1417: Split the tracker into stateful/stateless to reduce the overhead r=cwfitzgerald a=kvark
**Connections**
Implements https://github.com/gfx-rs/wgpu/issues/1413#issuecomment-851604344
Reduces the overhead for resource tracking in the Animometer benchmark by up to 50%.
**Description**
We used to use the full tracker set on the usage scopes associated with compute/render passes. A resource tracker has 2 responsibilities: ensuring the resource is held alive, and validating and recording the state transitions. This PR exploits the fact that the latter responsibility is only applicable for buffers and textures. So doing all the lifetime tracking for a pass is a waste: we can instead just attach the lifetimes to the parent command buffer, straight.
In the Animometer benchmark, there is one large buffer, and thousands of bind groups pointing to different offsets into it. The old code would fill up the pass tracker with those bind groups, and then merge it into the command buffer tracker. The new code would just fill up the command buffer tracker instead. Since there is only one buffer, the pass tracking becomes much lighter.
**Testing**
Untested. It would be nice to have some benchmarks here, possibly after #1397 ?
Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
1403: Trace render passes before they are ran r=cwfitzgerald a=kvark
**Connections**
Follow-up to #1245
**Description**
We want the API traces to reproduce issues. Previously, we moved all the resource creation to happen after the trace is written.
This PR follow-up by also recording the render pass commands even in the case where one of them triggers a validation error.
Another big part of the PR now is handling the command buffers that failed for some reason. The submission is skipping them now instead of erroring the whole thing.
**Testing**
Tested manually on wgpu-rs examples
Co-authored-by: Dzmitry Malyshau <dmalyshau@mozilla.com>
* Add command buffer fill buffer method
* Added test for buffer fill
* Added image clear + test
* image/buffer clear methods are now pure clear to zero
* fix cargo clippy & int type issues
* clear buffer alignment is now given by COPY_BUFFER_ALIGNMENT
* joined buffer/image clear player tests
* ImageSubresourceRange is now used in TextureViewDescriptor
* Pass [u8] instead of [u32] for push constants
Use 2-instruction aligner
Fixed alignment
Fixed offsets
Added alignment checks and fixed size
Use semantic constants
* Replaced all magic number with constant
* Fixed type cast
There was a lot of highly unsafe use of serialization based on peek-poke that we
weren't entirely happy with. It's replaced by just serializing the passes now.
Also, switch BufferSize to Option<NonZero>.
This is a major change in how the bundles are implemented. Instead of
transparently injecting them into the pass command stream, we are now
treating bundles as first-class API objects and API tracing them
accordingly. The bundle contains a normalized command stream that is
very easy to inject into a native command buffer multiple times.