Emboss Augmentation on HOST and HIP#675
Merged
LakshmiKumar23 merged 65 commits intoROCm:developfrom Mar 17, 2026
Merged
Conversation
Member
r-abishek
commented
Feb 20, 2026
- Adds Emboss Augmentation on HOST and HIP.
- Emboss applies a convolution operator to a 2D image array to generate an array containing difference values that represent edges in the original image.
- Support for kernel sizes 3, 5, 7, and 9 fo bit depths U8/F16/F32/I8
- Adds relevant unit/perf tests and qa test support.
…and resolved all the comment
… into apr/emboss
RPP Emboss on HOST and HIP
Contributor
There was a problem hiding this comment.
Pull request overview
Adds a new Emboss filter augmentation to the RPP tensor API, with HOST/HIP execution support and test-suite / QA wiring so the new op can be exercised via the existing test runners.
Changes:
- Adds
rppt_emboss(...)to the public tensor filter augmentations API and dispatches it inrppt_tensor_filter_augmentations.cppfor HOST and HIP backends. - Implements the HOST emboss kernel (CPU) and declares HOST/HIP executor entrypoints.
- Extends the Python/C++ test suite mappings and runners to include the new emboss case (ID 101) and kernel-size sweep (3/5/7/9), and adds doxygen output images for emboss.
Reviewed changes
Copilot reviewed 11 out of 24 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| utilities/test_suite/rpp_test_suite_image.h | Registers emboss in augmentation maps/enums and treats it as a kernel-size case. |
| utilities/test_suite/common.py | Adds emboss to Python augmentation map and filter augmentation group. |
| utilities/test_suite/HOST/runImageTests.py | Runs emboss across kernel sizes as part of HOST test matrix. |
| utilities/test_suite/HOST/Tensor_image_host.cpp | Adds emboss case execution path in the HOST C++ test binary and updates printed case range. |
| utilities/test_suite/HIP/runImageTests.py | Runs emboss across kernel sizes as part of HIP test/perf matrix. |
| utilities/test_suite/HIP/Tensor_image_hip.cpp | Adds emboss case execution path and alloc/free of strength/bias tensors in HIP test binary. |
| src/modules/tensor/rppt_tensor_filter_augmentations.cpp | Adds rppt_emboss entrypoint and backend dispatch (HOST/HIP). |
| src/modules/tensor/cpu/kernel/emboss.cpp | Adds the HOST emboss implementation (kernel creation + optimized paths). |
| src/include/tensor/host_tensor_executors.hpp | Declares emboss_host_tensor / emboss_generic_host_tensor. |
| src/include/tensor/hip_tensor_executors.hpp | Declares hip_exec_emboss_tensor. |
| docs/data/doxygenOutputs/filter_augmentations_emboss_kSize3_img150x150.png | Adds doxygen sample output image for emboss (3x3). |
| docs/data/doxygenOutputs/filter_augmentations_emboss_kSize5_img150x150.png | Adds doxygen sample output image for emboss (5x5). |
| docs/data/doxygenOutputs/filter_augmentations_emboss_kSize7_img150x150.png | Adds doxygen sample output image for emboss (7x7). |
| docs/data/doxygenOutputs/filter_augmentations_emboss_kSize9_img150x150.png | Adds doxygen sample output image for emboss (9x9). |
| api/rppt_tensor_filter_augmentations.h | Documents and exposes the new rppt_emboss public API. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+1218
to
+1220
| __m256 pRow[45], pDst[2]; | ||
| rpp_load_gaussian_filter_9x9_pkd_pln_host(pRow, srcPtrTemp, rowKernelLoopLimit, padIndex); | ||
| pDst[0] = avx_p0; |
There was a problem hiding this comment.
In the 9x9 NHWC->NCHW AVX2 path, this calls rpp_load_gaussian_filter_9x9_pkd_pln_host(...), which looks like a copy/paste from gaussian_filter and is inconsistent with emboss. This likely loads the wrong neighborhood/packing assumptions for emboss and can corrupt output. Please switch to the appropriate generic NxN loader used elsewhere in emboss (or add an emboss-specific 9x9 pkd->pln loader).
Contributor
There was a problem hiding this comment.
Both emboss and gaussian filter are algorithmically similar expect the filter values. So same loads and store funcs of gaussian filter are used here
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## develop #675 +/- ##
===========================================
- Coverage 92.46% 92.43% -0.02%
===========================================
Files 198 200 +2
Lines 89139 90759 +1620
===========================================
+ Hits 82415 83891 +1476
- Misses 6724 6868 +144
🚀 New features to boost your workflow:
|
Emboss - resolved review comments
Merged develop - Emboss
rrawther
approved these changes
Mar 16, 2026
AryanSalmanpour
approved these changes
Mar 17, 2026
LakshmiKumar23
approved these changes
Mar 17, 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.