Merge #2260

2260: Remote command sink in Metal r=grovesNL a=kvark

Fixes #2259
The results so far are not super promising - highly unstable (presumably, because of the dispatch), with performance around `Immediate` mark. We are still missing the most important follow-up here - to avoid any heap allocations when recording commands. Currently, it just goes with `Vec::new()` and grows it for each pass, which shows up in the profile quite a bit.

The PR also has a bunch of stuff in general optimizations:
  - HAL change in the descriptor allocation API to avoid the heap
  - lighten up Metal descriptor binding path (a bit) by making sure there is enough state slots in advance
  - simplification and refactoring of `CommandSink` implementations

PR checklist:
- [ ] `make` succeeds (on *nix)
- [ ] `make reftests` succeeds
- [x] tested examples with the following backends: Metal
- [ ] `rustfmt` run on changed code


Co-authored-by: Dzmitry Malyshau <[email protected]>

Author: bors[bot]

src/backend/gl/src/native.rs

@@ -174,15 +174,11 @@ pub struct DescriptorSet {
 pub struct DescriptorPool {}
 
 impl pso::DescriptorPool<Backend> for DescriptorPool {
-    fn allocate_sets<I>(&mut self, layouts: I) -> Vec<Result<DescriptorSet, pso::AllocationError>>
-    where
-        I: IntoIterator,
-        I::Item: Borrow<DescriptorSetLayout>,
-    {
-        layouts.into_iter().map(|layout| Ok(DescriptorSet {
-            layout: layout.borrow().clone(),
+    fn allocate_set(&mut self, layout: &DescriptorSetLayout) -> Result<DescriptorSet, pso::AllocationError> {
+        Ok(DescriptorSet {
+            layout: layout.clone(),
             bindings: Arc::new(Mutex::new(Vec::new())),
-        })).collect()
+        })
     }
 
     fn free_sets<I>(&mut self, _descriptor_sets: I)

src/backend/metal/Cargo.toml

@@ -30,6 +30,7 @@ block = "0.1"
 cocoa = "0.15"
 core-foundation = "0.6"
 core-graphics = "0.14"
+dispatch = "0.1"
 smallvec = "0.6"
 spirv_cross = "0.9"
 parking_lot = "0.6.3"