Bindless Tests (#6732)

* Move Partial Binding into Own File

* Texture Bindless Test

* Make It Work

* Tests

* Uniform Buffers

* BadCode

* Bugs!

* Exclude llvmpipe

* Combine Partial Binding Test

* MVK Issue

* Sampler Array Tests

* Make All Tests Partially Bound As Well

Author: cwfitzgerald

tests/Cargo.toml

@@ -34,6 +34,7 @@ ctor.workspace = true
 futures-lite.workspace = true
 glam.workspace = true
 itertools.workspace = true
+image.workspace = true
 libtest-mimic.workspace = true
 log.workspace = true
 parking_lot.workspace = true

tests/tests/binding_array/buffers.rs

@@ -0,0 +1,265 @@
+use std::num::{NonZeroU32, NonZeroU64};
+
+use wgpu::*;
+use wgpu_test::{gpu_test, FailureCase, GpuTestConfiguration, TestParameters, TestingContext};
+
+#[gpu_test]
+static BINDING_ARRAY_UNIFORM_BUFFERS: GpuTestConfiguration = GpuTestConfiguration::new()
+    .parameters(
+        TestParameters::default()
+            .features(
+                Features::BUFFER_BINDING_ARRAY
+                    | Features::UNIFORM_BUFFER_AND_STORAGE_TEXTURE_ARRAY_NON_UNIFORM_INDEXING,
+            )
+            .limits(Limits {
+                max_uniform_buffers_per_shader_stage: 16,
+                ..Limits::default()
+            })
+            // Naga bug on vulkan: https://github.com/gfx-rs/wgpu/issues/6733
+            //
+            // Causes varying errors on different devices, so we don't match more closely.
+            .expect_fail(FailureCase::backend(Backends::VULKAN))
+            // These issues cause a segfault on lavapipe
+            .skip(FailureCase::backend_adapter(Backends::VULKAN, "llvmpipe")),
+    )
+    .run_async(|ctx| async move { binding_array_buffers(ctx, BufferType::Uniform, false).await });
+
+#[gpu_test]
+static PARTIAL_BINDING_ARRAY_UNIFORM_BUFFERS: GpuTestConfiguration = GpuTestConfiguration::new()
+    .parameters(
+        TestParameters::default()
+            .features(
+                Features::BUFFER_BINDING_ARRAY
+                    | Features::PARTIALLY_BOUND_BINDING_ARRAY
+                    | Features::UNIFORM_BUFFER_AND_STORAGE_TEXTURE_ARRAY_NON_UNIFORM_INDEXING,
+            )
+            .limits(Limits {
+                max_uniform_buffers_per_shader_stage: 32,
+                ..Limits::default()
+            })
+            // Naga bug on vulkan: https://github.com/gfx-rs/wgpu/issues/6733
+            //
+            // Causes varying errors on different devices, so we don't match more closely.
+            .expect_fail(FailureCase::backend(Backends::VULKAN))
+            // These issues cause a segfault on lavapipe
+            .skip(FailureCase::backend_adapter(Backends::VULKAN, "llvmpipe")),
+    )
+    .run_async(|ctx| async move { binding_array_buffers(ctx, BufferType::Uniform, true).await });
+
+#[gpu_test]
+static BINDING_ARRAY_STORAGE_BUFFERS: GpuTestConfiguration = GpuTestConfiguration::new()
+    .parameters(
+        TestParameters::default()
+            .features(
+                Features::BUFFER_BINDING_ARRAY
+                    | Features::STORAGE_RESOURCE_BINDING_ARRAY
+                    | Features::SAMPLED_TEXTURE_AND_STORAGE_BUFFER_ARRAY_NON_UNIFORM_INDEXING,
+            )
+            .limits(Limits {
+                max_storage_buffers_per_shader_stage: 17,
+                ..Limits::default()
+            })
+            // See https://github.com/gfx-rs/wgpu/issues/6745.
+            .expect_fail(FailureCase::molten_vk()),
+    )
+    .run_async(|ctx| async move { binding_array_buffers(ctx, BufferType::Storage, false).await });
+
+#[gpu_test]
+static PARTIAL_BINDING_ARRAY_STORAGE_BUFFERS: GpuTestConfiguration = GpuTestConfiguration::new()
+    .parameters(
+        TestParameters::default()
+            .features(
+                Features::BUFFER_BINDING_ARRAY
+                    | Features::PARTIALLY_BOUND_BINDING_ARRAY
+                    | Features::STORAGE_RESOURCE_BINDING_ARRAY
+                    | Features::SAMPLED_TEXTURE_AND_STORAGE_BUFFER_ARRAY_NON_UNIFORM_INDEXING,
+            )
+            .limits(Limits {
+                max_storage_buffers_per_shader_stage: 33,
+                ..Limits::default()
+            })
+            // See https://github.com/gfx-rs/wgpu/issues/6745.
+            .expect_fail(FailureCase::molten_vk()),
+    )
+    .run_async(|ctx| async move { binding_array_buffers(ctx, BufferType::Storage, true).await });
+
+enum BufferType {
+    Storage,
+    Uniform,
+}
+
+async fn binding_array_buffers(
+    ctx: TestingContext,
+    buffer_type: BufferType,
+    partial_binding: bool,
+) {
+    let storage_mode = match buffer_type {
+        BufferType::Storage => "storage",
+        BufferType::Uniform => "uniform",
+    };
+
+    let shader = r#"
+        struct ImAU32 {
+            value: u32,
+            _padding: u32,
+            _padding2: u32,
+            _padding3: u32,
+        };
+
+        @group(0) @binding(0)
+        var<{storage_mode}> buffers: binding_array<ImAU32>;
+
+        @group(0) @binding(1)
+        var<storage, read_write> output_buffer: array<u32>;
+
+        @compute
+        @workgroup_size(16, 1, 1)
+        fn compMain(@builtin(global_invocation_id) id: vec3u) {
+            output_buffer[id.x] = buffers[id.x].value;
+        }
+    "#;
+    let shader = shader.replace("{storage_mode}", storage_mode);
+
+    let module = ctx
+        .device
+        .create_shader_module(wgpu::ShaderModuleDescriptor {
+            label: Some("Binding Array Buffer"),
+            source: wgpu::ShaderSource::Wgsl(shader.into()),
+        });
+
+    let image = image::load_from_memory(include_bytes!("../3x3_colors.png")).unwrap();
+    // Resize image to 4x4
+    let image = image
+        .resize_exact(4, 4, image::imageops::FilterType::Gaussian)
+        .into_rgba8();
+
+    // Create one buffer for each pixel
+    let mut buffers = Vec::with_capacity(64);
+    for data in image.pixels() {
+        let buffer = ctx.device.create_buffer(&BufferDescriptor {
+            label: None,
+            usage: match buffer_type {
+                BufferType::Storage => BufferUsages::STORAGE | BufferUsages::COPY_DST,
+                BufferType::Uniform => BufferUsages::UNIFORM | BufferUsages::COPY_DST,
+            },
+            // 16 to allow padding for uniform buffers
+            size: 16,
+            mapped_at_creation: true,
+        });
+        buffer.slice(..).get_mapped_range_mut()[0..4].copy_from_slice(&data.0);
+        buffer.unmap();
+        buffers.push(buffer);
+    }
+
+    let output_buffer = ctx.device.create_buffer(&BufferDescriptor {
+        label: None,
+        size: 4 * 4 * 4,
+        usage: BufferUsages::STORAGE | BufferUsages::COPY_SRC,
+        mapped_at_creation: false,
+    });
+
+    let multiplier = if partial_binding { 2 } else { 1 };
+
+    let bind_group_layout = ctx
+        .device
+        .create_bind_group_layout(&BindGroupLayoutDescriptor {
+            label: Some("Bind Group Layout"),
+            entries: &[
+                BindGroupLayoutEntry {
+                    binding: 0,
+                    visibility: ShaderStages::COMPUTE,
+                    ty: BindingType::Buffer {
+                        ty: match buffer_type {
+                            BufferType::Storage => BufferBindingType::Storage { read_only: true },
+                            BufferType::Uniform => BufferBindingType::Uniform,
+                        },
+                        has_dynamic_offset: false,
+                        min_binding_size: Some(NonZeroU64::new(16).unwrap()),
+                    },
+                    count: Some(NonZeroU32::new(16 * multiplier).unwrap()),
+                },
+                BindGroupLayoutEntry {
+                    binding: 1,
+                    visibility: ShaderStages::COMPUTE,
+                    ty: BindingType::Buffer {
+                        ty: BufferBindingType::Storage { read_only: false },
+                        has_dynamic_offset: false,
+                        min_binding_size: Some(NonZeroU64::new(4).unwrap()),
+                    },
+                    count: None,
+                },
+            ],
+        });
+
+    let buffer_references: Vec<_> = buffers
+        .iter()
+        .map(|b| b.as_entire_buffer_binding())
+        .collect();
+
+    let bind_group = ctx.device.create_bind_group(&BindGroupDescriptor {
+        label: Some("Bind Group"),
+        layout: &bind_group_layout,
+        entries: &[
+            BindGroupEntry {
+                binding: 0,
+                resource: BindingResource::BufferArray(&buffer_references),
+            },
+            BindGroupEntry {
+                binding: 1,
+                resource: output_buffer.as_entire_binding(),
+            },
+        ],
+    });
+
+    let pipeline_layout = ctx
+        .device
+        .create_pipeline_layout(&PipelineLayoutDescriptor {
+            label: Some("Pipeline Layout"),
+            bind_group_layouts: &[&bind_group_layout],
+            push_constant_ranges: &[],
+        });
+
+    let pipeline = ctx
+        .device
+        .create_compute_pipeline(&ComputePipelineDescriptor {
+            label: Some("Compute Pipeline"),
+            layout: Some(&pipeline_layout),
+            module: &module,
+            entry_point: Some("compMain"),
+            compilation_options: Default::default(),
+            cache: None,
+        });
+
+    let mut encoder = ctx
+        .device
+        .create_command_encoder(&CommandEncoderDescriptor { label: None });
+    {
+        let mut render_pass = encoder.begin_compute_pass(&ComputePassDescriptor {
+            label: None,
+            timestamp_writes: None,
+        });
+        render_pass.set_pipeline(&pipeline);
+        render_pass.set_bind_group(0, &bind_group, &[]);
+        render_pass.dispatch_workgroups(1, 1, 1);
+    }
+
+    let readback_buffer = ctx.device.create_buffer(&BufferDescriptor {
+        label: None,
+        size: 4 * 4 * 4,
+        usage: BufferUsages::MAP_READ | BufferUsages::COPY_DST,
+        mapped_at_creation: false,
+    });
+
+    encoder.copy_buffer_to_buffer(&output_buffer, 0, &readback_buffer, 0, 4 * 4 * 4);
+
+    ctx.queue.submit(Some(encoder.finish()));
+
+    let slice = readback_buffer.slice(..);
+    slice.map_async(MapMode::Read, |_| {});
+
+    ctx.device.poll(Maintain::Wait);
+
+    let data = slice.get_mapped_range();
+
+    assert_eq!(&data[..], &*image);
+}

tests/tests/binding_array/mod.rs

@@ -0,0 +1,4 @@
+mod buffers;
+mod sampled_textures;
+mod samplers;
+mod storage_textures;