wg_engine: alpha premultiplied bug fix #4066
Conversation
There was a problem hiding this comment.
Pull request overview
This PR fixes an alpha premultiplication bug in the WebGPU rendering engine by introducing a new sampler specifically configured for image rendering. The fix addresses issue #4041 where images were being rendered with incorrect alpha blending due to using a generic linear clamp sampler instead of a properly configured image sampler.
Key Changes:
- Added a new
samplerImageswith different min/mag filter settings optimized for image rendering - Updated
WgImageDatato use the new image-specific sampler instead of the generic linear clamp sampler - Introduced a new overloaded
createSamplermethod to support different min/mag filter configurations
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| tvgWgRenderData.cpp | Updates bind group creation to use the new samplerImages instead of samplerLinearClamp |
| tvgWgCommon.h | Declares the new samplerImages member and overloaded createSampler method |
| tvgWgCommon.cpp | Implements initialization, cleanup, and creation logic for the new image sampler |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
96529cc to
9777920
Compare
Fix sampler setings for un-premultiplied images #4041
9777920 to
dffcc08
Compare
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
|
Visual testing 👉 Rendering correctly in web. |
@SergeyLebedkin Please test on the thorvg viewer. Thanks. |
|
Hi @SergeyLebedkin, Thank you for the PR. I tested it and noticed a regression when scaling the image to 4×. To address this, I’d like to suggest pre-multiplying the alpha channel before storing the image into the drawing texture. Would you be open to handling this step with GPGPU via WebGPU in the PR? I think doing the premultiply as a lightweight compute pass (or equivalent) could keep the cost low and avoid CPU-side preprocessing. Thanks again for taking a look. |
Fix sampler setings for un-premultiplied images by updating sampler for images
#4041