Two storage buffer bindings point to the same buffer

[robostep9829]

Description
I'm trying to access vertex and index buffers from fragment shader. I tried to get vertex world position without getting data from vertex shader, only using acceleration structure. For that i needed to get primitive id by using RayQuery, then get vertex id based on primitive id and then get position of that vertex. So i added two buffer bindings to my render pipeline - one for indices and another for vertices. They are pointers to index and vertex buffers respectively.
Honorable mention: #1040. I think this is relevant, because i'm trying to replicate some concepts from vk_mini_path_tracer (See ./checkpoints/11_specular/. The shader makes use of two storage buffers which are pointing to vertex and index buffers).

Repro steps
This is probably runnable:
https://gitlab.com/robostep9829/hal-path-tracing-rs/-/tree/ssbo-error
See ./shaders/shader.frag, ./src/main.rs and ./src/resources.rs

Expected vs observed behavior
I noticed that no matter what value of vertex buffer i'm trying to access - it's always zero, however index values are proper and indbuf[rayQueryGetIntersectionPrimitiveIndexEXT(rq, true)] returns correct vertex number. I investigated further and found out that layout(set = 0, binding = 2) buffer ind_buf { uint indbuf[]; }; and layout(set = 0, binding = 3) buffer vtx_buf { Vertex vtxbuf[]; }; point to the same address (see nsight screenshot). In my program it is stated like this:

&wgpu_hal::BindGroupLayoutDescriptor {
                    label: Some("Bind Group layout"),
                    flags: wgpu_hal::BindGroupLayoutFlags::empty(),
                    entries: &[
                        wgt::BindGroupLayoutEntry {
                            binding: 0,
                            visibility: wgt::ShaderStages::FRAGMENT,
                            ty: wgt::BindingType::Texture {
                                sample_type: wgt::TextureSampleType::Float { filterable: true },
                                view_dimension: wgt::TextureViewDimension::D2,
                                multisampled: false,
                            },
                            count: None,
                        },
                        wgt::BindGroupLayoutEntry {
                            binding: 1,
                            visibility: wgt::ShaderStages::FRAGMENT,
                            ty: wgt::BindingType::Sampler(wgt::SamplerBindingType::Filtering),
                            count: None,
                        },
                        wgt::BindGroupLayoutEntry{
                            binding: 2,
                            visibility: wgt::ShaderStages::FRAGMENT,
                            ty: wgt::BindingType::Buffer {
                                ty: wgt::BufferBindingType::Storage { read_only: true },
                                has_dynamic_offset: false,
                                min_binding_size: None,
                            },
                            count: None,
                        },
                        wgt::BindGroupLayoutEntry{
                            binding: 3,
                            visibility: wgt::ShaderStages::FRAGMENT,
                            ty: wgt::BindingType::Buffer {
                                ty: wgt::BufferBindingType::Storage { read_only: true },
                                has_dynamic_offset: false,
                                min_binding_size: None,
                            },
                            count: None,
                        }
                    ],
                }

&wgpu_hal::BindGroupDescriptor {
                    layout,
                    buffers: &[
                        wgpu_hal::BufferBinding {
                            buffer: &index_buffer,
                            offset: 0,
                            size: None,
                        },
                        wgpu_hal::BufferBinding {
                            buffer: &vertex_buffer,
                            offset: 0,
                            size: None,
                        },
                    ],
                    samplers: &[&diffuse_texture.sampler],
                    textures: &[wgpu_hal::TextureBinding {
                        view: &diffuse_texture.view,
                        usage: wgpu_hal::TextureUses::RESOURCE,
                    }],
                    entries: &[
                        wgpu_hal::BindGroupEntry {
                            binding: 0,
                            resource_index: 0,
                            count: 1,
                        },
                        wgpu_hal::BindGroupEntry {
                            binding: 1,
                            resource_index: 0,
                            count: 1,
                        },
                        wgpu_hal::BindGroupEntry {
                            binding: 2,
                            resource_index: 0,
                            count: 1,
                        },
                        wgpu_hal::BindGroupEntry {
                            binding: 3,
                            resource_index: 0,
                            count: 1,
                        }
                    ],
                    label: Some(&format!("{} Bind group", m.name)),
                    acceleration_structures: &[],
                }

Extra materials


Here we can see that ind_buf and vtx_buf are the same buffer

Graphics-wgpu__2024_03_09__19_25_05.zip

Platform
OS: Windows 11 22H2, GPU: Radeon RX 6600, IDE: RustRover 233.14015.155, Vulkan SDK: 1.3.275.0, Rust: 1.71-msvc.

[robostep9829]

Sorry for congestion. The solution was to change resource_index in the fourth BindGroupEntry. Is it worth documenting? Like if you want to bind multiple instances of the same binding type you should mark them with respective resource index.