Core(Device(Lost)) error when launching with nsight graphics GPU profiler

[SarthakSingh31]

Description
Launching the wgpu-examples from NVIDIA Nsight Graphics with the GPU Trace Profiler Activity leads to the application crashing due to RequestDeviceError { inner: Core(Device(Lost)) } error when requesting the device when forcing the vulkan backend.

2024-12-21T09:14:52.470204Z  INFO bevy_render::renderer: AdapterInfo { name: "NVIDIA GeForce GTX 960", vendor: 4318, device: 5121, device_type: DiscreteGpu, driver: "NVIDIA", driver_info: "565.77", backend: Vulkan }
thread 'main' panicked at /home/ssingh/.cargo/registry/src/index.crates.io-6f17d22bba15001f/bevy_render-0.15.0/src/renderer/mod.rs:363:10:
called `Result::unwrap()` on an `Err` value: RequestDeviceError { inner: Core(Device(Lost)) }
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

When I allow it to use any backend it errors out with

thread 'main' panicked at /home/ssingh/.cargo/registry/src/index.crates.io-6f17d22bba15001f/khronos-egl-6.0.0/src/lib.rs:1095:27:
called `Option::unwrap()` on a `None` value
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::panicking::panic
   3: core::option::unwrap_failed
   4: <wgpu_hal::gles::egl::Instance as wgpu_hal::Instance>::enumerate_adapters
   5: <I as wgpu_hal::dynamic::instance::DynInstance>::enumerate_adapters
   6: wgpu_core::instance::Instance::request_adapter
   7: wgpu_core::instance::<impl wgpu_core::global::Global>::request_adapter
   8: <T as wgpu::context::DynContext>::instance_request_adapter
   9: wgpu::api::instance::Instance::request_adapter
  10: futures_lite::future::block_on
  11: <bevy_render::RenderPlugin as bevy_app::plugin::Plugin>::build
  12: bevy_app::app::App::add_boxed_plugin
  13: bevy_app::plugin_group::PluginGroupBuilder::finish
  14: <(S0,S1,S2) as bevy_app::plugin::sealed::Plugins<(bevy_app::plugin::sealed::PluginsTupleMarker,P0,P1,P2)>>::add_to_app
  15: bevy_app::app::App::add_plugins
  16: simple::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

Repro steps

1. Launch NVIDIA Nsight Graphics (version: 2024.3.0.0).
2. Click on start activity.
3. Fill in the application path and application workspace.
4. Select the GPU Trace Profiler from the activity menu on the bottom left.
5. Click on Launch GPU Trace button at the bottom right.

Expected vs observed behavior
Expected: The application runs without any issues and I can profile the shader.
Observed: The application crashes and fails to request a device.

Platform
Platform: Arch Linux (Linux version 6.12.4-arch1-1 (linux@archlinux) (gcc (GCC) 14.2.1 20240910, GNU ld (GNU Binutils) 2.43.0) #1 SMP PREEMPT_DYNAMIC Mon, 09 Dec 2024 14:31:57 +0000)
Wgpu: main branch
GPU: GTX 960 6GB

[SarthakSingh31]

If anyone could recommend a different shader profiler that works that would also unblock me. I just started using NVIDIA Nsight Graphics so I am not married to it.

[cwfitzgerald]

I have no idea what would cause this - but I know that nsight doesn't support profiling vulkan on 10xx or older, they only support 20xx or newer.

I think this isn't our issue.

As for a recommendation, I would look into tracy and tracy's gpu spans. You can use https://github.com/Wumpf/wgpu-profiler to hook into tracy's gpu infrastructure.