Mesh shaders - initial wgpu hal changes

CHANGELOG.md

@@ -130,6 +130,8 @@ By @jamienicol in [#6929](https://github.com/gfx-rs/wgpu/pull/6929) and [#7080](
 
 ### New Features
 
+- Added mesh shader support to `wgpu_hal`. By @SupaMaggie70Incorporated in [#7089](https://github.com/gfx-rs/wgpu/pull/7089)
+
 #### General
 
 - It is now possible to create a dummy `wgpu` device even when no GPU is available. This may be useful for testing of code which manages graphics resources. Currently, it supports reading and writing buffers, and other resource types can be created but do nothing.

wgpu-hal/src/dx12/command.rs

@@ -1179,6 +1179,14 @@ impl crate::CommandEncoder for super::CommandEncoder {
             )
         }
     }
+    unsafe fn draw_mesh_tasks(
+        &mut self,
+        _group_count_x: u32,
+        _group_count_y: u32,
+        _group_count_z: u32,
+    ) {
+        unreachable!()
+    }
     unsafe fn draw_indirect(
         &mut self,
         buffer: &super::Buffer,
@@ -1215,6 +1223,14 @@ impl crate::CommandEncoder for super::CommandEncoder {
             )
         }
     }
+    unsafe fn draw_mesh_tasks_indirect(
+        &mut self,
+        _buffer: &<Self::A as crate::Api>::Buffer,
+        _offset: wgt::BufferAddress,
+        _draw_count: u32,
+    ) {
+        unreachable!()
+    }
     unsafe fn draw_indirect_count(
         &mut self,
         buffer: &super::Buffer,
@@ -1255,6 +1271,16 @@ impl crate::CommandEncoder for super::CommandEncoder {
             )
         }
     }
+    unsafe fn draw_mesh_tasks_indirect_count(
+        &mut self,
+        _buffer: &<Self::A as crate::Api>::Buffer,
+        _offset: wgt::BufferAddress,
+        _count_buffer: &<Self::A as crate::Api>::Buffer,
+        _count_offset: wgt::BufferAddress,
+        _max_count: u32,
+    ) {
+        unreachable!()
+    }
 
     // compute
 

wgpu-hal/src/dx12/device.rs

@@ -1883,6 +1883,18 @@ impl crate::Device for super::Device {
             vertex_strides,
         })
     }
+
+    unsafe fn create_mesh_pipeline(
+        &self,
+        _desc: &crate::MeshPipelineDescriptor<
+            <Self::A as crate::Api>::PipelineLayout,
+            <Self::A as crate::Api>::ShaderModule,
+            <Self::A as crate::Api>::PipelineCache,
+        >,
+    ) -> Result<<Self::A as crate::Api>::RenderPipeline, crate::PipelineError> {
+        unreachable!()
+    }
+
     unsafe fn destroy_render_pipeline(&self, _pipeline: super::RenderPipeline) {
         self.counters.render_pipelines.sub(1);
     }

wgpu-hal/src/dynamic/command.rs

@@ -130,6 +130,12 @@ pub trait DynCommandEncoder: DynResource + core::fmt::Debug {
         first_instance: u32,
         instance_count: u32,
     );
+    unsafe fn draw_mesh_tasks(
+        &mut self,
+        group_count_x: u32,
+        group_count_y: u32,
+        group_count_z: u32,
+    );
     unsafe fn draw_indirect(
         &mut self,
         buffer: &dyn DynBuffer,
@@ -142,6 +148,12 @@ pub trait DynCommandEncoder: DynResource + core::fmt::Debug {
         offset: wgt::BufferAddress,
         draw_count: u32,
     );
+    unsafe fn draw_mesh_tasks_indirect(
+        &mut self,
+        buffer: &dyn DynBuffer,
+        offset: wgt::BufferAddress,
+        draw_count: u32,
+    );
     unsafe fn draw_indirect_count(
         &mut self,
         buffer: &dyn DynBuffer,
@@ -158,6 +170,14 @@ pub trait DynCommandEncoder: DynResource + core::fmt::Debug {
         count_offset: wgt::BufferAddress,
         max_count: u32,
     );
+    unsafe fn draw_mesh_tasks_indirect_count(
+        &mut self,
+        buffer: &dyn DynBuffer,
+        offset: wgt::BufferAddress,
+        count_buffer: &dyn DynBuffer,
+        count_offset: wgt::BufferAddress,
+        max_count: u32,
+    );
 
     unsafe fn begin_compute_pass(&mut self, desc: &ComputePassDescriptor<dyn DynQuerySet>);
     unsafe fn end_compute_pass(&mut self);
@@ -473,6 +493,15 @@ impl<C: CommandEncoder + DynResource> DynCommandEncoder for C {
         };
     }
 
+    unsafe fn draw_mesh_tasks(
+        &mut self,
+        group_count_x: u32,
+        group_count_y: u32,
+        group_count_z: u32,
+    ) {
+        unsafe { C::draw_mesh_tasks(self, group_count_x, group_count_y, group_count_z) };
+    }
+
     unsafe fn draw_indirect(
         &mut self,
         buffer: &dyn DynBuffer,
@@ -493,6 +522,16 @@ impl<C: CommandEncoder + DynResource> DynCommandEncoder for C {
         unsafe { C::draw_indexed_indirect(self, buffer, offset, draw_count) };
     }
 
+    unsafe fn draw_mesh_tasks_indirect(
+        &mut self,
+        buffer: &dyn DynBuffer,
+        offset: wgt::BufferAddress,
+        draw_count: u32,
+    ) {
+        let buffer = buffer.expect_downcast_ref();
+        unsafe { C::draw_mesh_tasks_indirect(self, buffer, offset, draw_count) };
+    }
+
     unsafe fn draw_indirect_count(
         &mut self,
         buffer: &dyn DynBuffer,
@@ -530,6 +569,28 @@ impl<C: CommandEncoder + DynResource> DynCommandEncoder for C {
         };
     }
 
+    unsafe fn draw_mesh_tasks_indirect_count(
+        &mut self,
+        buffer: &dyn DynBuffer,
+        offset: wgt::BufferAddress,
+        count_buffer: &dyn DynBuffer,
+        count_offset: wgt::BufferAddress,
+        max_count: u32,
+    ) {
+        let buffer = buffer.expect_downcast_ref();
+        let count_buffer = count_buffer.expect_downcast_ref();
+        unsafe {
+            C::draw_mesh_tasks_indirect_count(
+                self,
+                buffer,
+                offset,
+                count_buffer,
+                count_offset,
+                max_count,
+            )
+        };
+    }
+
     unsafe fn begin_compute_pass(&mut self, desc: &ComputePassDescriptor<dyn DynQuerySet>) {
         let desc = ComputePassDescriptor {
             label: desc.label,

wgpu-hal/src/dynamic/device.rs

@@ -4,10 +4,10 @@ use crate::{
     AccelerationStructureBuildSizes, AccelerationStructureDescriptor, Api, BindGroupDescriptor,
     BindGroupLayoutDescriptor, BufferDescriptor, BufferMapping, CommandEncoderDescriptor,
     ComputePipelineDescriptor, Device, DeviceError, FenceValue,
-    GetAccelerationStructureBuildSizesDescriptor, Label, MemoryRange, PipelineCacheDescriptor,
-    PipelineCacheError, PipelineError, PipelineLayoutDescriptor, RenderPipelineDescriptor,
-    SamplerDescriptor, ShaderError, ShaderInput, ShaderModuleDescriptor, TextureDescriptor,
-    TextureViewDescriptor, TlasInstance,
+    GetAccelerationStructureBuildSizesDescriptor, Label, MemoryRange, MeshPipelineDescriptor,
+    PipelineCacheDescriptor, PipelineCacheError, PipelineError, PipelineLayoutDescriptor,
+    RenderPipelineDescriptor, SamplerDescriptor, ShaderError, ShaderInput, ShaderModuleDescriptor,
+    TextureDescriptor, TextureViewDescriptor, TlasInstance,
 };
 
 use super::{
@@ -100,6 +100,14 @@ pub trait DynDevice: DynResource {
             dyn DynPipelineCache,
         >,
     ) -> Result<Box<dyn DynRenderPipeline>, PipelineError>;
+    unsafe fn create_mesh_pipeline(
+        &self,
+        desc: &MeshPipelineDescriptor<
+            dyn DynPipelineLayout,
+            dyn DynShaderModule,
+            dyn DynPipelineCache,
+        >,
+    ) -> Result<Box<dyn DynRenderPipeline>, PipelineError>;
     unsafe fn destroy_render_pipeline(&self, pipeline: Box<dyn DynRenderPipeline>);
 
     unsafe fn create_compute_pipeline(
@@ -393,6 +401,32 @@ impl<D: Device + DynResource> DynDevice for D {
             .map(|b| -> Box<dyn DynRenderPipeline> { Box::new(b) })
     }
 
+    unsafe fn create_mesh_pipeline(
+        &self,
+        desc: &MeshPipelineDescriptor<
+            dyn DynPipelineLayout,
+            dyn DynShaderModule,
+            dyn DynPipelineCache,
+        >,
+    ) -> Result<Box<dyn DynRenderPipeline>, PipelineError> {
+        let desc = MeshPipelineDescriptor {
+            label: desc.label,
+            layout: desc.layout.expect_downcast_ref(),
+            task_stage: desc.task_stage.clone().map(|f| f.expect_downcast()),
+            mesh_stage: desc.mesh_stage.clone().expect_downcast(),
+            primitive: desc.primitive,
+            depth_stencil: desc.depth_stencil.clone(),
+            multisample: desc.multisample,
+            fragment_stage: desc.fragment_stage.clone().map(|f| f.expect_downcast()),
+            color_targets: desc.color_targets,
+            multiview: desc.multiview,
+            cache: desc.cache.map(|c| c.expect_downcast_ref()),
+        };
+
+        unsafe { D::create_mesh_pipeline(self, &desc) }
+            .map(|b| -> Box<dyn DynRenderPipeline> { Box::new(b) })
+    }
+
     unsafe fn destroy_render_pipeline(&self, pipeline: Box<dyn DynRenderPipeline>) {
         unsafe { D::destroy_render_pipeline(self, pipeline.unbox()) };
     }

wgpu-hal/src/gles/command.rs

@@ -1073,6 +1073,14 @@ impl crate::CommandEncoder for super::CommandEncoder {
             first_instance_location: self.state.first_instance_location.clone(),
         });
     }
+    unsafe fn draw_mesh_tasks(
+        &mut self,
+        _group_count_x: u32,
+        _group_count_y: u32,
+        _group_count_z: u32,
+    ) {
+        unreachable!()
+    }
     unsafe fn draw_indirect(
         &mut self,
         buffer: &super::Buffer,
@@ -1116,6 +1124,14 @@ impl crate::CommandEncoder for super::CommandEncoder {
             });
         }
     }
+    unsafe fn draw_mesh_tasks_indirect(
+        &mut self,
+        _buffer: &<Self::A as crate::Api>::Buffer,
+        _offset: wgt::BufferAddress,
+        _draw_count: u32,
+    ) {
+        unreachable!()
+    }
     unsafe fn draw_indirect_count(
         &mut self,
         _buffer: &super::Buffer,
@@ -1136,6 +1152,16 @@ impl crate::CommandEncoder for super::CommandEncoder {
     ) {
         unreachable!()
     }
+    unsafe fn draw_mesh_tasks_indirect_count(
+        &mut self,
+        _buffer: &<Self::A as crate::Api>::Buffer,
+        _offset: wgt::BufferAddress,
+        _count_buffer: &<Self::A as crate::Api>::Buffer,
+        _count_offset: wgt::BufferAddress,
+        _max_count: u32,
+    ) {
+        unreachable!()
+    }
 
     // compute
 

wgpu-hal/src/gles/device.rs

@@ -1415,6 +1415,16 @@ impl crate::Device for super::Device {
             alpha_to_coverage_enabled: desc.multisample.alpha_to_coverage_enabled,
         })
     }
+    unsafe fn create_mesh_pipeline(
+        &self,
+        _desc: &crate::MeshPipelineDescriptor<
+            <Self::A as crate::Api>::PipelineLayout,
+            <Self::A as crate::Api>::ShaderModule,
+            <Self::A as crate::Api>::PipelineCache,
+        >,
+    ) -> Result<<Self::A as crate::Api>::RenderPipeline, crate::PipelineError> {
+        unreachable!()
+    }
 
     unsafe fn destroy_render_pipeline(&self, pipeline: super::RenderPipeline) {
         // If the pipeline only has 2 strong references remaining, they're `pipeline` and `program_cache`

wgpu-hal/src/lib.rs

@@ -305,7 +305,8 @@ use wgt::WasmNotSendSync;
 
 // - Vertex + Fragment
 // - Compute
-pub const MAX_CONCURRENT_SHADER_STAGES: usize = 2;
+// Task + Mesh + Fragment
+pub const MAX_CONCURRENT_SHADER_STAGES: usize = 3;
 pub const MAX_ANISOTROPY: u8 = 16;
 pub const MAX_BIND_GROUPS: usize = 8;
 pub const MAX_VERTEX_BUFFERS: usize = 16;
@@ -901,6 +902,15 @@ pub trait Device: WasmNotSendSync {
             <Self::A as Api>::PipelineCache,
         >,
     ) -> Result<<Self::A as Api>::RenderPipeline, PipelineError>;
+    #[allow(clippy::type_complexity)]
+    unsafe fn create_mesh_pipeline(
+        &self,
+        desc: &MeshPipelineDescriptor<
+            <Self::A as Api>::PipelineLayout,
+            <Self::A as Api>::ShaderModule,
+            <Self::A as Api>::PipelineCache,
+        >,
+    ) -> Result<<Self::A as Api>::RenderPipeline, PipelineError>;
     unsafe fn destroy_render_pipeline(&self, pipeline: <Self::A as Api>::RenderPipeline);
 
     #[allow(clippy::type_complexity)]
@@ -1458,6 +1468,26 @@ pub trait CommandEncoder: WasmNotSendSync + fmt::Debug {
         count_offset: wgt::BufferAddress,
         max_count: u32,
     );
+    unsafe fn draw_mesh_tasks(
+        &mut self,
+        group_count_x: u32,
+        group_count_y: u32,
+        group_count_z: u32,
+    );
+    unsafe fn draw_mesh_tasks_indirect(
+        &mut self,
+        buffer: &<Self::A as Api>::Buffer,
+        offset: wgt::BufferAddress,
+        draw_count: u32,
+    );
+    unsafe fn draw_mesh_tasks_indirect_count(
+        &mut self,
+        buffer: &<Self::A as Api>::Buffer,
+        offset: wgt::BufferAddress,
+        count_buffer: &<Self::A as Api>::Buffer,
+        count_offset: wgt::BufferAddress,
+        max_count: u32,
+    );
 
     // compute passes
 
@@ -2158,6 +2188,33 @@ pub struct RenderPipelineDescriptor<
     /// The cache which will be used and filled when compiling this pipeline
     pub cache: Option<&'a Pc>,
 }
+pub struct MeshPipelineDescriptor<
+    'a,
+    Pl: DynPipelineLayout + ?Sized,
+    M: DynShaderModule + ?Sized,
+    Pc: DynPipelineCache + ?Sized,
+> {
+    pub label: Label<'a>,
+    /// The layout of bind groups for this pipeline.
+    pub layout: &'a Pl,
+    pub task_stage: Option<ProgrammableStage<'a, M>>,
+    pub mesh_stage: ProgrammableStage<'a, M>,
+    /// The properties of the pipeline at the primitive assembly and rasterization level.
+    pub primitive: wgt::PrimitiveState,
+    /// The effect of draw calls on the depth and stencil aspects of the output target, if any.
+    pub depth_stencil: Option<wgt::DepthStencilState>,
+    /// The multi-sampling properties of the pipeline.
+    pub multisample: wgt::MultisampleState,
+    /// The fragment stage for this pipeline.
+    pub fragment_stage: Option<ProgrammableStage<'a, M>>,
+    /// The effect of draw calls on the color aspect of the output target.
+    pub color_targets: &'a [Option<wgt::ColorTargetState>],
+    /// If the pipeline will be used with a multiview render pass, this indicates how many array
+    /// layers the attachments will have.
+    pub multiview: Option<NonZeroU32>,
+    /// The cache which will be used and filled when compiling this pipeline
+    pub cache: Option<&'a Pc>,
+}
 
 #[derive(Debug, Clone)]
 pub struct SurfaceConfiguration {