H.264 videos have artifacts when rendered using a Braswell GPU on Chromium-based browsers.

[Smu1zel]

When attempting to decode an H.264 video on Chromium-based browsers, artifacts will begin to appear. The issue becomes much more pronounced the more "fast-paced" the video is. Higher resolutions tend to exhibit this much more often as well, but even 240p can have issues.

Info

Browser: Opera/Chromium (both tested)
Platform: Wayland
Driver version: Latest from master
Inxi -FG output:

  Host: lynden-aspirer3131t Kernel: 6.12.44-1-lts arch: x86_64 bits: 64
  Desktop: KDE Plasma v: 6.4.4 Distro: EndeavourOS
Machine:
  Type: Laptop System: Acer product: Aspire R3-131T v: V1.09
    serial: <superuser required>
  Mobo: Acer model: Aspire R3-131T v: V1.09 serial: <superuser required>
    UEFI: Insyde v: 1.09 date: 07/21/2015
Battery:
  ID-1: BAT0 charge: N/A condition: N/A
CPU:
  Info: dual core model: Intel Celeron N3050 bits: 64 type: MCP cache:
    L2: 2 MiB
  Speed (MHz): avg: 2160 min/max: 480/2160 cores: 1: 2160 2: 2160
Graphics:
  Device-1: Intel Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx
    Integrated Graphics driver: i915 v: kernel
  Device-2: Realtek HD Webcam - driver: uvcvideo type: USB
  Display: wayland server: X.org v: 1.21.1.18 with: Xwayland v: 24.1.8
    compositor: kwin_wayland driver: X: loaded: modesetting dri: crocus
    gpu: i915 resolution: 1366x768~60Hz
  API: EGL v: 1.5 drivers: crocus,swrast
    platforms: gbm,wayland,x11,surfaceless,device
  API: OpenGL v: 4.6 compat-v: 4.5 vendor: intel mesa v: 25.2.1-arch1.5
    renderer: Mesa Intel HD Graphics 400 (BSW)
  API: Vulkan v: 1.4.321 drivers: intel surfaces: N/A
  Info: Tools: api: clinfo, eglinfo, glxinfo, vulkaninfo
    de: kscreen-console,kscreen-doctor gpu: gputop, intel_gpu_top, lsgpu
    wl: wayland-info x11: xdpyinfo, xprop, xrandr
Audio:
  Device-1: Intel Atom/Celeron/Pentium Processor x5-E8000/J3xxx/N3xxx Series
    High Definition Audio driver: snd_hda_intel
  API: ALSA v: k6.12.44-1-lts status: kernel-api
  Server-1: PipeWire v: 1.4.7 status: active
Network:
  Device-1: Realtek RTL8111/8168/8211/8411 PCI Express Gigabit Ethernet
    driver: r8169
  IF: enp1s0 state: down mac: 30:65:ec:8e:a6:20
  Device-2: Intel Wireless 3165 driver: iwlwifi
  IF: wlan0 state: up mac: 94:65:9c:01:62:c4
Bluetooth:
  Device-1: Intel Bluetooth wireless interface driver: btusb type: USB
  Report: btmgmt ID: hci0 state: up address: N/A
Drives:
  Local Storage: total: 465.76 GiB used: 16.06 GiB (3.4%)
  ID-1: /dev/sda vendor: Seagate model: ST500LT012-9WS142 size: 465.76 GiB
Partition:
  ID-1: / size: 97.66 GiB used: 16.04 GiB (16.4%) fs: btrfs dev: /dev/sda3
  ID-2: /boot/efi size: 96 MiB used: 26.2 MiB (27.3%) fs: vfat
    dev: /dev/sda1
  ID-3: /home size: 97.66 GiB used: 16.04 GiB (16.4%) fs: btrfs
    dev: /dev/sda3
  ID-4: /var/log size: 97.66 GiB used: 16.04 GiB (16.4%) fs: btrfs
    dev: /dev/sda3
Swap:
  Alert: No swap data was found.
Sensors:
  System Temperatures: cpu: 46.0 C mobo: N/A
  Fan Speeds (rpm): N/A
Info:
  Memory: total: 4 GiB note: est. available: 3.65 GiB used: 2.33 GiB (63.7%)
  Processes: 216 Uptime: 39m Shell: Bash inxi: 3.3.38

Screenshots

PS

Sorry for taking so long to make this. I kept jumping around unrelated projects. I'll update this post if I find any new information.

[irql-notlessorequal]

I think I had this issue too on a Cherry Trail x86 tablet. If I recall correctly, setting Ozone to Wayland and adding AcceleratedVideoDecodeLinuxZeroCopyGL on top of AcceleratedVideoDecodeLinuxGL fixed it since it bypassed the VPP layer in the driver which seems to just be completely broken on low-power hardware.

[Smu1zel]

I think I had this issue too on a Cherry Trail x86 tablet. If I recall correctly, setting Ozone to Wayland and adding AcceleratedVideoDecodeLinuxZeroCopyGL on top of AcceleratedVideoDecodeLinuxGL fixed it since it bypassed the VPP layer in the driver which seems to just be completely broken on low-power hardware.

I added that flag and that indeed works around the issue. I can stress the system all I want and nothing ever corrupts. Intel_gpu_top does report that the Video engine is at work as well, so I know it's not just Chromium tricking me again.

I do wonder if this is a regression that was never fixed, or a bug that has always existed that Chromium just happens to trigger. Given that ChromeOS exists, this had to have worked at some point for them, but they use VA-API a bit differently (they might zero-copy by default, I'm not sure), so they also could've just sidestepped the bug and never had to fix it. The Braswell Chromebooks also only run up to 102, which uses the old backend (I've been trying to get 109 on Linux to work with acceleration enabled, but no success so far).

[luisfer153]

any flag combination suggest for xorg?

[irql-notlessorequal]

Not to my knowledge since the zero-copy flag is limited to Wayland as it uses EGL over GLX, I believe Chrome still blocks EGL from being used on X11 with Intel for some reason (still)

You could try with Vulkan as it might work (haven't checked) if your iGPU is Haswell-based or newer.

--enable-features=Vulkan,DefaultANGLEVulkan,VulkanFromANGLE should get Chrome to use Vulkan for everything

[Smu1zel]

Not sure if this helps any, but I've recently acquired a Bay Trail (Celeron N2840) device, and this issue doesn't affect it. As far as I can tell, this driver works perfectly with this thing. I expected this line of low-end SoCs to have the same issues as the others, but apparently not.