CMake-only fork of https://dawn.googlesource.com/dawn with cleaned, vendored dependencies
df4c60546e
It may not be the root cause of CTS memory leak, but we found some issues in D3D12 memory allocation for large buffers (> 4M). The memory allocation is not freed correctly when we call buffer destroy. When we create a large buffer, it always allocate committed resources from the system memory and append it to LRU cache, and will try to free the expired memory in LRU cache when the left memory size is not enough for the current allocation. When we call buffer destroy, we remove its heap from LRU cache, release its allocation, but will add the allocation reference to a delete queue and wait to be freed together in the next device.tick. At this time, the destroyed memory is not really released to the system memory. If no tick is triggered during the buffer allocation and deallocation, these memories will never be reused again and there will be a memory leak. We need to free these memories when we detect there is not enough memory, do not wait for device.tick to do that. Add a stress case to track the issue, and need to discuss how to fix it. Bug: dawn:957 Change-Id: Id4bcc97496d372ad7c27c31675327ca12e133217 Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/55740 Commit-Queue: Hao Li <hao.x.li@intel.com> Reviewed-by: Corentin Wallez <cwallez@chromium.org> |
||
|---|---|---|
| .vscode | ||
| build_overrides | ||
| docs | ||
| examples | ||
| generator | ||
| infra/config | ||
| scripts | ||
| src | ||
| third_party | ||
| tools/memory/asan | ||
| .clang-format | ||
| .gitattributes | ||
| .gitignore | ||
| .gn | ||
| AUTHORS | ||
| BUILD.gn | ||
| CMakeLists.txt | ||
| DEPS | ||
| DIR_METADATA | ||
| LICENSE | ||
| OWNERS | ||
| PRESUBMIT.py | ||
| README.chromium | ||
| README.md | ||
| codereview.settings | ||
| dawn.json | ||
| dawn_wire.json | ||
README.md
Dawn, a WebGPU implementation
Dawn is an open-source and cross-platform implementation of the work-in-progress WebGPU standard.
More precisely it implements webgpu.h that is a one-to-one mapping with the WebGPU IDL.
Dawn is meant to be integrated as part of a larger system and is the underlying implementation of WebGPU in Chromium.
Dawn provides several WebGPU building blocks:
- WebGPU C/C++ headers that applications and other building blocks use.
- The
webgpu.hversion that Dawn implements. - A C++ wrapper for the
webgpu.h.
- The
- A "native" implementation of WebGPU using platforms' GPU APIs:
- D3D12 on Windows 10
- Metal on macOS and iOS
- Vulkan on Windows, Linux, ChromeOS, Android and Fuchsia
- OpenGL as best effort where available
- A client-server implementation of WebGPU for applications that are in a sandbox without access to native drivers
Helpful links:
- Dawn's bug tracker if you find issues with Dawn.
- Dawn's mailing list for other discussions related to Dawn.
- Dawn's source code
- Dawn's Matrix chatroom for live discussion around contributing or using Dawn.
- WebGPU's Matrix chatroom
Documentation table of content
Developer documentation:
User documentation: (TODO, figure out what overlaps with the webgpu.h docs)
Status
(TODO)
License
Apache 2.0 Public License, please see LICENSE.
Disclaimer
This is not an officially supported Google product.