Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

.finish() stage of the shader compilation segfaults on NVIDIA Vulkan driver #59

Open
mikidep opened this issue Jul 17, 2022 · 1 comment
Open

.finish() stage of the shader compilation segfaults on NVIDIA Vulkan driver #59

mikidep opened this issue Jul 17, 2022 · 1 comment

Comments

@mikidep
Copy link

mikidep commented Jul 17, 2022 •
edited
Loading

I am running on Ubuntu 22.04 with emu_core 0.1.1, info()?.name is "NVIDIA GeForce RTX 3050 Ti Laptop GPU", the driver is version 515 of the official NVIDIA Linux driver, installed through APT.

The problem seems to be related to the presence of a storage buffer, the one called prec_mat: if I remove it in both in the shader and in the SpirvBuilder, the issue does not manifest. I am using rust-gpu to write my shader. Note that if my integrated AMD GPU is selected, the code runs fine.

Below is a comprehensive stack trace:

___lldb_unnamed_symbol462 (@___lldb_unnamed_symbol462:301)
___lldb_unnamed_symbol11106 (@___lldb_unnamed_symbol11106:2200)
___lldb_unnamed_symbol11107 (@___lldb_unnamed_symbol11107:19)
___lldb_unnamed_symbol16036 (@___lldb_unnamed_symbol16036:120)
___lldb_unnamed_symbol11528 (@___lldb_unnamed_symbol11528:60)
___lldb_unnamed_symbol11308 (@___lldb_unnamed_symbol11308:258)
_nv002nvvm (@_nv002nvvm:11)
___lldb_unnamed_symbol58166 (@___lldb_unnamed_symbol58166:66)
___lldb_unnamed_symbol58168 (@___lldb_unnamed_symbol58168:583)
___lldb_unnamed_symbol58169 (@___lldb_unnamed_symbol58169:146)
___lldb_unnamed_symbol58181 (@___lldb_unnamed_symbol58181:164)
___lldb_unnamed_symbol58182 (@___lldb_unnamed_symbol58182:8)
___lldb_unnamed_symbol58172 (@___lldb_unnamed_symbol58172:148)
___lldb_unnamed_symbol58204 (@___lldb_unnamed_symbol58204:91)
___lldb_unnamed_symbol57964 (@___lldb_unnamed_symbol57964:70)
___lldb_unnamed_symbol57965 (@___lldb_unnamed_symbol57965:28)
ash::vk::features::DeviceFnV1_0::create_compute_pipelines (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/ash-0.31.0/src/vk/features.rs:5094)
gfx_backend_vulkan::device::<impl gfx_hal::device::Device<gfx_backend_vulkan::Backend> for gfx_backend_vulkan::Device>::create_compute_pipeline (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/gfx-backend-vulkan-0.5.11/src/device.rs:1044)
wgpu_core::device::<impl wgpu_core::hub::Global<G>>::device_create_compute_pipeline (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-core-0.5.6/src/device/mod.rs:1932)
wgpu_device_create_compute_pipeline (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-native-0.5.1/src/device.rs:347)
wgpu::Device::create_compute_pipeline (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/wgpu-0.5.2/src/lib.rs:906)
emu_core::device::Device::compile (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/emu_core-0.1.1/src/device.rs:611)
emu_core::compile::SpirvOrFinished<P,C>::finish (/home/mikidep/.cargo/registry/src/github.com-1ecc6299db9ec823/emu_core-0.1.1/src/compile.rs:305)
scene_emu::main (/home/mikidep/Documenti/Codice/scene-emu/src/main.rs:104)
core::ops::function::FnOnce::call_once (@core::ops::function::FnOnce::call_once:6)
std::sys_common::backtrace::__rust_begin_short_backtrace (@std::sys_common::backtrace::__rust_begin_short_backtrace:6)
std::rt::lang_start::{{closure}} (@std::rt::lang_start::{{closure}}:7)
core::ops::function::impls::<impl core::ops::function::FnOnce<A> for &F>::call_once (@std::rt::lang_start_internal:184)
std::panicking::try::do_call (@std::rt::lang_start_internal:183)
std::panicking::try (@std::rt::lang_start_internal:183)
std::panic::catch_unwind (@std::rt::lang_start_internal:183)
std::rt::lang_start_internal::{{closure}} (@std::rt::lang_start_internal:183)
std::panicking::try::do_call (@std::rt::lang_start_internal:183)
std::panicking::try (@std::rt::lang_start_internal:183)
std::panic::catch_unwind (@std::rt::lang_start_internal:183)
std::rt::lang_start_internal (@std::rt::lang_start_internal:183)
std::rt::lang_start (@std::rt::lang_start:13)
main (@main:10)
__libc_start_call_main (@__libc_start_call_main:29)
__libc_start_main_impl (@__libc_start_main@@GLIBC_2.34:43)
_start (@_start:15)

I am also attaching relevant Rust code and disassembled shader SPIR-V code:

  • main.rs.txt - This is the entrypoint of my main crate, which uses Emu;
  • stack.rs.txt - This is a module shared between my shader and my main crate, which is used in shader parameter and DeviceBox definitions;
  • lib.rs.txt - This is the entrypoint of my shader, where the accepted storage buffers are listed;
  • main_shader_disass.txt - This is the disassembled version of the SPIR-V shader compiled with rust-gpu.

Below are extracts from the above source files, in which the incriminated parameter is declared:

(in main.rs)

    let spirv = SpirvBuilder::new()
        .set_entry_point_name("main")
        .add_param_mut::<[u32]>() // alpha
        .add_param_mut::<[StackSym]>() // stack
        .add_param_mut::<[usize]>() // gives_stack
        .add_param_mut::<[u32]>() // prec_mat
        .add_param::<usize>() // length
        .add_param::<usize>() // chunk_size
        .add_param::<u32>() // term_thresh
        .set_code_with_u8(std::io::Cursor::new(code))?
        .build();
    let c = compile::<Spirv<_>, SpirvCompile, _, GlobalCache>(spirv)?.finish()?;

Segfault happens on the last line.

(in lib.rs)

#[spirv(compute(threads(4)))]
pub fn main(
    #[spirv(global_invocation_id)] id: UVec3,
    #[spirv(storage_buffer, descriptor_set = 0, binding = 0)] alpha: &mut [u32],
    #[spirv(storage_buffer, descriptor_set = 0, binding = 1)] stack: &mut [StackSym],
    #[spirv(storage_buffer, descriptor_set = 0, binding = 2)] gives_stack: &mut [usize],
    #[spirv(storage_buffer, descriptor_set = 0, binding = 3)] prec_mat: &mut [u32],
    #[spirv(storage_buffer, descriptor_set = 0, binding = 4)] length: &mut usize,
    #[spirv(storage_buffer, descriptor_set = 0, binding = 5)] chunk_size: &mut usize,
    #[spirv(storage_buffer, descriptor_set = 0, binding = 6)] term_thresh: &mut u32,
) { // ...

I understand that the issue should be related to NVIDIA's Vulkan implementation, but maybe you know something about this kind of issue. Thank you in advance.
@calebwin
Copy link
Owner

Interesting, thanks for reporting the issue. Not sure if it can be fixed within Emu though. May be an issue in WGPU or the Vulkan implementation.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@mikidep @calebwin