_color_conversion_draw(color_conversion, self->cursor.texture_id,
cursor_pos, self->cursor.size,
cursor_pos, (vec2i){self->cursor.size.x * scale.x, self->cursor.size.y * scale.y},
(vec2i){0, 0}, self->cursor.size,
0.0f, false);

235
src/codec_query/nvenc.c Normal file
View File

@@ -0,0 +1,235 @@
#include "../../include/codec_query/nvenc.h"
#include "../../include/cuda.h"
#include "../../external/nvEncodeAPI.h"
#include <dlfcn.h>
#include <stdio.h>
#include <string.h>
static void* open_nvenc_library(void) {
dlerror(); /* clear */
void *lib = dlopen("libnvidia-encode.so.1", RTLD_LAZY);
if(!lib) {
lib = dlopen("libnvidia-encode.so", RTLD_LAZY);
if(!lib) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc failed: failed to load libnvidia-encode.so/libnvidia-encode.so.1, error: %s\n", dlerror());
return NULL;
}
}
return lib;
}
static bool profile_is_h264(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_H264_PROFILE_BASELINE_GUID,
&NV_ENC_H264_PROFILE_MAIN_GUID,
&NV_ENC_H264_PROFILE_HIGH_GUID,
&NV_ENC_H264_PROFILE_PROGRESSIVE_HIGH_GUID,
&NV_ENC_H264_PROFILE_CONSTRAINED_HIGH_GUID
};
for(int i = 0; i < 5; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_hevc(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_HEVC_PROFILE_MAIN_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_hevc_10bit(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_HEVC_PROFILE_MAIN10_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_av1(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_AV1_PROFILE_MAIN_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool encoder_get_supported_profiles(const NV_ENCODE_API_FUNCTION_LIST *function_list, void *nvenc_encoder, const GUID *encoder_guid, gsr_supported_video_codecs *supported_video_codecs) {
bool success = false;
GUID *profile_guids = NULL;
uint32_t profile_guid_count = 0;
if(function_list->nvEncGetEncodeProfileGUIDCount(nvenc_encoder, *encoder_guid, &profile_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncGetEncodeProfileGUIDCount failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
if(profile_guid_count == 0)
goto fail;
profile_guids = calloc(profile_guid_count, sizeof(GUID));
if(!profile_guids) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: failed to allocate %d guids\n", (int)profile_guid_count);
goto fail;
}
if(function_list->nvEncGetEncodeProfileGUIDs(nvenc_encoder, *encoder_guid, profile_guids, profile_guid_count, &profile_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncGetEncodeProfileGUIDs failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
for(uint32_t i = 0; i < profile_guid_count; ++i) {
if(profile_is_h264(&profile_guids[i])) {
supported_video_codecs->h264 = (gsr_supported_video_codec){ true, false };
} else if(profile_is_hevc(&profile_guids[i])) {
supported_video_codecs->hevc = (gsr_supported_video_codec){ true, false };
} else if(profile_is_hevc_10bit(&profile_guids[i])) {
supported_video_codecs->hevc_hdr = (gsr_supported_video_codec){ true, false };
supported_video_codecs->hevc_10bit = (gsr_supported_video_codec){ true, false };
} else if(profile_is_av1(&profile_guids[i])) {
supported_video_codecs->av1 = (gsr_supported_video_codec){ true, false };
supported_video_codecs->av1_hdr = (gsr_supported_video_codec){ true, false };
supported_video_codecs->av1_10bit = (gsr_supported_video_codec){ true, false };
}
}
success = true;
fail:
if(profile_guids)
free(profile_guids);
return success;
}
static bool get_supported_video_codecs(const NV_ENCODE_API_FUNCTION_LIST *function_list, void *nvenc_encoder, gsr_supported_video_codecs *supported_video_codecs) {
bool success = false;
GUID *encoder_guids = NULL;
*supported_video_codecs = (gsr_supported_video_codecs){0};
uint32_t encode_guid_count = 0;
if(function_list->nvEncGetEncodeGUIDCount(nvenc_encoder, &encode_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncGetEncodeGUIDCount failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
if(encode_guid_count == 0)
goto fail;
encoder_guids = calloc(encode_guid_count, sizeof(GUID));
if(!encoder_guids) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: failed to allocate %d guids\n", (int)encode_guid_count);
goto fail;
}
if(function_list->nvEncGetEncodeGUIDs(nvenc_encoder, encoder_guids, encode_guid_count, &encode_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncGetEncodeGUIDs failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
for(uint32_t i = 0; i < encode_guid_count; ++i) {
encoder_get_supported_profiles(function_list, nvenc_encoder, &encoder_guids[i], supported_video_codecs);
}
success = true;
fail:
if(encoder_guids)
free(encoder_guids);
return success;
}
#define NVENCAPI_VERSION_470 (11 | (1 << 24))
#define NVENCAPI_STRUCT_VERSION_470(ver) ((uint32_t)NVENCAPI_VERSION_470 | ((ver)<<16) | (0x7 << 28))
bool gsr_get_supported_video_codecs_nvenc(gsr_supported_video_codecs *video_codecs, bool cleanup) {
memset(video_codecs, 0, sizeof(*video_codecs));
bool success = false;
void *nvenc_lib = NULL;
void *nvenc_encoder = NULL;
gsr_cuda cuda;
memset(&cuda, 0, sizeof(cuda));
if(!gsr_cuda_load(&cuda, NULL, false)) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: failed to load cuda\n");
goto done;
}
nvenc_lib = open_nvenc_library();
if(!nvenc_lib)
goto done;
typedef NVENCSTATUS NVENCAPI (*FUNC_NvEncodeAPICreateInstance)(NV_ENCODE_API_FUNCTION_LIST *functionList);
FUNC_NvEncodeAPICreateInstance nvEncodeAPICreateInstance = (FUNC_NvEncodeAPICreateInstance)dlsym(nvenc_lib, "NvEncodeAPICreateInstance");
if(!nvEncodeAPICreateInstance) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: failed to find NvEncodeAPICreateInstance in libnvidia-encode.so\n");
goto done;
}
NV_ENCODE_API_FUNCTION_LIST function_list;
memset(&function_list, 0, sizeof(function_list));
function_list.version = NVENCAPI_STRUCT_VERSION(2);
if(nvEncodeAPICreateInstance(&function_list) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncodeAPICreateInstance failed\n");
goto done;
}
NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS params;
memset(&params, 0, sizeof(params));
params.version = NVENCAPI_STRUCT_VERSION(1);
params.deviceType = NV_ENC_DEVICE_TYPE_CUDA;
params.device = cuda.cu_ctx;
params.apiVersion = NVENCAPI_VERSION;
if(function_list.nvEncOpenEncodeSessionEx(&params, &nvenc_encoder) != NV_ENC_SUCCESS) {
// Old nvidia gpus dont support the new nvenc api (which is required for av1).
// In such cases fallback to old api version if possible and try again.
function_list.version = NVENCAPI_STRUCT_VERSION_470(2);
if(nvEncodeAPICreateInstance(&function_list) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncodeAPICreateInstance (retry) failed\n");
goto done;
}
params.version = NVENCAPI_STRUCT_VERSION_470(1);
params.apiVersion = NVENCAPI_VERSION_470;
if(function_list.nvEncOpenEncodeSessionEx(&params, &nvenc_encoder) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_nvenc: nvEncOpenEncodeSessionEx (retry) failed\n");
goto done;
}
}
success = get_supported_video_codecs(&function_list, nvenc_encoder, video_codecs);
done:
if(cleanup) {
if(nvenc_encoder)
function_list.nvEncDestroyEncoder(nvenc_encoder);
if(nvenc_lib)
dlclose(nvenc_lib);
gsr_cuda_unload(&cuda);
}
return success;
}

203
src/codec_query/vaapi.c Normal file
View File

@@ -0,0 +1,203 @@
#include "../../include/codec_query/vaapi.h"
#include "../../include/utils.h"
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <fcntl.h>
#include <unistd.h>
#include <va/va.h>
#include <va/va_drm.h>
static bool profile_is_h264(VAProfile profile) {
switch(profile) {
case 5: // VAProfileH264Baseline
case VAProfileH264Main:
case VAProfileH264High:
case VAProfileH264ConstrainedBaseline:
return true;
default:
return false;
}
}
static bool profile_is_hevc_8bit(VAProfile profile) {
switch(profile) {
case VAProfileHEVCMain:
return true;
default:
return false;
}
}
static bool profile_is_hevc_10bit(VAProfile profile) {
switch(profile) {
case VAProfileHEVCMain10:
//case VAProfileHEVCMain12:
//case VAProfileHEVCMain422_10:
//case VAProfileHEVCMain422_12:
//case VAProfileHEVCMain444:
//case VAProfileHEVCMain444_10:
//case VAProfileHEVCMain444_12:
return true;
default:
return false;
}
}
static bool profile_is_av1(VAProfile profile) {
switch(profile) {
case VAProfileAV1Profile0:
case VAProfileAV1Profile1:
return true;
default:
return false;
}
}
static bool profile_is_vp8(VAProfile profile) {
switch(profile) {
case VAProfileVP8Version0_3:
return true;
default:
return false;
}
}
static bool profile_is_vp9(VAProfile profile) {
switch(profile) {
case VAProfileVP9Profile0:
case VAProfileVP9Profile1:
case VAProfileVP9Profile2:
case VAProfileVP9Profile3:
return true;
default:
return false;
}
}
static bool profile_supports_video_encoding(VADisplay va_dpy, VAProfile profile, bool *low_power) {
*low_power = false;
int num_entrypoints = vaMaxNumEntrypoints(va_dpy);
if(num_entrypoints <= 0)
return false;
VAEntrypoint *entrypoint_list = calloc(num_entrypoints, sizeof(VAEntrypoint));
if(!entrypoint_list)
return false;
bool supports_encoding = false;
bool supports_low_power_encoding = false;
if(vaQueryConfigEntrypoints(va_dpy, profile, entrypoint_list, &num_entrypoints) == VA_STATUS_SUCCESS) {
for(int i = 0; i < num_entrypoints; ++i) {
if(entrypoint_list[i] == VAEntrypointEncSlice)
supports_encoding = true;
else if(entrypoint_list[i] == VAEntrypointEncSliceLP)
supports_low_power_encoding = true;
}
}
if(!supports_encoding && supports_low_power_encoding)
*low_power = true;
free(entrypoint_list);
return supports_encoding || supports_low_power_encoding;
}
static bool get_supported_video_codecs(VADisplay va_dpy, gsr_supported_video_codecs *video_codecs, bool cleanup) {
*video_codecs = (gsr_supported_video_codecs){0};
bool success = false;
VAProfile *profile_list = NULL;
vaSetInfoCallback(va_dpy, NULL, NULL);
int va_major = 0;
int va_minor = 0;
if(vaInitialize(va_dpy, &va_major, &va_minor) != VA_STATUS_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vaapi: vaInitialize failed\n");
goto fail;
}
int num_profiles = vaMaxNumProfiles(va_dpy);
if(num_profiles <= 0)
goto fail;
profile_list = calloc(num_profiles, sizeof(VAProfile));
if(!profile_list || vaQueryConfigProfiles(va_dpy, profile_list, &num_profiles) != VA_STATUS_SUCCESS)
goto fail;
for(int i = 0; i < num_profiles; ++i) {
bool low_power = false;
if(profile_is_h264(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power)) {
video_codecs->h264 = (gsr_supported_video_codec){ true, low_power };
}
} else if(profile_is_hevc_8bit(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power))
video_codecs->hevc = (gsr_supported_video_codec){ true, low_power };
} else if(profile_is_hevc_10bit(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power)) {
video_codecs->hevc_hdr = (gsr_supported_video_codec){ true, low_power };
video_codecs->hevc_10bit = (gsr_supported_video_codec){ true, low_power };
}
} else if(profile_is_av1(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power)) {
video_codecs->av1 = (gsr_supported_video_codec){ true, low_power };
video_codecs->av1_hdr = (gsr_supported_video_codec){ true, low_power };
video_codecs->av1_10bit = (gsr_supported_video_codec){ true, low_power };
}
} else if(profile_is_vp8(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power))
video_codecs->vp8 = (gsr_supported_video_codec){ true, low_power };
} else if(profile_is_vp9(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i], &low_power))
video_codecs->vp9 = (gsr_supported_video_codec){ true, low_power };
}
}
success = true;
fail:
if(profile_list)
free(profile_list);
if(cleanup)
vaTerminate(va_dpy);
return success;
}
bool gsr_get_supported_video_codecs_vaapi(gsr_supported_video_codecs *video_codecs, const char *card_path, bool cleanup) {
memset(video_codecs, 0, sizeof(*video_codecs));
bool success = false;
int drm_fd = -1;
char render_path[128];
if(!gsr_card_path_get_render_path(card_path, render_path)) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vaapi: failed to get /dev/dri/renderDXXX file from %s\n", card_path);
goto done;
}
drm_fd = open(render_path, O_RDWR);
if(drm_fd == -1) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vaapi: failed to open device %s\n", render_path);
goto done;
}
VADisplay va_dpy = vaGetDisplayDRM(drm_fd);
if(va_dpy) {
if(!get_supported_video_codecs(va_dpy, video_codecs, cleanup)) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vaapi: failed to query supported video codecs for device %s\n", render_path);
goto done;
}
success = true;
}
done:
if(cleanup) {
if(drm_fd > 0)
close(drm_fd);
}
return success;
}

156
src/codec_query/vulkan.c Normal file
View File

@@ -0,0 +1,156 @@
#include "../../include/codec_query/vulkan.h"
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <xf86drm.h>
#define VK_NO_PROTOTYPES
//#include <vulkan/vulkan.h>
#define MAX_PHYSICAL_DEVICES 32
static const char *required_device_extensions[] = {
"VK_KHR_external_memory_fd",
"VK_KHR_external_semaphore_fd",
"VK_KHR_video_encode_queue",
"VK_KHR_video_queue",
"VK_KHR_video_maintenance1",
"VK_EXT_external_memory_dma_buf",
"VK_EXT_external_memory_host",
"VK_EXT_image_drm_format_modifier"
};
static int num_required_device_extensions = 8;
bool gsr_get_supported_video_codecs_vulkan(gsr_supported_video_codecs *video_codecs, const char *card_path, bool cleanup) {
memset(video_codecs, 0, sizeof(*video_codecs));
#if 0
bool success = false;
VkInstance instance = NULL;
VkPhysicalDevice physical_devices[MAX_PHYSICAL_DEVICES];
VkDevice device = NULL;
VkExtensionProperties *device_extensions = NULL;
const VkApplicationInfo app_info = {
.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO,
.pApplicationName = "GPU Screen Recorder",
.applicationVersion = VK_MAKE_VERSION(1, 0, 0),
.pEngineName = "GPU Screen Recorder",
.engineVersion = VK_MAKE_VERSION(1, 0, 0),
.apiVersion = VK_API_VERSION_1_3,
};
const VkInstanceCreateInfo instance_create_info = {
.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
.pApplicationInfo = &app_info
};
if(vkCreateInstance(&instance_create_info, NULL, &instance) != VK_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkCreateInstance failed\n");
goto done;
}
uint32_t num_devices = 0;
if(vkEnumeratePhysicalDevices(instance, &num_devices, NULL) != VK_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkEnumeratePhysicalDevices (query num devices) failed\n");
goto done;
}
if(num_devices == 0) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: no vulkan capable device found\n");
goto done;
}
if(num_devices > MAX_PHYSICAL_DEVICES)
num_devices = MAX_PHYSICAL_DEVICES;
if(vkEnumeratePhysicalDevices(instance, &num_devices, physical_devices) != VK_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkEnumeratePhysicalDevices (get data) failed\n");
goto done;
}
VkPhysicalDevice physical_device = NULL;
char device_card_path[128];
for(uint32_t i = 0; i < num_devices; ++i) {
VkPhysicalDeviceDrmPropertiesEXT device_drm_properties = {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRM_PROPERTIES_EXT
};
VkPhysicalDeviceProperties2 device_properties = {
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2,
.pNext = &device_drm_properties
};
vkGetPhysicalDeviceProperties2(physical_devices[i], &device_properties);
if(!device_drm_properties.hasPrimary)
continue;
snprintf(device_card_path, sizeof(device_card_path), DRM_DEV_NAME, DRM_DIR_NAME, (int)device_drm_properties.primaryMinor);
if(strcmp(device_card_path, card_path) == 0) {
physical_device = physical_devices[i];
break;
}
}
if(!physical_device) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: failed to find a vulkan device that matches opengl device %s\n", card_path);
goto done;
}
const VkDeviceCreateInfo device_create_info = {
.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO,
.enabledExtensionCount = num_required_device_extensions,
.ppEnabledExtensionNames = required_device_extensions
};
if(vkCreateDevice(physical_device, &device_create_info, NULL, &device) != VK_SUCCESS) {
//fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkCreateDevice failed. Device %s likely doesn't support vulkan video encoding\n", card_path);
goto done;
}
uint32_t num_device_extensions = 0;
if(vkEnumerateDeviceExtensionProperties(physical_device, NULL, &num_device_extensions, NULL) != VK_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkEnumerateDeviceExtensionProperties (query num device extensions) failed\n");
goto done;
}
device_extensions = calloc(num_device_extensions, sizeof(VkExtensionProperties));
if(!device_extensions) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: failed to allocate %d device extensions\n", num_device_extensions);
goto done;
}
if(vkEnumerateDeviceExtensionProperties(physical_device, NULL, &num_device_extensions, device_extensions) != VK_SUCCESS) {
fprintf(stderr, "gsr error: gsr_get_supported_video_codecs_vulkan: vkEnumerateDeviceExtensionProperties (get data) failed\n");
goto done;
}
for(uint32_t i = 0; i < num_device_extensions; ++i) {
if(strcmp(device_extensions[i].extensionName, "VK_KHR_video_encode_h264") == 0) {
video_codecs->h264 = true;
} else if(strcmp(device_extensions[i].extensionName, "VK_KHR_video_encode_h265") == 0) {
// TODO: Verify if 10bit and hdr are actually supported
video_codecs->hevc = true;
video_codecs->hevc_10bit = true;
video_codecs->hevc_hdr = true;
}
}
success = true;
done:
if(cleanup) {
if(device)
vkDestroyDevice(device, NULL);
if(instance)
vkDestroyInstance(instance, NULL);
}
if(device_extensions)
free(device_extensions);
return success;
#else
// TODO: Low power query
video_codecs->h264 = (gsr_supported_video_codec){ true, false };
video_codecs->hevc = (gsr_supported_video_codec){ true, false };
return true;
#endif
}

View File

@@ -223,7 +223,7 @@ static bool gsr_egl_create_window(gsr_egl *self, bool wayland) {
if(wayland) {
self->wayland.dpy = wl_display_connect(NULL);
if(!self->wayland.dpy) {
fprintf(stderr, "gsr error: gsr_egl_create_window failed: wl_display_connect failed\n");
fprintf(stderr, "gsr error: gsr_egl_create_window failed: failed to connect to the Wayland server\n");
goto fail;
}
@@ -262,12 +262,12 @@ static bool gsr_egl_create_window(gsr_egl *self, bool wayland) {
fprintf(stderr, "gsr error: gsr_egl_create_window failed: eglInitialize failed\n");
goto fail;
}
if(!self->eglChooseConfig(self->egl_display, attr, &ecfg, 1, &num_config) || num_config != 1) {
fprintf(stderr, "gsr error: gsr_egl_create_window failed: failed to find a matching config\n");
goto fail;
}
self->egl_context = self->eglCreateContext(self->egl_display, ecfg, NULL, ctxattr);
if(!self->egl_context) {
fprintf(stderr, "gsr error: gsr_egl_create_window failed: failed to create egl context\n");
@@ -499,6 +499,9 @@ static bool gsr_egl_load_gl(gsr_egl *self, void *library) {
{ (void**)&self->glUniform2f, "glUniform2f" },
{ (void**)&self->glDebugMessageCallback, "glDebugMessageCallback" },
{ (void**)&self->glScissor, "glScissor" },
{ (void**)&self->glReadPixels, "glReadPixels" },
{ (void**)&self->glMapBuffer, "glMapBuffer" },
{ (void**)&self->glUnmapBuffer, "glUnmapBuffer" },
{ NULL, NULL }
};
@@ -708,6 +711,10 @@ bool gsr_egl_process_event(gsr_egl *self) {
}
void gsr_egl_swap_buffers(gsr_egl *self) {
/* This uses less cpu than swap buffer on nvidia */
// TODO: Do these and remove swap
//self->glFlush();
//self->glFinish();
if(self->egl_display) {
self->eglSwapBuffers(self->egl_display, self->egl_surface);
} else if(self->x11.window) {

View File

@@ -1,468 +0,0 @@
#include "../../../include/encoder/video/cuda.h"
#include "../../../include/egl.h"
#include "../../../include/cuda.h"
#include "../../../external/nvEncodeAPI.h"
#include <libavcodec/avcodec.h>
#include <libavutil/hwcontext_cuda.h>
#include <stdlib.h>
#include <dlfcn.h>
typedef struct {
gsr_video_encoder_cuda_params params;
unsigned int target_textures[2];
AVBufferRef *device_ctx;
gsr_cuda cuda;
CUgraphicsResource cuda_graphics_resources[2];
CUarray mapped_arrays[2];
CUstream cuda_stream;
} gsr_video_encoder_cuda;
static bool gsr_video_encoder_cuda_setup_context(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context) {
self->device_ctx = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_CUDA);
if(!self->device_ctx) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hardware device context\n");
return false;
}
AVHWDeviceContext *hw_device_context = (AVHWDeviceContext*)self->device_ctx->data;
AVCUDADeviceContext *cuda_device_context = (AVCUDADeviceContext*)hw_device_context->hwctx;
cuda_device_context->cuda_ctx = self->cuda.cu_ctx;
if(av_hwdevice_ctx_init(self->device_ctx) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hardware device context\n");
av_buffer_unref(&self->device_ctx);
return false;
}
AVBufferRef *frame_context = av_hwframe_ctx_alloc(self->device_ctx);
if(!frame_context) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hwframe context\n");
av_buffer_unref(&self->device_ctx);
return false;
}
AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)frame_context->data;
hw_frame_context->width = video_codec_context->width;
hw_frame_context->height = video_codec_context->height;
hw_frame_context->sw_format = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? AV_PIX_FMT_P010LE : AV_PIX_FMT_NV12;
hw_frame_context->format = video_codec_context->pix_fmt;
hw_frame_context->device_ctx = (AVHWDeviceContext*)self->device_ctx->data;
if (av_hwframe_ctx_init(frame_context) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to initialize hardware frame context "
"(note: ffmpeg version needs to be > 4.0)\n");
av_buffer_unref(&self->device_ctx);
//av_buffer_unref(&frame_context);
return false;
}
self->cuda_stream = cuda_device_context->stream;
video_codec_context->hw_frames_ctx = av_buffer_ref(frame_context);
av_buffer_unref(&frame_context);
return true;
}
static unsigned int gl_create_texture(gsr_egl *egl, int width, int height, int internal_format, unsigned int format) {
unsigned int texture_id = 0;
egl->glGenTextures(1, &texture_id);
egl->glBindTexture(GL_TEXTURE_2D, texture_id);
egl->glTexImage2D(GL_TEXTURE_2D, 0, internal_format, width, height, 0, format, GL_UNSIGNED_BYTE, NULL);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
egl->glBindTexture(GL_TEXTURE_2D, 0);
return texture_id;
}
static bool cuda_register_opengl_texture(gsr_cuda *cuda, CUgraphicsResource *cuda_graphics_resource, CUarray *mapped_array, unsigned int texture_id) {
CUresult res;
res = cuda->cuGraphicsGLRegisterImage(cuda_graphics_resource, texture_id, GL_TEXTURE_2D, CU_GRAPHICS_REGISTER_FLAGS_NONE);
if (res != CUDA_SUCCESS) {
const char *err_str = "unknown";
cuda->cuGetErrorString(res, &err_str);
fprintf(stderr, "gsr error: cuda_register_opengl_texture: cuGraphicsGLRegisterImage failed, error: %s, texture " "id: %u\n", err_str, texture_id);
return false;
}
res = cuda->cuGraphicsResourceSetMapFlags(*cuda_graphics_resource, CU_GRAPHICS_MAP_RESOURCE_FLAGS_NONE);
res = cuda->cuGraphicsMapResources(1, cuda_graphics_resource, 0);
res = cuda->cuGraphicsSubResourceGetMappedArray(mapped_array, *cuda_graphics_resource, 0, 0);
return true;
}
static bool gsr_video_encoder_cuda_setup_textures(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context, AVFrame *frame) {
const int res = av_hwframe_get_buffer(video_codec_context->hw_frames_ctx, frame, 0);
if(res < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_textures: av_hwframe_get_buffer failed: %d\n", res);
return false;
}
const unsigned int internal_formats_nv12[2] = { GL_R8, GL_RG8 };
const unsigned int internal_formats_p010[2] = { GL_R16, GL_RG16 };
const unsigned int formats[2] = { GL_RED, GL_RG };
const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size
for(int i = 0; i < 2; ++i) {
self->target_textures[i] = gl_create_texture(self->params.egl, video_codec_context->width / div[i], video_codec_context->height / div[i], self->params.color_depth == GSR_COLOR_DEPTH_8_BITS ? internal_formats_nv12[i] : internal_formats_p010[i], formats[i]);
if(self->target_textures[i] == 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_textures: failed to create opengl texture\n");
return false;
}
if(!cuda_register_opengl_texture(&self->cuda, &self->cuda_graphics_resources[i], &self->mapped_arrays[i], self->target_textures[i])) {
return false;
}
}
return true;
}
static void* open_nvenc_library(void) {
dlerror(); /* clear */
void *lib = dlopen("libnvidia-encode.so.1", RTLD_LAZY);
if(!lib) {
lib = dlopen("libnvidia-encode.so", RTLD_LAZY);
if(!lib) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs failed: failed to load libnvidia-encode.so/libnvidia-encode.so.1, error: %s\n", dlerror());
return NULL;
}
}
return lib;
}
static bool profile_is_h264(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_H264_PROFILE_BASELINE_GUID,
&NV_ENC_H264_PROFILE_MAIN_GUID,
&NV_ENC_H264_PROFILE_HIGH_GUID,
&NV_ENC_H264_PROFILE_PROGRESSIVE_HIGH_GUID,
&NV_ENC_H264_PROFILE_CONSTRAINED_HIGH_GUID
};
for(int i = 0; i < 5; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_hevc(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_HEVC_PROFILE_MAIN_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_hevc_10bit(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_HEVC_PROFILE_MAIN10_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool profile_is_av1(const GUID *profile_guid) {
const GUID *h264_guids[] = {
&NV_ENC_AV1_PROFILE_MAIN_GUID,
};
for(int i = 0; i < 1; ++i) {
if(memcmp(profile_guid, h264_guids[i], sizeof(GUID)) == 0)
return true;
}
return false;
}
static bool encoder_get_supported_profiles(const NV_ENCODE_API_FUNCTION_LIST *function_list, void *nvenc_encoder, const GUID *encoder_guid, gsr_supported_video_codecs *supported_video_codecs) {
bool success = false;
GUID *profile_guids = NULL;
uint32_t profile_guid_count = 0;
if(function_list->nvEncGetEncodeProfileGUIDCount(nvenc_encoder, *encoder_guid, &profile_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncGetEncodeProfileGUIDCount failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
if(profile_guid_count == 0)
goto fail;
profile_guids = calloc(profile_guid_count, sizeof(GUID));
if(!profile_guids) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: failed to allocate %d guids\n", (int)profile_guid_count);
goto fail;
}
if(function_list->nvEncGetEncodeProfileGUIDs(nvenc_encoder, *encoder_guid, profile_guids, profile_guid_count, &profile_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncGetEncodeProfileGUIDs failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
for(uint32_t i = 0; i < profile_guid_count; ++i) {
if(profile_is_h264(&profile_guids[i])) {
supported_video_codecs->h264 = true;
} else if(profile_is_hevc(&profile_guids[i])) {
supported_video_codecs->hevc = true;
} else if(profile_is_hevc_10bit(&profile_guids[i])) {
supported_video_codecs->hevc_hdr = true;
supported_video_codecs->hevc_10bit = true;
} else if(profile_is_av1(&profile_guids[i])) {
supported_video_codecs->av1 = true;
supported_video_codecs->av1_hdr = true;
supported_video_codecs->av1_10bit = true;
}
}
success = true;
fail:
if(profile_guids)
free(profile_guids);
return success;
}
static bool get_supported_video_codecs(const NV_ENCODE_API_FUNCTION_LIST *function_list, void *nvenc_encoder, gsr_supported_video_codecs *supported_video_codecs) {
bool success = false;
GUID *encoder_guids = NULL;
*supported_video_codecs = (gsr_supported_video_codecs){0};
uint32_t encode_guid_count = 0;
if(function_list->nvEncGetEncodeGUIDCount(nvenc_encoder, &encode_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncGetEncodeGUIDCount failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
if(encode_guid_count == 0)
goto fail;
encoder_guids = calloc(encode_guid_count, sizeof(GUID));
if(!encoder_guids) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: failed to allocate %d guids\n", (int)encode_guid_count);
goto fail;
}
if(function_list->nvEncGetEncodeGUIDs(nvenc_encoder, encoder_guids, encode_guid_count, &encode_guid_count) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncGetEncodeGUIDs failed, error: %s\n", function_list->nvEncGetLastErrorString(nvenc_encoder));
goto fail;
}
for(uint32_t i = 0; i < encode_guid_count; ++i) {
encoder_get_supported_profiles(function_list, nvenc_encoder, &encoder_guids[i], supported_video_codecs);
}
success = true;
fail:
if(encoder_guids)
free(encoder_guids);
return success;
}
#define NVENCAPI_VERSION_470 (11 | (1 << 24))
#define NVENCAPI_STRUCT_VERSION_470(ver) ((uint32_t)NVENCAPI_VERSION_470 | ((ver)<<16) | (0x7 << 28))
static gsr_supported_video_codecs gsr_video_encoder_cuda_get_supported_codecs(gsr_video_encoder *encoder, bool cleanup) {
(void)encoder;
void *nvenc_lib = NULL;
void *nvenc_encoder = NULL;
gsr_cuda cuda;
memset(&cuda, 0, sizeof(cuda));
gsr_supported_video_codecs supported_video_codecs = {0};
if(!gsr_cuda_load(&cuda, NULL, false)) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: failed to load cuda\n");
goto done;
}
nvenc_lib = open_nvenc_library();
if(!nvenc_lib)
goto done;
typedef NVENCSTATUS NVENCAPI (*FUNC_NvEncodeAPICreateInstance)(NV_ENCODE_API_FUNCTION_LIST *functionList);
FUNC_NvEncodeAPICreateInstance nvEncodeAPICreateInstance = (FUNC_NvEncodeAPICreateInstance)dlsym(nvenc_lib, "NvEncodeAPICreateInstance");
if(!nvEncodeAPICreateInstance) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: failed to find NvEncodeAPICreateInstance in libnvidia-encode.so\n");
goto done;
}
NV_ENCODE_API_FUNCTION_LIST function_list;
memset(&function_list, 0, sizeof(function_list));
function_list.version = NVENCAPI_STRUCT_VERSION(2);
if(nvEncodeAPICreateInstance(&function_list) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncodeAPICreateInstance failed\n");
goto done;
}
NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS params;
memset(&params, 0, sizeof(params));
params.version = NVENCAPI_STRUCT_VERSION(1);
params.deviceType = NV_ENC_DEVICE_TYPE_CUDA;
params.device = cuda.cu_ctx;
params.apiVersion = NVENCAPI_VERSION;
if(function_list.nvEncOpenEncodeSessionEx(&params, &nvenc_encoder) != NV_ENC_SUCCESS) {
// Old nvidia gpus dont support the new nvenc api (which is required for av1).
// In such cases fallback to old api version if possible and try again.
function_list.version = NVENCAPI_STRUCT_VERSION_470(2);
if(nvEncodeAPICreateInstance(&function_list) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncodeAPICreateInstance (retry) failed\n");
goto done;
}
params.version = NVENCAPI_STRUCT_VERSION_470(1);
params.apiVersion = NVENCAPI_VERSION_470;
if(function_list.nvEncOpenEncodeSessionEx(&params, &nvenc_encoder) != NV_ENC_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_get_supported_codecs: nvEncOpenEncodeSessionEx (retry) failed\n");
goto done;
}
}
get_supported_video_codecs(&function_list, nvenc_encoder, &supported_video_codecs);
done:
if(cleanup) {
if(nvenc_encoder)
function_list.nvEncDestroyEncoder(nvenc_encoder);
if(nvenc_lib)
dlclose(nvenc_lib);
gsr_cuda_unload(&cuda);
}
return supported_video_codecs;
}
static void gsr_video_encoder_cuda_stop(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context);
static bool gsr_video_encoder_cuda_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_video_encoder_cuda *encoder_cuda = encoder->priv;
const bool overclock = gsr_egl_get_display_server(encoder_cuda->params.egl) == GSR_DISPLAY_SERVER_X11 ? encoder_cuda->params.overclock : false;
if(!gsr_cuda_load(&encoder_cuda->cuda, encoder_cuda->params.egl->x11.dpy, overclock)) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_start: failed to load cuda\n");
gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context);
return false;
}
if(!gsr_video_encoder_cuda_setup_context(encoder_cuda, video_codec_context)) {
gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context);
return false;
}
if(!gsr_video_encoder_cuda_setup_textures(encoder_cuda, video_codec_context, frame)) {
gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context);
return false;
}
return true;
}
void gsr_video_encoder_cuda_stop(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context) {
self->params.egl->glDeleteTextures(2, self->target_textures);
self->target_textures[0] = 0;
self->target_textures[1] = 0;
if(video_codec_context->hw_frames_ctx)
av_buffer_unref(&video_codec_context->hw_frames_ctx);
if(self->device_ctx)
av_buffer_unref(&self->device_ctx);
if(self->cuda.cu_ctx) {
for(int i = 0; i < 2; ++i) {
if(self->cuda_graphics_resources[i]) {
self->cuda.cuGraphicsUnmapResources(1, &self->cuda_graphics_resources[i], 0);
self->cuda.cuGraphicsUnregisterResource(self->cuda_graphics_resources[i]);
self->cuda_graphics_resources[i] = 0;
}
}
}
gsr_cuda_unload(&self->cuda);
}
static void gsr_video_encoder_cuda_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame) {
gsr_video_encoder_cuda *encoder_cuda = encoder->priv;
const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size
for(int i = 0; i < 2; ++i) {
CUDA_MEMCPY2D memcpy_struct;
memcpy_struct.srcXInBytes = 0;
memcpy_struct.srcY = 0;
memcpy_struct.srcMemoryType = CU_MEMORYTYPE_ARRAY;
memcpy_struct.dstXInBytes = 0;
memcpy_struct.dstY = 0;
memcpy_struct.dstMemoryType = CU_MEMORYTYPE_DEVICE;
memcpy_struct.srcArray = encoder_cuda->mapped_arrays[i];
memcpy_struct.srcPitch = frame->width / div[i];
memcpy_struct.dstDevice = (CUdeviceptr)frame->data[i];
memcpy_struct.dstPitch = frame->linesize[i];
memcpy_struct.WidthInBytes = frame->width * (encoder_cuda->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? 2 : 1);
memcpy_struct.Height = frame->height / div[i];
// TODO: Remove this copy if possible
encoder_cuda->cuda.cuMemcpy2DAsync_v2(&memcpy_struct, encoder_cuda->cuda_stream);
}
// TODO: needed?
encoder_cuda->cuda.cuStreamSynchronize(encoder_cuda->cuda_stream);
}
static void gsr_video_encoder_cuda_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {
gsr_video_encoder_cuda *encoder_cuda = encoder->priv;
textures[0] = encoder_cuda->target_textures[0];
textures[1] = encoder_cuda->target_textures[1];
*num_textures = 2;
*destination_color = encoder_cuda->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
}
static void gsr_video_encoder_cuda_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) {
gsr_video_encoder_cuda_stop(encoder->priv, video_codec_context);
free(encoder->priv);
free(encoder);
}
gsr_video_encoder* gsr_video_encoder_cuda_create(const gsr_video_encoder_cuda_params *params) {
gsr_video_encoder *encoder = calloc(1, sizeof(gsr_video_encoder));
if(!encoder)
return NULL;
gsr_video_encoder_cuda *encoder_cuda = calloc(1, sizeof(gsr_video_encoder_cuda));
if(!encoder_cuda) {
free(encoder);
return NULL;
}
encoder_cuda->params = *params;
*encoder = (gsr_video_encoder) {
.get_supported_codecs = gsr_video_encoder_cuda_get_supported_codecs,
.start = gsr_video_encoder_cuda_start,
.copy_textures_to_frame = gsr_video_encoder_cuda_copy_textures_to_frame,
.get_textures = gsr_video_encoder_cuda_get_textures,
.destroy = gsr_video_encoder_cuda_destroy,
.priv = encoder_cuda
};
return encoder;
}

237
src/encoder/video/nvenc.c Normal file
View File

@@ -0,0 +1,237 @@
#include "../../../include/encoder/video/nvenc.h"
#include "../../../include/egl.h"
#include "../../../include/cuda.h"
#include <libavcodec/avcodec.h>
#include <libavutil/hwcontext_cuda.h>
#include <stdlib.h>
typedef struct {
gsr_video_encoder_nvenc_params params;
unsigned int target_textures[2];
AVBufferRef *device_ctx;
gsr_cuda cuda;
CUgraphicsResource cuda_graphics_resources[2];
CUarray mapped_arrays[2];
CUstream cuda_stream;
} gsr_video_encoder_nvenc;
static bool gsr_video_encoder_nvenc_setup_context(gsr_video_encoder_nvenc *self, AVCodecContext *video_codec_context) {
self->device_ctx = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_CUDA);
if(!self->device_ctx) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_context failed: failed to create hardware device context\n");
return false;
}
AVHWDeviceContext *hw_device_context = (AVHWDeviceContext*)self->device_ctx->data;
AVCUDADeviceContext *cuda_device_context = (AVCUDADeviceContext*)hw_device_context->hwctx;
cuda_device_context->cuda_ctx = self->cuda.cu_ctx;
if(av_hwdevice_ctx_init(self->device_ctx) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_context failed: failed to create hardware device context\n");
av_buffer_unref(&self->device_ctx);
return false;
}
AVBufferRef *frame_context = av_hwframe_ctx_alloc(self->device_ctx);
if(!frame_context) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_context failed: failed to create hwframe context\n");
av_buffer_unref(&self->device_ctx);
return false;
}
AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)frame_context->data;
hw_frame_context->width = video_codec_context->width;
hw_frame_context->height = video_codec_context->height;
hw_frame_context->sw_format = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? AV_PIX_FMT_P010LE : AV_PIX_FMT_NV12;
hw_frame_context->format = video_codec_context->pix_fmt;
hw_frame_context->device_ctx = (AVHWDeviceContext*)self->device_ctx->data;
if (av_hwframe_ctx_init(frame_context) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_context failed: failed to initialize hardware frame context "
"(note: ffmpeg version needs to be > 4.0)\n");
av_buffer_unref(&self->device_ctx);
//av_buffer_unref(&frame_context);
return false;
}
self->cuda_stream = cuda_device_context->stream;
video_codec_context->hw_frames_ctx = av_buffer_ref(frame_context);
av_buffer_unref(&frame_context);
return true;
}
static unsigned int gl_create_texture(gsr_egl *egl, int width, int height, int internal_format, unsigned int format) {
unsigned int texture_id = 0;
egl->glGenTextures(1, &texture_id);
egl->glBindTexture(GL_TEXTURE_2D, texture_id);
egl->glTexImage2D(GL_TEXTURE_2D, 0, internal_format, width, height, 0, format, GL_UNSIGNED_BYTE, NULL);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
egl->glBindTexture(GL_TEXTURE_2D, 0);
return texture_id;
}
static bool cuda_register_opengl_texture(gsr_cuda *cuda, CUgraphicsResource *cuda_graphics_resource, CUarray *mapped_array, unsigned int texture_id) {
CUresult res;
res = cuda->cuGraphicsGLRegisterImage(cuda_graphics_resource, texture_id, GL_TEXTURE_2D, CU_GRAPHICS_REGISTER_FLAGS_NONE);
if (res != CUDA_SUCCESS) {
const char *err_str = "unknown";
cuda->cuGetErrorString(res, &err_str);
fprintf(stderr, "gsr error: cuda_register_opengl_texture: cuGraphicsGLRegisterImage failed, error: %s, texture " "id: %u\n", err_str, texture_id);
return false;
}
res = cuda->cuGraphicsResourceSetMapFlags(*cuda_graphics_resource, CU_GRAPHICS_MAP_RESOURCE_FLAGS_NONE);
res = cuda->cuGraphicsMapResources(1, cuda_graphics_resource, 0);
res = cuda->cuGraphicsSubResourceGetMappedArray(mapped_array, *cuda_graphics_resource, 0, 0);
return true;
}
static bool gsr_video_encoder_nvenc_setup_textures(gsr_video_encoder_nvenc *self, AVCodecContext *video_codec_context, AVFrame *frame) {
const int res = av_hwframe_get_buffer(video_codec_context->hw_frames_ctx, frame, 0);
if(res < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_textures: av_hwframe_get_buffer failed: %d\n", res);
return false;
}
const unsigned int internal_formats_nv12[2] = { GL_R8, GL_RG8 };
const unsigned int internal_formats_p010[2] = { GL_R16, GL_RG16 };
const unsigned int formats[2] = { GL_RED, GL_RG };
const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size
for(int i = 0; i < 2; ++i) {
self->target_textures[i] = gl_create_texture(self->params.egl, video_codec_context->width / div[i], video_codec_context->height / div[i], self->params.color_depth == GSR_COLOR_DEPTH_8_BITS ? internal_formats_nv12[i] : internal_formats_p010[i], formats[i]);
if(self->target_textures[i] == 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_setup_textures: failed to create opengl texture\n");
return false;
}
if(!cuda_register_opengl_texture(&self->cuda, &self->cuda_graphics_resources[i], &self->mapped_arrays[i], self->target_textures[i])) {
return false;
}
}
return true;
}
static void gsr_video_encoder_nvenc_stop(gsr_video_encoder_nvenc *self, AVCodecContext *video_codec_context);
static bool gsr_video_encoder_nvenc_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_video_encoder_nvenc *self = encoder->priv;
const bool overclock = gsr_egl_get_display_server(self->params.egl) == GSR_DISPLAY_SERVER_X11 ? self->params.overclock : false;
if(!gsr_cuda_load(&self->cuda, self->params.egl->x11.dpy, overclock)) {
fprintf(stderr, "gsr error: gsr_video_encoder_nvenc_start: failed to load cuda\n");
gsr_video_encoder_nvenc_stop(self, video_codec_context);
return false;
}
if(!gsr_video_encoder_nvenc_setup_context(self, video_codec_context)) {
gsr_video_encoder_nvenc_stop(self, video_codec_context);
return false;
}
if(!gsr_video_encoder_nvenc_setup_textures(self, video_codec_context, frame)) {
gsr_video_encoder_nvenc_stop(self, video_codec_context);
return false;
}
return true;
}
void gsr_video_encoder_nvenc_stop(gsr_video_encoder_nvenc *self, AVCodecContext *video_codec_context) {
self->params.egl->glDeleteTextures(2, self->target_textures);
self->target_textures[0] = 0;
self->target_textures[1] = 0;
if(video_codec_context->hw_frames_ctx)
av_buffer_unref(&video_codec_context->hw_frames_ctx);
if(self->device_ctx)
av_buffer_unref(&self->device_ctx);
if(self->cuda.cu_ctx) {
for(int i = 0; i < 2; ++i) {
if(self->cuda_graphics_resources[i]) {
self->cuda.cuGraphicsUnmapResources(1, &self->cuda_graphics_resources[i], 0);
self->cuda.cuGraphicsUnregisterResource(self->cuda_graphics_resources[i]);
self->cuda_graphics_resources[i] = 0;
}
}
}
gsr_cuda_unload(&self->cuda);
}
static void gsr_video_encoder_nvenc_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion) {
gsr_video_encoder_nvenc *self = encoder->priv;
const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size
for(int i = 0; i < 2; ++i) {
CUDA_MEMCPY2D memcpy_struct;
memcpy_struct.srcXInBytes = 0;
memcpy_struct.srcY = 0;
memcpy_struct.srcMemoryType = CU_MEMORYTYPE_ARRAY;
memcpy_struct.dstXInBytes = 0;
memcpy_struct.dstY = 0;
memcpy_struct.dstMemoryType = CU_MEMORYTYPE_DEVICE;
memcpy_struct.srcArray = self->mapped_arrays[i];
memcpy_struct.srcPitch = frame->width / div[i];
memcpy_struct.dstDevice = (CUdeviceptr)frame->data[i];
memcpy_struct.dstPitch = frame->linesize[i];
memcpy_struct.WidthInBytes = frame->width * (self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? 2 : 1);
memcpy_struct.Height = frame->height / div[i];
// TODO: Remove this copy if possible
self->cuda.cuMemcpy2DAsync_v2(&memcpy_struct, self->cuda_stream);
}
// TODO: needed?
self->cuda.cuStreamSynchronize(self->cuda_stream);
}
static void gsr_video_encoder_nvenc_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {
gsr_video_encoder_nvenc *self = encoder->priv;
textures[0] = self->target_textures[0];
textures[1] = self->target_textures[1];
*num_textures = 2;
*destination_color = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
}
static void gsr_video_encoder_nvenc_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) {
gsr_video_encoder_nvenc_stop(encoder->priv, video_codec_context);
free(encoder->priv);
free(encoder);
}
gsr_video_encoder* gsr_video_encoder_nvenc_create(const gsr_video_encoder_nvenc_params *params) {
gsr_video_encoder *encoder = calloc(1, sizeof(gsr_video_encoder));
if(!encoder)
return NULL;
gsr_video_encoder_nvenc *encoder_cuda = calloc(1, sizeof(gsr_video_encoder_nvenc));
if(!encoder_cuda) {
free(encoder);
return NULL;
}
encoder_cuda->params = *params;
*encoder = (gsr_video_encoder) {
.start = gsr_video_encoder_nvenc_start,
.copy_textures_to_frame = gsr_video_encoder_nvenc_copy_textures_to_frame,
.get_textures = gsr_video_encoder_nvenc_get_textures,
.destroy = gsr_video_encoder_nvenc_destroy,
.priv = encoder_cuda
};
return encoder;
}

View File

@@ -58,26 +58,10 @@ static bool gsr_video_encoder_software_setup_textures(gsr_video_encoder_software
return true;
}
static gsr_supported_video_codecs gsr_video_encoder_software_get_supported_codecs(gsr_video_encoder *encoder, bool cleanup) {
(void)encoder;
(void)cleanup;
return (gsr_supported_video_codecs) {
.h264 = true,
.hevc = false,
.hevc_hdr = false,
.hevc_10bit = false,
.av1 = false,
.av1_hdr = false,
.av1_10bit = false,
.vp8 = false,
.vp9 = false
};
}
static void gsr_video_encoder_software_stop(gsr_video_encoder_software *self, AVCodecContext *video_codec_context);
static bool gsr_video_encoder_software_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_video_encoder_software *encoder_software = encoder->priv;
gsr_video_encoder_software *self = encoder->priv;
video_codec_context->width = FFALIGN(video_codec_context->width, LINESIZE_ALIGNMENT);
video_codec_context->height = FFALIGN(video_codec_context->height, 2);
@@ -85,8 +69,8 @@ static bool gsr_video_encoder_software_start(gsr_video_encoder *encoder, AVCodec
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
if(!gsr_video_encoder_software_setup_textures(encoder_software, video_codec_context, frame)) {
gsr_video_encoder_software_stop(encoder_software, video_codec_context);
if(!gsr_video_encoder_software_setup_textures(self, video_codec_context, frame)) {
gsr_video_encoder_software_stop(self, video_codec_context);
return false;
}
@@ -100,29 +84,29 @@ void gsr_video_encoder_software_stop(gsr_video_encoder_software *self, AVCodecCo
self->target_textures[1] = 0;
}
static void gsr_video_encoder_software_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame) {
gsr_video_encoder_software *encoder_software = encoder->priv;
static void gsr_video_encoder_software_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion) {
gsr_video_encoder_software *self = encoder->priv;
// TODO: hdr support
const unsigned int formats[2] = { GL_RED, GL_RG };
for(int i = 0; i < 2; ++i) {
encoder_software->params.egl->glBindTexture(GL_TEXTURE_2D, encoder_software->target_textures[i]);
self->params.egl->glBindTexture(GL_TEXTURE_2D, self->target_textures[i]);
// We could use glGetTexSubImage and then we wouldn't have to use a specific linesize (LINESIZE_ALIGNMENT) that adds padding,
// but glGetTexSubImage is only available starting from opengl 4.5.
encoder_software->params.egl->glGetTexImage(GL_TEXTURE_2D, 0, formats[i], GL_UNSIGNED_BYTE, frame->data[i]);
self->params.egl->glGetTexImage(GL_TEXTURE_2D, 0, formats[i], GL_UNSIGNED_BYTE, frame->data[i]);
}
encoder_software->params.egl->glBindTexture(GL_TEXTURE_2D, 0);
self->params.egl->glBindTexture(GL_TEXTURE_2D, 0);
// cap_kms->kms.base.egl->eglSwapBuffers(cap_kms->kms.base.egl->egl_display, cap_kms->kms.base.egl->egl_surface);
encoder_software->params.egl->glFlush();
encoder_software->params.egl->glFinish();
self->params.egl->glFlush();
self->params.egl->glFinish();
}
static void gsr_video_encoder_software_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {
gsr_video_encoder_software *encoder_software = encoder->priv;
textures[0] = encoder_software->target_textures[0];
textures[1] = encoder_software->target_textures[1];
gsr_video_encoder_software *self = encoder->priv;
textures[0] = self->target_textures[0];
textures[1] = self->target_textures[1];
*num_textures = 2;
*destination_color = encoder_software->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
*destination_color = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
}
static void gsr_video_encoder_software_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) {
@@ -145,7 +129,6 @@ gsr_video_encoder* gsr_video_encoder_software_create(const gsr_video_encoder_sof
encoder_software->params = *params;
*encoder = (gsr_video_encoder) {
.get_supported_codecs = gsr_video_encoder_software_get_supported_codecs,
.start = gsr_video_encoder_software_start,
.copy_textures_to_frame = gsr_video_encoder_software_copy_textures_to_frame,
.get_textures = gsr_video_encoder_software_get_textures,

View File

@@ -4,9 +4,9 @@
#include <libavcodec/avcodec.h>
#include <libavutil/hwcontext_vaapi.h>
#include <libavutil/intreadwrite.h>
#include <va/va_drmcommon.h>
#include <va/va_drm.h>
#include <stdlib.h>
#include <unistd.h>
@@ -102,6 +102,7 @@ static bool gsr_video_encoder_vaapi_setup_textures(gsr_video_encoder_vaapi *self
uint32_t pitches[4];
uint64_t modifiers[4];
for(uint32_t j = 0; j < self->prime.layers[layer].num_planes; ++j) {
// TODO: Close these? in _stop, using self->prime
fds[j] = self->prime.objects[self->prime.layers[layer].object_index[j]].fd;
offsets[j] = self->prime.layers[layer].offset[j];
pitches[j] = self->prime.layers[layer].pitch[j];
@@ -147,194 +148,16 @@ static bool gsr_video_encoder_vaapi_setup_textures(gsr_video_encoder_vaapi *self
}
}
static bool profile_is_h264(VAProfile profile) {
switch(profile) {
case 5: // VAProfileH264Baseline
case VAProfileH264Main:
case VAProfileH264High:
case VAProfileH264ConstrainedBaseline:
return true;
default:
return false;
}
}
static bool profile_is_hevc_8bit(VAProfile profile) {
switch(profile) {
case VAProfileHEVCMain:
return true;
default:
return false;
}
}
static bool profile_is_hevc_10bit(VAProfile profile) {
switch(profile) {
case VAProfileHEVCMain10:
//case VAProfileHEVCMain12:
//case VAProfileHEVCMain422_10:
//case VAProfileHEVCMain422_12:
//case VAProfileHEVCMain444:
//case VAProfileHEVCMain444_10:
//case VAProfileHEVCMain444_12:
return true;
default:
return false;
}
}
static bool profile_is_av1(VAProfile profile) {
switch(profile) {
case VAProfileAV1Profile0:
case VAProfileAV1Profile1:
return true;
default:
return false;
}
}
static bool profile_is_vp8(VAProfile profile) {
switch(profile) {
case VAProfileVP8Version0_3:
return true;
default:
return false;
}
}
static bool profile_is_vp9(VAProfile profile) {
switch(profile) {
case VAProfileVP9Profile0:
case VAProfileVP9Profile1:
case VAProfileVP9Profile2:
case VAProfileVP9Profile3:
return true;
default:
return false;
}
}
static bool profile_supports_video_encoding(VADisplay va_dpy, VAProfile profile) {
int num_entrypoints = vaMaxNumEntrypoints(va_dpy);
if(num_entrypoints <= 0)
return false;
VAEntrypoint *entrypoint_list = calloc(num_entrypoints, sizeof(VAEntrypoint));
if(!entrypoint_list)
return false;
bool supported = false;
if(vaQueryConfigEntrypoints(va_dpy, profile, entrypoint_list, &num_entrypoints) == VA_STATUS_SUCCESS) {
for(int i = 0; i < num_entrypoints; ++i) {
if(entrypoint_list[i] == VAEntrypointEncSlice) {
supported = true;
break;
}
}
}
free(entrypoint_list);
return supported;
}
static bool get_supported_video_codecs(VADisplay va_dpy, gsr_supported_video_codecs *video_codecs, bool cleanup) {
*video_codecs = (gsr_supported_video_codecs){0};
bool success = false;
VAProfile *profile_list = NULL;
vaSetInfoCallback(va_dpy, NULL, NULL);
int va_major = 0;
int va_minor = 0;
if(vaInitialize(va_dpy, &va_major, &va_minor) != VA_STATUS_SUCCESS) {
fprintf(stderr, "gsr error: gsr_video_encoder_vaapi_get_supported_codecs: vaInitialize failed\n");
goto fail;
}
int num_profiles = vaMaxNumProfiles(va_dpy);
if(num_profiles <= 0)
goto fail;
profile_list = calloc(num_profiles, sizeof(VAProfile));
if(!profile_list || vaQueryConfigProfiles(va_dpy, profile_list, &num_profiles) != VA_STATUS_SUCCESS)
goto fail;
for(int i = 0; i < num_profiles; ++i) {
if(profile_is_h264(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i]))
video_codecs->h264 = true;
} else if(profile_is_hevc_8bit(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i]))
video_codecs->hevc = true;
} else if(profile_is_hevc_10bit(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i])) {
video_codecs->hevc_hdr = true;
video_codecs->hevc_10bit = true;
}
} else if(profile_is_av1(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i])) {
video_codecs->av1 = true;
video_codecs->av1_hdr = true;
video_codecs->av1_10bit = true;
}
} else if(profile_is_vp8(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i]))
video_codecs->vp8 = true;
} else if(profile_is_vp9(profile_list[i])) {
if(profile_supports_video_encoding(va_dpy, profile_list[i]))
video_codecs->vp9 = true;
}
}
success = true;
fail:
if(profile_list)
free(profile_list);
if(cleanup)
vaTerminate(va_dpy);
return success;
}
static gsr_supported_video_codecs gsr_video_encoder_vaapi_get_supported_codecs(gsr_video_encoder *encoder, bool cleanup) {
gsr_video_encoder_vaapi *encoder_vaapi = encoder->priv;
gsr_supported_video_codecs supported_video_codecs = {0};
char render_path[128];
if(!gsr_card_path_get_render_path(encoder_vaapi->params.egl->card_path, render_path)) {
fprintf(stderr, "gsr error: gsr_video_encoder_vaapi_get_supported_codecs: failed to get /dev/dri/renderDXXX file from %s\n", encoder_vaapi->params.egl->card_path);
return supported_video_codecs;
}
const int drm_fd = open(render_path, O_RDWR);
if(drm_fd == -1) {
fprintf(stderr, "gsr error: gsr_video_encoder_vaapi_get_supported_codecs: failed to open device %s\n", render_path);
return supported_video_codecs;
}
VADisplay va_dpy = vaGetDisplayDRM(drm_fd);
if(va_dpy) {
if(!get_supported_video_codecs(va_dpy, &supported_video_codecs, cleanup))
fprintf(stderr, "gsr error: gsr_video_encoder_vaapi_get_supported_codecs: failed to query supported video codecs for device %s\n", render_path);
}
if(cleanup)
close(drm_fd);
return supported_video_codecs;
}
static void gsr_video_encoder_vaapi_stop(gsr_video_encoder_vaapi *self, AVCodecContext *video_codec_context);
static bool gsr_video_encoder_vaapi_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_video_encoder_vaapi *encoder_vaapi = encoder->priv;
gsr_video_encoder_vaapi *self = encoder->priv;
if(encoder_vaapi->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && video_codec_context->codec_id == AV_CODEC_ID_HEVC) {
if(self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && video_codec_context->codec_id == AV_CODEC_ID_HEVC) {
// TODO: dont do this if using ffmpeg reports that this is not needed (AMD driver bug that was fixed recently)
video_codec_context->width = FFALIGN(video_codec_context->width, 64);
video_codec_context->height = FFALIGN(video_codec_context->height, 16);
} else if(encoder_vaapi->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && video_codec_context->codec_id == AV_CODEC_ID_AV1) {
} else if(self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && video_codec_context->codec_id == AV_CODEC_ID_AV1) {
// TODO: Dont do this for VCN 5 and forward which should fix this hardware bug
video_codec_context->width = FFALIGN(video_codec_context->width, 64);
// AMD driver has special case handling for 1080 height to set it to 1082 instead of 1088 (1080 aligned to 16).
@@ -346,20 +169,40 @@ static bool gsr_video_encoder_vaapi_start(gsr_video_encoder *encoder, AVCodecCon
}
}
if(video_codec_context->width != frame->width || video_codec_context->height != frame->height) {
const int crop_top = (video_codec_context->height - frame->height) / 2;
const int crop_left = (video_codec_context->width - frame->width) / 2;
if(crop_top != 0 || crop_left != 0) {
fprintf(stderr, "gsr warning: gsr_video_encoder_vaapi_start: black bars have been added to the video because of a bug in AMD drivers/hardware. Record with h264 codec instead (-k h264) to get around this issue\n");
#if 0
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(61, 10, 100)
const int crop_bottom = crop_top;
const int crop_right = crop_left;
fprintf(stderr, "gsr info: cropping metadata has been added to the file to try and workaround this issue. Video players that support this will remove the black bars when the video is playing\n");
const int frame_cropping_data_size = 4 * sizeof(uint32_t);
uint8_t *frame_cropping = av_malloc(frame_cropping_data_size);
if(frame_cropping) {
AV_WL32(frame_cropping + 0, crop_top);
AV_WL32(frame_cropping + 4, crop_bottom);
AV_WL32(frame_cropping + 8, crop_left);
AV_WL32(frame_cropping + 12, crop_right);
const bool sidedata_added = av_packet_side_data_add(&video_stream->codecpar->coded_side_data, &video_stream->codecpar->nb_coded_side_data, AV_PKT_DATA_FRAME_CROPPING, frame_cropping, frame_cropping_data_size, 0) != NULL;
if(!sidedata_added)
av_free(frame_cropping);
}
#endif
#endif
}
frame->width = video_codec_context->width;
frame->height = video_codec_context->height;
if(!gsr_video_encoder_vaapi_setup_context(encoder_vaapi, video_codec_context)) {
gsr_video_encoder_vaapi_stop(encoder_vaapi, video_codec_context);
if(!gsr_video_encoder_vaapi_setup_context(self, video_codec_context)) {
gsr_video_encoder_vaapi_stop(self, video_codec_context);
return false;
}
if(!gsr_video_encoder_vaapi_setup_textures(encoder_vaapi, video_codec_context, frame)) {
gsr_video_encoder_vaapi_stop(encoder_vaapi, video_codec_context);
if(!gsr_video_encoder_vaapi_setup_textures(self, video_codec_context, frame)) {
gsr_video_encoder_vaapi_stop(self, video_codec_context);
return false;
}
@@ -385,11 +228,11 @@ void gsr_video_encoder_vaapi_stop(gsr_video_encoder_vaapi *self, AVCodecContext
}
static void gsr_video_encoder_vaapi_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {
gsr_video_encoder_vaapi *encoder_vaapi = encoder->priv;
textures[0] = encoder_vaapi->target_textures[0];
textures[1] = encoder_vaapi->target_textures[1];
gsr_video_encoder_vaapi *self = encoder->priv;
textures[0] = self->target_textures[0];
textures[1] = self->target_textures[1];
*num_textures = 2;
*destination_color = encoder_vaapi->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
*destination_color = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
}
static void gsr_video_encoder_vaapi_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) {
@@ -412,9 +255,7 @@ gsr_video_encoder* gsr_video_encoder_vaapi_create(const gsr_video_encoder_vaapi_
encoder_vaapi->params = *params;
*encoder = (gsr_video_encoder) {
.get_supported_codecs = gsr_video_encoder_vaapi_get_supported_codecs,
.start = gsr_video_encoder_vaapi_start,
.copy_textures_to_frame = NULL,
.get_textures = gsr_video_encoder_vaapi_get_textures,
.destroy = gsr_video_encoder_vaapi_destroy,
.priv = encoder_vaapi

View File

@@ -1,10 +1,6 @@
#include "../../../include/encoder/video/video.h"
#include <assert.h>
gsr_supported_video_codecs gsr_video_encoder_get_supported_codecs(gsr_video_encoder *encoder, bool cleanup) {
return encoder->get_supported_codecs(encoder, cleanup);
}
bool gsr_video_encoder_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
assert(!encoder->started);
bool res = encoder->start(encoder, video_codec_context, frame);
@@ -13,10 +9,10 @@ bool gsr_video_encoder_start(gsr_video_encoder *encoder, AVCodecContext *video_c
return res;
}
void gsr_video_encoder_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame) {
void gsr_video_encoder_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion) {
assert(encoder->started);
if(encoder->copy_textures_to_frame)
encoder->copy_textures_to_frame(encoder, frame);
encoder->copy_textures_to_frame(encoder, frame, color_conversion);
}
void gsr_video_encoder_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {

313
src/encoder/video/vulkan.c Normal file
View File

@@ -0,0 +1,313 @@
#include "../../../include/encoder/video/vulkan.h"
#include "../../../include/utils.h"
#include "../../../include/egl.h"
#include <libavcodec/avcodec.h>
#define VK_NO_PROTOTYPES
#include <libavutil/hwcontext_vulkan.h>
//#include <vulkan/vulkan_core.h>
#define GL_TEXTURE_TILING_EXT 0x9580
#define GL_OPTIMAL_TILING_EXT 0x9584
#define GL_LINEAR_TILING_EXT 0x9585
#define GL_PIXEL_PACK_BUFFER 0x88EB
#define GL_PIXEL_UNPACK_BUFFER 0x88EC
#define GL_STREAM_READ 0x88E1
#define GL_STREAM_DRAW 0x88E0
#define GL_READ_ONLY 0x88B8
#define GL_WRITE_ONLY 0x88B9
#define GL_READ_FRAMEBUFFER 0x8CA8
typedef struct {
gsr_video_encoder_vulkan_params params;
unsigned int target_textures[2];
AVBufferRef *device_ctx;
AVVulkanDeviceContext* vv;
unsigned int pbo_y[2];
unsigned int pbo_uv[2];
AVFrame *sw_frame;
} gsr_video_encoder_vulkan;
static bool gsr_video_encoder_vulkan_setup_context(gsr_video_encoder_vulkan *self, AVCodecContext *video_codec_context) {
AVDictionary *options = NULL;
//av_dict_set(&options, "linear_images", "1", 0);
//av_dict_set(&options, "disable_multiplane", "1", 0);
// TODO: Use correct device
if(av_hwdevice_ctx_create(&self->device_ctx, AV_HWDEVICE_TYPE_VULKAN, NULL, options, 0) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_vulkan_setup_context: failed to create hardware device context\n");
return false;
}
AVBufferRef *frame_context = av_hwframe_ctx_alloc(self->device_ctx);
if(!frame_context) {
fprintf(stderr, "gsr error: gsr_video_encoder_vulkan_setup_context: failed to create hwframe context\n");
av_buffer_unref(&self->device_ctx);
return false;
}
AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)frame_context->data;
hw_frame_context->width = video_codec_context->width;
hw_frame_context->height = video_codec_context->height;
hw_frame_context->sw_format = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? AV_PIX_FMT_P010LE : AV_PIX_FMT_NV12;
hw_frame_context->format = video_codec_context->pix_fmt;
hw_frame_context->device_ctx = (AVHWDeviceContext*)self->device_ctx->data;
//AVVulkanFramesContext *vk_frame_ctx = (AVVulkanFramesContext*)hw_frame_context->hwctx;
//hw_frame_context->initial_pool_size = 20;
if (av_hwframe_ctx_init(frame_context) < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_vulkan_setup_context: failed to initialize hardware frame context "
"(note: ffmpeg version needs to be > 4.0)\n");
av_buffer_unref(&self->device_ctx);
//av_buffer_unref(&frame_context);
return false;
}
video_codec_context->hw_frames_ctx = av_buffer_ref(frame_context);
av_buffer_unref(&frame_context);
return true;
}
static unsigned int gl_create_texture(gsr_egl *egl, int width, int height, int internal_format, unsigned int format) {
unsigned int texture_id = 0;
egl->glGenTextures(1, &texture_id);
egl->glBindTexture(GL_TEXTURE_2D, texture_id);
//egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_TILING_EXT, GL_OPTIMAL_TILING_EXT);
egl->glTexImage2D(GL_TEXTURE_2D, 0, internal_format, width, height, 0, format, GL_UNSIGNED_BYTE, NULL);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
egl->glBindTexture(GL_TEXTURE_2D, 0);
return texture_id;
}
static AVVulkanDeviceContext* video_codec_context_get_vulkan_data(AVCodecContext *video_codec_context) {
AVBufferRef *hw_frames_ctx = video_codec_context->hw_frames_ctx;
if(!hw_frames_ctx)
return NULL;
AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)hw_frames_ctx->data;
AVHWDeviceContext *device_context = (AVHWDeviceContext*)hw_frame_context->device_ctx;
if(device_context->type != AV_HWDEVICE_TYPE_VULKAN)
return NULL;
return (AVVulkanDeviceContext*)device_context->hwctx;
}
static bool gsr_video_encoder_vulkan_setup_textures(gsr_video_encoder_vulkan *self, AVCodecContext *video_codec_context, AVFrame *frame) {
const int res = av_hwframe_get_buffer(video_codec_context->hw_frames_ctx, frame, 0);
if(res < 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_vulkan_setup_textures: av_hwframe_get_buffer failed: %d\n", res);
return false;
}
//AVVkFrame *target_surface_id = (AVVkFrame*)frame->data[0];
self->vv = video_codec_context_get_vulkan_data(video_codec_context);
const unsigned int internal_formats_nv12[2] = { GL_RGBA8, GL_RGBA8 };
const unsigned int internal_formats_p010[2] = { GL_R16, GL_RG16 };
const unsigned int formats[2] = { GL_RED, GL_RG };
const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size
for(int i = 0; i < 2; ++i) {
self->target_textures[i] = gl_create_texture(self->params.egl, video_codec_context->width / div[i], video_codec_context->height / div[i], self->params.color_depth == GSR_COLOR_DEPTH_8_BITS ? internal_formats_nv12[i] : internal_formats_p010[i], formats[i]);
if(self->target_textures[i] == 0) {
fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_textures: failed to create opengl texture\n");
return false;
}
}
self->params.egl->glGenBuffers(2, self->pbo_y);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[0]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, frame->width * frame->height, 0, GL_STREAM_READ);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[1]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, frame->width * frame->height, 0, GL_STREAM_READ);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
self->params.egl->glGenBuffers(2, self->pbo_uv);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_uv[0]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, (frame->width/2 * frame->height/2) * 2, 0, GL_STREAM_READ);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_uv[1]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, (frame->width/2 * frame->height/2) * 2, 0, GL_STREAM_READ);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
self->sw_frame = av_frame_alloc();
self->sw_frame->format = AV_PIX_FMT_NV12;
self->sw_frame->width = frame->width;
self->sw_frame->height = frame->height;
// TODO: Remove
if(av_frame_get_buffer(self->sw_frame, 0) < 0) {
fprintf(stderr, "failed to allocate sw frame\n");
}
// TODO: Remove
if(av_frame_make_writable(self->sw_frame) < 0) {
fprintf(stderr, "failed to make writable\n");
}
return true;
}
static void gsr_video_encoder_vulkan_stop(gsr_video_encoder_vulkan *self, AVCodecContext *video_codec_context);
static bool gsr_video_encoder_vulkan_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) {
gsr_video_encoder_vulkan *self = encoder->priv;
if(!gsr_video_encoder_vulkan_setup_context(self, video_codec_context)) {
gsr_video_encoder_vulkan_stop(self, video_codec_context);
return false;
}
if(!gsr_video_encoder_vulkan_setup_textures(self, video_codec_context, frame)) {
gsr_video_encoder_vulkan_stop(self, video_codec_context);
return false;
}
return true;
}
void gsr_video_encoder_vulkan_stop(gsr_video_encoder_vulkan *self, AVCodecContext *video_codec_context) {
self->params.egl->glDeleteTextures(2, self->target_textures);
self->target_textures[0] = 0;
self->target_textures[1] = 0;
if(video_codec_context->hw_frames_ctx)
av_buffer_unref(&video_codec_context->hw_frames_ctx);
if(self->device_ctx)
av_buffer_unref(&self->device_ctx);
}
static void nop_free(void *opaque, uint8_t *data) {
}
static void gsr_video_encoder_vulkan_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame, gsr_color_conversion *color_conversion) {
gsr_video_encoder_vulkan *self = encoder->priv;
static int counter = 0;
++counter;
// AVBufferRef *av_buffer_create(uint8_t *data, size_t size,
// void (*free)(void *opaque, uint8_t *data),
// void *opaque, int flags);
while(self->params.egl->glGetError()){}
self->params.egl->glBindFramebuffer(GL_READ_FRAMEBUFFER, color_conversion->framebuffers[0]);
//fprintf(stderr, "1 gl err: %d\n", self->params.egl->glGetError());
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[counter % 2]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, frame->width * frame->height, 0, GL_STREAM_READ);
self->params.egl->glReadPixels(0, 0, frame->width, frame->height, GL_RED, GL_UNSIGNED_BYTE, 0);
//fprintf(stderr, "2 gl err: %d\n", self->params.egl->glGetError());
const int next_pbo_y = (counter + 1) % 2;
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[next_pbo_y]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, frame->width * frame->height, 0, GL_STREAM_READ);
//fprintf(stderr, "3 gl err: %d\n", self->params.egl->glGetError());
uint8_t *ptr_y = (uint8_t*)self->params.egl->glMapBuffer(GL_PIXEL_PACK_BUFFER, GL_READ_ONLY);
//fprintf(stderr, "4 gl err: %d\n", self->params.egl->glGetError());
if(!ptr_y) {
fprintf(stderr, "failed to map buffer y!\n");
}
while(self->params.egl->glGetError()){}
self->params.egl->glBindFramebuffer(GL_READ_FRAMEBUFFER, color_conversion->framebuffers[1]);
//fprintf(stderr, "5 gl err: %d\n", self->params.egl->glGetError());
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_uv[counter % 2]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, (frame->width/2 * frame->height/2) * 2, 0, GL_STREAM_READ);
//fprintf(stderr, "5.5 gl err: %d\n", self->params.egl->glGetError());
self->params.egl->glReadPixels(0, 0, frame->width/2, frame->height/2, GL_RG, GL_UNSIGNED_BYTE, 0);
//fprintf(stderr, "6 gl err: %d\n", self->params.egl->glGetError());
const int next_pbo_uv = (counter + 1) % 2;
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_uv[next_pbo_uv]);
self->params.egl->glBufferData(GL_PIXEL_PACK_BUFFER, (frame->width/2 * frame->height/2) * 2, 0, GL_STREAM_READ);
//fprintf(stderr, "7 gl err: %d\n", self->params.egl->glGetError());
uint8_t *ptr_uv = (uint8_t*)self->params.egl->glMapBuffer(GL_PIXEL_PACK_BUFFER, GL_READ_ONLY);
//fprintf(stderr, "8 gl err: %d\n", self->params.egl->glGetError());
if(!ptr_uv) {
fprintf(stderr, "failed to map buffer uv!\n");
}
//self->sw_frame->buf[0] = av_buffer_create(ptr_y, 3840 * 2160, nop_free, NULL, 0);
//self->sw_frame->buf[1] = av_buffer_create(ptr_uv, 1920 * 1080 * 2, nop_free, NULL, 0);
//self->sw_frame->data[0] = self->sw_frame->buf[0]->data;
//self->sw_frame->data[1] = self->sw_frame->buf[1]->data;
//self->sw_frame->extended_data[0] = self->sw_frame->data[0];
//self->sw_frame->extended_data[1] = self->sw_frame->data[1];
self->sw_frame->data[0] = ptr_y;
self->sw_frame->data[1] = ptr_uv;
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
self->params.egl->glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
//self->params.egl->glBindTexture(GL_TEXTURE_2D, self->target_textures[1]);
//self->params.egl->glGetTexImage(GL_TEXTURE_2D, 0, GL_RG, GL_UNSIGNED_BYTE, sw_frame->data[1]);
//self->params.egl->glBindTexture(GL_TEXTURE_2D, 0);
int ret = av_hwframe_transfer_data(frame, self->sw_frame, 0);
if(ret < 0) {
fprintf(stderr, "transfer data failed, error: %s\n", av_err2str(ret));
}
//av_buffer_unref(&self->sw_frame->buf[0]);
//av_buffer_unref(&self->sw_frame->buf[1]);
//av_frame_free(&sw_frame);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[next_pbo_y]);
self->params.egl->glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, self->pbo_y[next_pbo_uv]);
self->params.egl->glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
self->params.egl->glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
}
static void gsr_video_encoder_vulkan_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) {
gsr_video_encoder_vulkan *self = encoder->priv;
textures[0] = self->target_textures[0];
textures[1] = self->target_textures[1];
*num_textures = 2;
*destination_color = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12;
}
static void gsr_video_encoder_vulkan_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) {
gsr_video_encoder_vulkan_stop(encoder->priv, video_codec_context);
free(encoder->priv);
free(encoder);
}
gsr_video_encoder* gsr_video_encoder_vulkan_create(const gsr_video_encoder_vulkan_params *params) {
gsr_video_encoder *encoder = calloc(1, sizeof(gsr_video_encoder));
if(!encoder)
return NULL;
gsr_video_encoder_vulkan *encoder_vulkan = calloc(1, sizeof(gsr_video_encoder_vulkan));
if(!encoder_vulkan) {
free(encoder);
return NULL;
}
encoder_vulkan->params = *params;
*encoder = (gsr_video_encoder) {
.start = gsr_video_encoder_vulkan_start,
.copy_textures_to_frame = gsr_video_encoder_vulkan_copy_textures_to_frame,
.get_textures = gsr_video_encoder_vulkan_get_textures,
.destroy = gsr_video_encoder_vulkan_destroy,
.priv = encoder_vulkan
};
return encoder;
}

File diff suppressed because it is too large Load Diff

View File

@@ -340,16 +340,18 @@ static void pa_server_info_cb(pa_context*, const pa_server_info *server_info, vo
}
static void get_pulseaudio_default_inputs(AudioDevices &audio_devices) {
pa_mainloop *main_loop = pa_mainloop_new();
pa_context *ctx = pa_context_new(pa_mainloop_get_api(main_loop), "gpu-screen-recorder-gtk");
pa_context_connect(ctx, NULL, PA_CONTEXT_NOFLAGS, NULL);
int state = 0;
int pa_ready = 0;
pa_context_set_state_callback(ctx, pa_state_cb, &pa_ready);
pa_operation *pa_op = NULL;
pa_mainloop *main_loop = pa_mainloop_new();
pa_context *ctx = pa_context_new(pa_mainloop_get_api(main_loop), "gpu-screen-recorder");
if(pa_context_connect(ctx, NULL, PA_CONTEXT_NOFLAGS, NULL) < 0)
goto done;
pa_context_set_state_callback(ctx, pa_state_cb, &pa_ready);
for(;;) {
// Not ready
if(pa_ready == 0) {
@@ -366,23 +368,25 @@ static void get_pulseaudio_default_inputs(AudioDevices &audio_devices) {
}
// Couldn't get connection to the server
if(pa_ready == 2 || (state == 1 && pa_op && pa_operation_get_state(pa_op) == PA_OPERATION_DONE)) {
if(pa_op)
pa_operation_unref(pa_op);
pa_context_disconnect(ctx);
pa_context_unref(ctx);
pa_mainloop_free(main_loop);
return;
}
if(pa_ready == 2 || (state == 1 && pa_op && pa_operation_get_state(pa_op) == PA_OPERATION_DONE))
break;
pa_mainloop_iterate(main_loop, 1, NULL);
}
done:
if(pa_op)
pa_operation_unref(pa_op);
pa_context_disconnect(ctx);
pa_context_unref(ctx);
pa_mainloop_free(main_loop);
}
AudioDevices get_pulseaudio_inputs() {
AudioDevices audio_devices;
int state = 0;
int pa_ready = 0;
pa_operation *pa_op = NULL;
// TODO: Do this in the same connection below instead of two separate connections
get_pulseaudio_default_inputs(audio_devices);
@@ -390,12 +394,10 @@ AudioDevices get_pulseaudio_inputs() {
pa_mainloop *main_loop = pa_mainloop_new();
pa_context *ctx = pa_context_new(pa_mainloop_get_api(main_loop), "gpu-screen-recorder");
pa_context_connect(ctx, NULL, PA_CONTEXT_NOFLAGS, NULL);
int state = 0;
int pa_ready = 0;
pa_context_set_state_callback(ctx, pa_state_cb, &pa_ready);
if(pa_context_connect(ctx, NULL, PA_CONTEXT_NOFLAGS, NULL) < 0)
goto done;
pa_operation *pa_op = NULL;
pa_context_set_state_callback(ctx, pa_state_cb, &pa_ready);
for(;;) {
// Not ready
@@ -413,17 +415,17 @@ AudioDevices get_pulseaudio_inputs() {
}
// Couldn't get connection to the server
if(pa_ready == 2 || (state == 1 && pa_op && pa_operation_get_state(pa_op) == PA_OPERATION_DONE)) {
if(pa_op)
pa_operation_unref(pa_op);
pa_context_disconnect(ctx);
pa_context_unref(ctx);
if(pa_ready == 2 || (state == 1 && pa_op && pa_operation_get_state(pa_op) == PA_OPERATION_DONE))
break;
}
pa_mainloop_iterate(main_loop, 1, NULL);
}
done:
if(pa_op)
pa_operation_unref(pa_op);
pa_context_disconnect(ctx);
pa_context_unref(ctx);
pa_mainloop_free(main_loop);
return audio_devices;
}

View File

@@ -165,7 +165,7 @@ static bool connector_get_property_by_name(int drmfd, drmModeConnectorPtr props,
return false;
}
/* TODO: Support more connector types*/
/* TODO: Support more connector types */
static int get_connector_type_by_name(const char *name) {
int len = strlen(name);
if(len >= 5 && strncmp(name, "HDMI-", 5) == 0)
@@ -400,6 +400,8 @@ bool gl_get_gpu_info(gsr_egl *egl, gsr_gpu_info *info) {
if(strstr((const char*)gl_vendor, "AMD"))
info->vendor = GSR_GPU_VENDOR_AMD;
else if(strstr((const char*)gl_vendor, "Mesa") && gl_renderer && strstr((const char*)gl_renderer, "AMD"))
info->vendor = GSR_GPU_VENDOR_AMD;
else if(strstr((const char*)gl_vendor, "Intel"))
info->vendor = GSR_GPU_VENDOR_INTEL;
else if(strstr((const char*)gl_vendor, "NVIDIA"))
@@ -464,23 +466,14 @@ static bool try_card_has_valid_plane(const char *card_path) {
return false;
}
static void string_copy(char *dst, const char *src, int len) {
int src_len = strlen(src);
int min_len = src_len;
if(len - 1 < min_len)
min_len = len - 1;
memcpy(dst, src, min_len);
dst[min_len] = '\0';
}
bool gsr_get_valid_card_path(gsr_egl *egl, char *output, bool is_monitor_capture) {
if(egl->dri_card_path) {
string_copy(output, egl->dri_card_path, 127);
snprintf(output, 128, "%s", egl->dri_card_path);
return is_monitor_capture ? try_card_has_valid_plane(output) : true;
}
for(int i = 0; i < 10; ++i) {
snprintf(output, 127, DRM_DEV_NAME, DRM_DIR_NAME, i);
snprintf(output, 128, DRM_DEV_NAME, DRM_DIR_NAME, i);
if(try_card_has_valid_plane(output))
return true;
}
@@ -494,7 +487,7 @@ bool gsr_card_path_get_render_path(const char *card_path, char *render_path) {
char *render_path_tmp = drmGetRenderDeviceNameFromFd(fd);
if(render_path_tmp) {
string_copy(render_path, render_path_tmp, 127);
snprintf(render_path, 128, "%s", render_path_tmp);
free(render_path_tmp);
close(fd);
return true;
@@ -745,6 +738,8 @@ bool vaapi_copy_drm_planes_to_video_surface(AVCodecContext *video_codec_context,
.height = dest_size.y
};
const bool scaled = dest_size.x != source_size.x || dest_size.y != source_size.y;
// Copying a surface to another surface will automatically perform the color conversion. Thanks vaapi!
VAProcPipelineParameterBuffer params = {0};
params.surface = input_surface_id;
@@ -752,7 +747,7 @@ bool vaapi_copy_drm_planes_to_video_surface(AVCodecContext *video_codec_context,
params.surface_region = &source_region;
params.output_region = &output_region;
params.output_background_color = 0;
params.filter_flags = VA_FRAME_PICTURE;
params.filter_flags = scaled ? (VA_FILTER_SCALING_HQ | VA_FILTER_INTERPOLATION_BILINEAR) : 0;
params.pipeline_flags = VA_PROC_PIPELINE_FAST;
params.input_color_properties.colour_primaries = 1;
@@ -884,3 +879,20 @@ bool vaapi_copy_egl_image_to_video_surface(gsr_egl *egl, EGLImage image, vec2i s
return success;
}
vec2i scale_keep_aspect_ratio(vec2i from, vec2i to) {
if(from.x == 0 || from.y == 0)
return (vec2i){0, 0};
const double height_to_width_ratio = (double)from.y / (double)from.x;
from.x = to.x;
from.y = from.x * height_to_width_ratio;
if(from.y > to.y) {
const double width_height_ratio = (double)from.x / (double)from.y;
from.y = to.y;
from.x = from.y * width_height_ratio;
}
return from;
}