Add Enhanced Broadcasting support for Linux #11541

lexano-ivs · 2024-11-20T19:59:09Z

Description

TEB (Twitch Enhanced Broadcasting) requires the client to gather information about the host, such as which GPU(s) are available, memory sizes, OS version, etc..., and send that information to a server which then responds with a configuration set for the client to use with multitrack encoding. TEB originally supported Windows, and this PR adds support for Linux-based operating systems.

TEB on Linux supports NVIDIA and AMD GPUs, and Intel GPUs as of 2024.12.17. Intel GPUs with Linux seem to work at least with a flatpak build. For NVIDIA GPUs, NVENC is used and the configurations from the server are currently identical between Windows and Linux. For AMD GPUs, AMF is used on Windows and VAAPI is used on Linux; AMF on Linux is not currently supported for TEB. AMF on Linux might be supported in the future.

Motivation and Context

From the client encoding perspective, Enhanced Broadcasting was originally envisioned to support Windows, Linux, and MacOS. This PR addresses the market need to expand support for Linux-based users.

How Has This Been Tested?

The code has been tested in the TEB beta program and also by the primary developer (@lexano-ivs) on a host setup specifically to test TEB on Linux. This host includes:

iGPU (Intel)
NVIDIA 4060
AMD 6650 XT and AMD 7800 XT
Intel Arc A770
Multi-boot system with Ubuntu 24.04, Linux Mint 22, and EndeavourOS (Arch-based)

The code was tested against multiple configurations, including single & dual dGPU configurations, swapping the primary/secondary GPU between NVIDIA and AMD, and across the 3 Linux operating systems. DEB-based builds, local builds, and flatpak builds were also tested.

Refer to TEB Installation Instructions for additional context.

Types of changes

New feature (non-breaking change which adds functionality)

Checklist:

My code has been run through clang-format.
I have read the contributing document.
My code is not on the master branch.
The code has been tested.
All commit messages are properly formatted and commits squashed where appropriate.
I have included updates to all appropriate documentation.

plugins/obs-ffmpeg/obs-ffmpeg-vaapi.c

tytan652 · 2024-11-20T20:18:04Z

UI/system-info-posix.cpp

+
+static bool get_distribution_info(string &distro, string &version)
+{
+	ifstream file("/etc/os-release");


In a Flatpak context, this file should not be relied on.

Thanks tytan652. Is there something else within a flatpak that can be relied upon?

The presence of /.flatpak-info and it's content can be used.

Thank you, I'll look into that.

I've added parsing of the /.flatpak-info in addition to /etc/os-release.

Lain-B

In addition to my comment on Tytan's comment about VAAPI settings migration, these are the only issues I see so far.

Lain-B · 2024-11-24T21:38:52Z

UI/system-info-posix.cpp

Okay, so for the all of the UI/system-info-posix.cpp changes, there seems to be a bit of a mix between using raw pointers and references.

static void trim_ws(std::string &s)

static bool get_distribution_info(string &distro, string &version)

static bool get_cpu_freq(uint32_t *cpu_freq)

static bool get_drm_card_info(uint8_t *card_idx, struct drm_card_info *dci)

static void adjust_gpu_model(std::string *model)

bool compare_match_strength(const drm_card_info &a, const drm_card_info &b)

I'm not entirely sure why pointers were used in certain cases here like this versus references. In C this would be understandable but in C++ this just adds unnecessarily unsafe code. Might want to fix these (and any other similar cases in here I haven't spotted) to use references.

If you need the ability to pass nullptr (which I don't quite see), that would also make sense, but I would probably try to see if that's absolutely necessary, and if so, to check to see if there's a safer way of passing that (e.g. std::optional).

Thank you. I've cleaned this up with the latest commits. It's more consistent now and using references instead of pointers.

@Lain-B Looks like these have been addressed?

Lain-B · 2024-11-24T21:45:14Z

UI/system-info-posix.cpp

+	}
+}
+
+bool compare_match_strength(const drm_card_info &a, const drm_card_info &b)


Unless this function is used in a different translation unit, might want to mark it as static.

Also recommend checking other newly added non-exported functions elsewhere where this situation might apply, but this is just the one I noticed at first glance.

Unless this function is used in a different translation unit, might want to mark it as static.

Or for better C++-isms, put them into an anonymous (unnamed) namespace.

I've put almost all functions in an anonymous namespace and dropped "static" for these. It's cleaner from perspective. Thanks for the suggestion, it makes sense.

@Lain-B looks like this has been addressed?

tytan652 · 2024-12-19T11:50:15Z

UI/system-info-posix.cpp

+	distro = "";
+	version = "";
+
+	if (std::filesystem::exists(systemd_file)) {


Wrong priority, Flatpak should be checked first.

Runtime can add an os-release file for compatibility, but has been proven to be an issue (e.g. 6.2 was still showing 5.15 in a lsb-release made for compat) , let's avoid it to happen again.

Thanks tytan652. I reversed the parsing order and pushed the update.

@tytan652 has this been addressed?

UI/system-info-posix.cpp

mihawk90 · 2025-01-19T08:44:11Z

Just wanted to add my feedback for future reviewers:
I've applied this to my build a couple weeks ago and just ran a 9h stream (on an AMD RX 7700XT) with 0 issues, it's working great.
Thanks for putting the work into this.

The only thing that was odd was that the Twitch Inspector doesn't show any of the streams when in Bandwidth Testing mode, but it's showing them fine for a Live Broadcast. I don't think that's related to this PR though and more of an Inspector thing.

dsaedtler · 2025-01-24T00:52:13Z

The only thing that was odd was that the Twitch Inspector doesn't show any of the streams when in Bandwidth Testing mode, but it's showing them fine for a Live Broadcast. I don't think that's related to this PR though and more of an Inspector thing.

Yeah in bandwidth testing mode the data doesn't actually get processed, so it doesn't log codecs etc.

Add graphics APIs to obtain the GPU driver version and renderer strings, as well as GDDR memory sizes. The GDDR memory sizes include the dmem (dedicated memory on the GPU) and smem (shared CPU memory used by the GPU but resides in the CPU DDR). The version and renderer strings are needed for identification purposes, for example enhanced broadcasting used by Twitch, to associate the GPU used by OBS with the PCIe-based identification values such as device_id and vendor_id.

`os_get_free_size()` was simply returning 0. For Linux, implement the free size calculation based on the `sysinfo()` system call.

`get_rc_mode()` compares the incoming rate control mode string to the .name member of rc_mode_t, and the table entries are all upper-case. This caused a crash when the incoming string is set to "cbr" instead of "CBR". Make the string comparison case insensitive.

VAAPI encoders deviate from other encoders (e.g. AMF, NVENC) with the "profile" setting being an integer instead of a string. With enhanced broadcasting, "profile" is signalled as a string. Convert the string-based profile to the appropriate integer-based profile for VAAPI encoders as a workaround, until VAAPI supports string-based "profile" (if ever).

Enhanced broadcasting requires system information to be gathered on the client and submitted to the GetClientConfiguration request in order to obtain a valid response from the server. This commit adds support for gathering the required information on Linux-based systems.

RytoEX · 2024-11-20T20:06:55Z

libobs/graphics/graphics-imports.c

+	GRAPHICS_IMPORT_OPTIONAL(gpu_get_smem);
+	GRAPHICS_IMPORT_OPTIONAL(gpu_get_dmem);
+	GRAPHICS_IMPORT_OPTIONAL(gpu_get_driver_version);
+	GRAPHICS_IMPORT_OPTIONAL(gpu_get_renderer);


Prefer these in the same order they are declared/defined.

RytoEX · 2024-11-20T20:07:37Z

libobs/graphics/graphics.c

+uint64_t gs_get_gpu_dmem(void)
+{
+	return gs_valid("gs_get_gpu_dmem") ? thread_graphics->exports.gpu_get_dmem() : 0;
+}
+
+uint64_t gs_get_gpu_smem(void)
+{
+	return gs_valid("gs_get_gpu_smem") ? thread_graphics->exports.gpu_get_smem() : 0;
+}
+
+const char *gs_get_driver_version(void)
+{
+	return gs_valid("gs_get_driver_version") ? thread_graphics->exports.gpu_get_driver_version() : NULL;
+}
+
+const char *gs_get_renderer(void)
+{
+	return gs_valid("gs_get_renderer") ? thread_graphics->exports.gpu_get_renderer() : NULL;
+}
+


Prefer these in the same order they are declared/defined in the header.

RytoEX · 2024-11-20T20:10:34Z

UI/window-basic-settings.cpp

+	// Enhanced Broadcasting works on Windows and Linux. For other
+	// OS variants, only enable the GUI controls if developer mode
+	// was invoked.
+#if (!defined(__linux__) && !defined(_WIN32))


This will probably conflict with the PR to enable TEB on macOS.

RytoEX · 2025-02-12T17:41:03Z

frontend/settings/OBSBasicSettings.cpp

-#ifndef _WIN32
+	// Enhanced Broadcasting works on Windows and Linux. For other
+	// OS variants, only enable the GUI controls if developer mode
+	// was invoked.
+#if (!defined(__linux__) && !defined(_WIN32))


This will need updated to account for #11440 being merged.

RytoEX · 2025-02-12T17:46:45Z

libobs-opengl/gl-subsystem.c

+	 */
+	if (dmem < 0)
+		dmem = 0;
+	return (uint64_t)dmem;


Prefer a blank line after an un-bracketed if, or just enclose the if in braces.

RytoEX · 2025-02-12T17:50:24Z

plugins/obs-ffmpeg/obs-ffmpeg-vaapi.c

-	while (!!rc_mode->name && strcmp(rc_mode->name, name) != 0)
+	while (!!rc_mode->name && astrcmpi(rc_mode->name, name) != 0)


What is the rationale for this change?

RytoEX · 2025-02-12T18:15:58Z

frontend/utility/MultitrackVideoOutput.cpp

+	/* VAAPI-based encoders unfortunately use an integer for "profile".
+	 * Until a string-based "profile" can be used with VAAPI, find the
+	 * corresponding integer value and update the settings with an
+	 * integer-based "profile".
+	 */


Lines can be wrapped at 120 columns.

RytoEX · 2025-02-12T18:16:27Z

frontend/utility/MultitrackVideoOutput.cpp

+	 * integer-based "profile".
+	 */
+	if (strstr(encoder_type, "vaapi")) {
+		// Move the "profile" string to "profile_str"


For comments using complete sentences, add final punctuation at the end.

RytoEX · 2025-02-12T18:16:56Z

libobs-opengl/gl-subsystem.c

+/* Get the amount of dedicated GDDR memory, aka VRAM, in
+ * units of kilobytes.
+ */


Lines can be wrapped at 120 columns.

RytoEX · 2025-02-12T18:17:34Z

libobs-opengl/gl-subsystem.c

+		dmem = 0;
+	if (total_mem < 0)
+		total_mem = 0;
+	return (uint64_t)total_mem - dmem;


Prefer a blank line after an un-bracketed if, or enclose all ifs in braces.

tytan652 suggested changes Nov 20, 2024

View reviewed changes

plugins/obs-ffmpeg/obs-ffmpeg-vaapi.c Outdated Show resolved Hide resolved

tytan652 reviewed Nov 20, 2024

View reviewed changes

WizardCM added Linux Affects Linux New Feature New feature or plugin labels Nov 24, 2024

Lain-B requested changes Nov 24, 2024

View reviewed changes

lexano-ivs force-pushed the lexano/teb-linux branch 2 times, most recently from 69c3d29 to efd4f5e Compare December 18, 2024 19:17

lexano-ivs requested review from Lain-B, tytan652 and gxalpha December 18, 2024 19:38

tytan652 reviewed Dec 19, 2024

View reviewed changes

eiymba reviewed Dec 23, 2024

View reviewed changes

UI/system-info-posix.cpp Outdated Show resolved Hide resolved

lexano-ivs force-pushed the lexano/teb-linux branch from efd4f5e to dd12c61 Compare January 6, 2025 22:15

lexano-ivs added 6 commits January 31, 2025 09:53

libobs/util: Update os_get_free_size()

999843c

`os_get_free_size()` was simply returning 0. For Linux, implement the free size calculation based on the `sysinfo()` system call.

frontend: Enable multitrack video controls on Linux

326db9f

lexano-ivs force-pushed the lexano/teb-linux branch from dd12c61 to 326db9f Compare January 31, 2025 15:55

RytoEX requested changes Feb 12, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Enhanced Broadcasting support for Linux #11541

Add Enhanced Broadcasting support for Linux #11541

lexano-ivs commented Nov 20, 2024 •

edited

Loading

tytan652 Nov 20, 2024 •

edited

Loading

lexano-ivs Nov 22, 2024

tytan652 Nov 22, 2024

lexano-ivs Nov 22, 2024

lexano-ivs Dec 18, 2024

Lain-B left a comment

Lain-B Nov 24, 2024

lexano-ivs Dec 18, 2024

RytoEX Feb 14, 2025

Lain-B Nov 24, 2024

Lain-B Nov 24, 2024

gxalpha Nov 25, 2024

lexano-ivs Dec 18, 2024

RytoEX Feb 14, 2025

tytan652 Dec 19, 2024 •

edited

Loading

lexano-ivs Jan 6, 2025

RytoEX Feb 14, 2025

mihawk90 commented Jan 19, 2025 •

edited

Loading

dsaedtler commented Jan 24, 2025

RytoEX Nov 20, 2024

RytoEX Nov 20, 2024

RytoEX Nov 20, 2024

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

RytoEX Feb 12, 2025

		while (!!rc_mode->name && strcmp(rc_mode->name, name) != 0)
		while (!!rc_mode->name && astrcmpi(rc_mode->name, name) != 0)

Add Enhanced Broadcasting support for Linux #11541

Are you sure you want to change the base?

Add Enhanced Broadcasting support for Linux #11541

Conversation

lexano-ivs commented Nov 20, 2024 • edited Loading

Description

Motivation and Context

How Has This Been Tested?

Types of changes

Checklist:

tytan652 Nov 20, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Lain-B left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tytan652 Dec 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mihawk90 commented Jan 19, 2025 • edited Loading

dsaedtler commented Jan 24, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lexano-ivs commented Nov 20, 2024 •

edited

Loading

tytan652 Nov 20, 2024 •

edited

Loading

tytan652 Dec 19, 2024 •

edited

Loading

mihawk90 commented Jan 19, 2025 •

edited

Loading