Adds full frame implementation to this node operation.
No functional changes.
1.05x faster than tiled fallback.
Details
Details
Diff Detail
Diff Detail
- Repository
- rB Blender
- Branch
- cmp-node-bilateral-blur (branched from master)
- Build Status
Buildable 15360 Build 15360: arc lint + arc unit
Event Timeline
Comment Actions
- Refactor method to reduce complexity
- Merge branch 'master' into cmp-node-bilateral-blur
- After merge fixes
| source/blender/compositor/operations/COM_BilateralBlurOperation.cc | ||
|---|---|---|
| 158 | I think we need feedback from artists. Don't know if they find useful the edit and render quality settings and whether it's confusing that it only has effect on some nodes. Only blurs. We can keep it for the time being and once we see how much performance improves, we can decide better. | |
Comment Actions
Add performance test to main post.
I only do performance tests to ensure no regressions.
Comment Actions
It is unclear to me what the performance measurements are about?
| source/blender/compositor/operations/COM_BilateralBlurOperation.cc | ||
|---|---|---|
| 142 | Add whiteline. | |
Comment Actions
To compare with tiled implementation. Executes tiled implementation fallback and fullframe comparing both. I use this patch:
Comment Actions
- Merge branch 'master' into cmp-node-bilateral-blur
- After merge fixes
- Move output writing to blur_pixel. Makes more sense.