Median and Gaussian Filter optimizations#681
Merged
rrawther merged 20 commits intoROCm:developfrom Mar 30, 2026
Merged
Conversation
RPP Median Filter and Gaussian Optimizations
Contributor
There was a problem hiding this comment.
Pull request overview
This PR implements significant performance optimizations for median and gaussian filters targeting AVX2-level performance comparable to OpenCV. The changes introduce sorting networks for small kernels (3×3, 5×5), histogram-based methods for large kernels with U8 types, and various SIMD optimizations.
Changes:
- Median filter: Added specialized 5×5 sorting network implementations for HIP and CPU with AVX2 vectorization
- Median filter: Introduced histogram-based median calculation for U8 types with large kernels (7×7, 9×9+)
- Gaussian filter: Pre-computed filter coefficients, optimized kernel generation, added prefetching and loop unrolling
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 9 comments.
| File | Description |
|---|---|
| src/modules/tensor/hip/kernel/median_filter.cpp | Fixed double semicolon, added optimized 5×5 median implementation with sorting network, refactored compute_median with histogram method for large U8 kernels |
| src/modules/tensor/cpu/kernel/median_filter.cpp | Added AVX2-vectorized sorting networks for 3×3 and 5×5 median filters across multiple data types (U8, I8, F32, F16), implemented histogram-based median for large U8 kernels |
| src/modules/tensor/hip/kernel/gaussian_filter.cpp | Changed expf to __expf intrinsic, optimized kernel generation with single-pass normalization |
| src/modules/tensor/cpu/kernel/gaussian_filter.cpp | Pre-allocated and broadcast filter coefficients, optimized kernel generation, added prefetching, unrolled convolution loops for 3×3 kernel |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## develop #681 +/- ##
===========================================
+ Coverage 92.45% 92.48% +0.03%
===========================================
Files 215 215
Lines 94897 95987 +1090
===========================================
+ Hits 87729 88767 +1038
- Misses 7168 7220 +52
🚀 New features to boost your workflow:
|
rrawther
requested changes
Mar 19, 2026
rrawther
reviewed
Mar 20, 2026
rrawther
reviewed
Mar 20, 2026
rrawther
reviewed
Mar 20, 2026
rrawther
reviewed
Mar 20, 2026
rrawther
reviewed
Mar 20, 2026
Contributor
rrawther
left a comment
There was a problem hiding this comment.
please address the review comments
LakshmiKumar23
approved these changes
Mar 27, 2026
Contributor
|
@r-abishek @HazarathKumarM code coverage drop 7% . Please take a look |
Contributor
|
@LakshmiKumar23 With the recent CI runs, we are observing the code coverage of this PR as 92.48%. we are not observing any drop |
LakshmiKumar23
approved these changes
Mar 30, 2026
rrawther
approved these changes
Mar 30, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This PR contains specific powerful optimizations in Median and Gaussian filter to match and marginally beat AVX2 level performance using OpenCV.
PERFORMANCE COMPARISONS FOR MEDIAN FILTER AND GAUSSIAN FILTER