Internal tiling for OpenCL guided filter #20245
Open
+255
−169
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jenshannoschwalm
added
scope: performance
ralfbrown
reviewed
Jan 29, 2026
jenshannoschwalm
force-pushed
the
guided_cl_filter
branch
from
357b23d to
033c1fb
Compare
Collaborator
Author
|
Force-pushed updated version (@ralfbrown spotted) and updated main text |
Collaborator
Author
|
Release note: Increased performance for OpenCL guided filter by internal tiling. |
The OpenCL guided filter (used in blending and hazeremoval) requires 18 floats/pixel internally thus often leads to CPU fallbacks (overall requirements are ~30 floats/pixel) while processing large images in HQ darkroom or while exporting. This commit introduces internal tiling for improved performance. The overlap - thus the efficiacy of internal tiling - depends on the guided filter radius, still performance is much better. We use the device specific advantage hint if provided. With tiled OpenCL code we relax calculation of blending tiling requirements. - some formatting - updated kernels to access a specific part of the guide
jenshannoschwalm
force-pushed
the
guided_cl_filter
branch
from
033c1fb to
9b632f3
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The OpenCL guided filter (used in blending and hazeremoval) requires 21 floats/pixel internally thus often leads to CPU fallbacks (overall requirements are >30 floats/pixel) while processing large images in HQ darkroom or while exporting.
This commit introduces internal tiling for improved performance. The overlap - thus the efficiacy of internal tiling - depends on the guided filter radius, still performance is much better.
We use the device specific advantage hint if provided.
With tiled OpenCL code we relax calculation of blending tiling requirements.
BTW this work was inspired by the first proposed "benchmark" in https://discuss.pixls.us/t/dt-performance-analyzer-v0-6/55563/54