native: bind groups for both compute and graphics

wgpu-native/src/command/bind.rs

@@ -1,12 +1,10 @@
-use crate::registry::{HUB, Items, ConcreteItems};
+use crate::registry::{HUB, Items};
 use crate::{
     B, Stored, WeaklyStored,
-    BindGroup, PipelineLayout,
     BindGroupId, BindGroupLayoutId, PipelineLayoutId,
 };
 
 use hal;
-use parking_lot::RwLockReadGuard;
 
 
 #[derive(Clone, Default)]
@@ -21,35 +19,20 @@ pub struct Binder {
     entries: Vec<BindGroupEntry>,
 }
 
-pub struct NewBind<'a, B: hal::Backend> {
-    pipeline_layout_guard: RwLockReadGuard<'a, ConcreteItems<PipelineLayout<B>>>,
-    pipeline_layout_id: PipelineLayoutId,
-    bind_group_guard: RwLockReadGuard<'a, ConcreteItems<BindGroup<B>>>,
-    bind_group_id: BindGroupId,
-}
-
-impl<'a, B: hal::Backend> NewBind<'a, B> {
-    pub fn pipeline_layout(&self) -> &B::PipelineLayout {
-        &self.pipeline_layout_guard.get(self.pipeline_layout_id).raw
-    }
-
-    pub fn descriptor_set(&self) -> &B::DescriptorSet {
-        &self.bind_group_guard.get(self.bind_group_id).raw
-    }
-}
+//Note: we can probably make this much better than passing an `FnMut`
 
 impl Binder {
-    //Note: `'a` is need to avoid inheriting the lifetime from `self`
-    pub fn bind_group<'a>(
-        &mut self, index: u32, bind_group_id: BindGroupId
-    ) -> Option<NewBind<'a, B>> {
+    pub fn bind_group<F>(&mut self, index: usize, bind_group_id: BindGroupId, mut fun: F)
+    where
+        F: FnMut(&<B as hal::Backend>::PipelineLayout, &<B as hal::Backend>::DescriptorSet),
+    {
         let bind_group_guard = HUB.bind_groups.read();
         let bind_group = bind_group_guard.get(bind_group_id);
 
-        while self.entries.len() <= index as usize {
+        while self.entries.len() <= index {
             self.entries.push(BindGroupEntry::default());
         }
-        *self.entries.get_mut(index as usize).unwrap() = BindGroupEntry {
+        *self.entries.get_mut(index).unwrap() = BindGroupEntry {
             layout: Some(bind_group.layout_id.clone()),
             data: Some(Stored {
                 value: bind_group_id,
@@ -61,16 +44,41 @@ impl Binder {
             //TODO: we can cache the group layout ids of the current pipeline in `Binder` itself
             let pipeline_layout_guard = HUB.pipeline_layouts.read();
             let pipeline_layout = pipeline_layout_guard.get(pipeline_layout_id);
-            if pipeline_layout.bind_group_layout_ids[index as usize] == bind_group.layout_id {
-                return Some(NewBind {
-                    pipeline_layout_guard,
-                    pipeline_layout_id,
-                    bind_group_guard,
-                    bind_group_id,
-                })
+            if pipeline_layout.bind_group_layout_ids[index] == bind_group.layout_id {
+                fun(&pipeline_layout.raw, &bind_group.raw);
             }
         }
+    }
 
-        None
+    pub fn change_layout<F>(&mut self, pipeline_layout_id: PipelineLayoutId, mut fun: F)
+    where
+        F: FnMut(&<B as hal::Backend>::PipelineLayout, usize, &<B as hal::Backend>::DescriptorSet),
+    {
+        if self.pipeline_layout_id == Some(WeaklyStored(pipeline_layout_id)) {
+            return
+        }
+
+        self.pipeline_layout_id = Some(WeaklyStored(pipeline_layout_id));
+        let pipeline_layout_guard = HUB.pipeline_layouts.read();
+        let pipeline_layout = pipeline_layout_guard.get(pipeline_layout_id);
+        let bing_group_guard = HUB.bind_groups.read();
+
+        while self.entries.len() < pipeline_layout.bind_group_layout_ids.len() {
+            self.entries.push(BindGroupEntry::default());
+        }
+        for (index, (entry, bgl_id)) in self.entries
+            .iter_mut()
+            .zip(&pipeline_layout.bind_group_layout_ids)
+            .enumerate()
+        {
+            if entry.layout == Some(bgl_id.clone()) {
+                continue
+            }
+            entry.layout = Some(bgl_id.clone());
+            if let Some(ref bg_id) = entry.data {
+                let bind_group = bing_group_guard.get(bg_id.value);
+                fun(&pipeline_layout.raw, index, &bind_group.raw);
+            }
+        }
     }
 }

wgpu-native/src/command/compute.rs

@@ -38,43 +38,6 @@ pub extern "C" fn wgpu_compute_pass_end_pass(pass_id: ComputePassId) -> CommandB
     pass.cmb_id.value
 }
 
-#[no_mangle]
-pub extern "C" fn wgpu_compute_pass_set_bind_group(
-    pass_id: ComputePassId,
-    index: u32,
-    bind_group_id: BindGroupId,
-) {
-    let mut pass_guard = HUB.compute_passes.write();
-    let pass = pass_guard.get_mut(pass_id);
-    if let Some(bind) = pass.binder.bind_group(index, bind_group_id) {
-        unsafe {
-            pass.raw.bind_compute_descriptor_sets(
-                bind.pipeline_layout(),
-                index as usize,
-                iter::once(bind.descriptor_set()),
-                &[],
-            );
-        }
-    }
-}
-
-#[no_mangle]
-pub extern "C" fn wgpu_compute_pass_set_pipeline(
-    pass_id: ComputePassId,
-    pipeline_id: ComputePipelineId,
-) {
-    let pipeline_guard = HUB.compute_pipelines.read();
-    let pipeline = &pipeline_guard.get(pipeline_id).raw;
-
-    unsafe {
-        HUB.compute_passes
-            .write()
-            .get_mut(pass_id)
-            .raw
-            .bind_compute_pipeline(pipeline);
-    }
-}
-
 #[no_mangle]
 pub extern "C" fn wgpu_compute_pass_dispatch(pass_id: ComputePassId, x: u32, y: u32, z: u32) {
     unsafe {
@@ -85,3 +48,46 @@ pub extern "C" fn wgpu_compute_pass_dispatch(pass_id: ComputePassId, x: u32, y:
             .dispatch([x, y, z]);
     }
 }
+
+#[no_mangle]
+pub extern "C" fn wgpu_compute_pass_set_bind_group(
+    pass_id: ComputePassId,
+    index: u32,
+    bind_group_id: BindGroupId,
+) {
+    let mut pass_guard = HUB.compute_passes.write();
+    let ComputePass { ref mut raw, ref mut binder, .. } = *pass_guard.get_mut(pass_id);
+
+    binder.bind_group(index as usize, bind_group_id, |pipeline_layout, desc_set| unsafe {
+        raw.bind_compute_descriptor_sets(
+            pipeline_layout,
+            index as usize,
+            iter::once(desc_set),
+            &[],
+        );
+    });
+}
+
+#[no_mangle]
+pub extern "C" fn wgpu_compute_pass_set_pipeline(
+    pass_id: ComputePassId,
+    pipeline_id: ComputePipelineId,
+) {
+    let mut pass_guard = HUB.compute_passes.write();
+    let ComputePass { ref mut raw, ref mut binder, .. } = *pass_guard.get_mut(pass_id);
+
+    let pipeline_guard = HUB.compute_pipelines.read();
+    let pipeline = pipeline_guard.get(pipeline_id);
+
+    unsafe {
+        raw.bind_compute_pipeline(&pipeline.raw);
+    }
+    binder.change_layout(pipeline.layout_id.0, |pipeline_layout, index, desc_set| unsafe {
+        raw.bind_compute_descriptor_sets(
+            pipeline_layout,
+            index,
+            iter::once(desc_set),
+            &[],
+        );
+    });
+}

wgpu-native/src/command/render.rs

@@ -1,17 +1,21 @@
+use crate::command::bind::Binder;
 use crate::resource::BufferUsageFlags;
 use crate::registry::{Items, HUB};
 use crate::track::{BufferTracker, TextureTracker, TrackPermit};
 use crate::{
     CommandBuffer, Stored,
-    BufferId, CommandBufferId, RenderPassId,
+    BindGroupId, BufferId, CommandBufferId, RenderPassId, RenderPipelineId,
 };
 
 use hal::command::RawCommandBuffer;
 
+use std::iter;
+
 
 pub struct RenderPass<B: hal::Backend> {
     raw: B::CommandBuffer,
     cmb_id: Stored<CommandBufferId>,
+    binder: Binder,
     buffer_tracker: BufferTracker,
     texture_tracker: TextureTracker,
 }
@@ -21,6 +25,7 @@ impl<B: hal::Backend> RenderPass<B> {
         RenderPass {
             raw,
             cmb_id,
+            binder: Binder::default(),
             buffer_tracker: BufferTracker::new(),
             texture_tracker: TextureTracker::new(),
         }
@@ -152,3 +157,46 @@ pub extern "C" fn wgpu_render_pass_draw_indexed(
             );
     }
 }
+
+#[no_mangle]
+pub extern "C" fn wgpu_render_pass_set_bind_group(
+    pass_id: RenderPassId,
+    index: u32,
+    bind_group_id: BindGroupId,
+) {
+    let mut pass_guard = HUB.render_passes.write();
+    let RenderPass { ref mut raw, ref mut binder, .. } = *pass_guard.get_mut(pass_id);
+
+    binder.bind_group(index as usize, bind_group_id, |pipeline_layout, desc_set| unsafe {
+        raw.bind_compute_descriptor_sets(
+            pipeline_layout,
+            index as usize,
+            iter::once(desc_set),
+            &[],
+        );
+    });
+}
+
+#[no_mangle]
+pub extern "C" fn wgpu_render_pass_set_pipeline(
+    pass_id: RenderPassId,
+    pipeline_id: RenderPipelineId,
+) {
+    let mut pass_guard = HUB.render_passes.write();
+    let RenderPass { ref mut raw, ref mut binder, .. } = *pass_guard.get_mut(pass_id);
+
+    let pipeline_guard = HUB.render_pipelines.read();
+    let pipeline = pipeline_guard.get(pipeline_id);
+
+    unsafe {
+        raw.bind_graphics_pipeline(&pipeline.raw);
+    }
+    binder.change_layout(pipeline.layout_id.0, |pipeline_layout, index, desc_set| unsafe {
+        raw.bind_graphics_descriptor_sets(
+            pipeline_layout,
+            index,
+            iter::once(desc_set),
+            &[],
+        );
+    });
+}

wgpu-native/src/device.rs

@@ -986,7 +986,10 @@ pub extern "C" fn wgpu_device_create_render_pipeline(
 
     HUB.render_pipelines
         .write()
-        .register(pipeline::RenderPipeline { raw: pipeline })
+        .register(pipeline::RenderPipeline {
+            raw: pipeline,
+            layout_id: WeaklyStored(desc.layout),
+        })
 }
 
 #[no_mangle]

wgpu-native/src/pipeline.rs

@@ -1,5 +1,8 @@
 use crate::resource;
-use crate::{BlendStateId, ByteArray, DepthStencilStateId, PipelineLayoutId, ShaderModuleId};
+use crate::{
+    ByteArray, WeaklyStored,
+    BlendStateId, DepthStencilStateId, PipelineLayoutId, ShaderModuleId,
+};
 
 use bitflags::bitflags;
 
@@ -212,6 +215,7 @@ pub struct ComputePipelineDescriptor {
 
 pub(crate) struct ComputePipeline<B: hal::Backend> {
     pub raw: B::ComputePipeline,
+    pub layout_id: WeaklyStored<PipelineLayoutId>,
 }
 
 #[repr(C)]
@@ -251,4 +255,5 @@ pub struct RenderPipelineDescriptor {
 
 pub(crate) struct RenderPipeline<B: hal::Backend> {
     pub raw: B::GraphicsPipeline,
+    pub layout_id: WeaklyStored<PipelineLayoutId>,
 }