Further optimization of cv::merge RVV HAL for 8U and 16S#26923
Merged
asmorkalov merged 3 commits intoopencv:4.xfrom Feb 20, 2025
Merged
Further optimization of cv::merge RVV HAL for 8U and 16S#26923asmorkalov merged 3 commits intoopencv:4.xfrom
asmorkalov merged 3 commits intoopencv:4.xfrom
Conversation
Member
Author
|
@hanliutong @amane-ame , may I ask you review this proposal and verify that MUSE-PI will also have performance improvement? |
Contributor
|
Hi @dkurt , The performance on MUSE-Pi is as follows, it also have improvement: Clang 19.1.7GCC 14.2However, seems Clang 17, which is the minimum compiler version that supports rvv intrinsic v0.12 (currently used for CI), does not support Maybe we should add compiler version check |
This was referenced Feb 16, 2025
Contributor
|
I agree, we need to somehow keep compatibility with clang 17. Perhaps we can raise version i precommit builders, but keep old image in weekly builds. I'll take a look. |
Member
Author
|
@mshabunin, I can modify this PR to be compatible with clang 17, so we can postpone upgrade to newer version until it’s critical. |
asmorkalov
approved these changes
Feb 18, 2025
Member
Author
|
@asmorkalov just a note that GitHub updated UI and CI mark is misleading (2 jobs failed and was restarted, but mark is ✔️) : |
Contributor
|
My performance results for Spacemit Muse Pi v 30 (GCC 14.2.1): |
Contributor
|
@mshabunin Could I merge the PR? |
NanQin555
pushed a commit
to NanQin555/opencv
that referenced
this pull request
Feb 24, 2025
Further optimization of cv::merge RVV HAL for 8U and 16S opencv#26923 ### Pull Request Readiness Checklist * Banana Pi BF3 (SpacemiT K1) RISC-V * Compiler: Syntacore Clang 18.1.4 (build 2024.12) ``` Geometric mean (ms) Name of Test baseline pr pr merge vs baseline merge (x-factor) merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 2) 0.013 0.003 3.76 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 3) 0.020 0.006 3.46 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 4) 0.026 0.010 2.61 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 5) 0.043 0.028 1.56 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 6) 0.054 0.035 1.53 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 7) 0.065 0.050 1.30 merge::Size_SrcDepth_DstChannels::(127x61, 8UC1, 8) 0.070 0.036 1.95 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 2) 0.015 0.008 1.82 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 3) 0.022 0.015 1.48 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 4) 0.029 0.018 1.63 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 5) 0.067 0.044 1.54 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 6) 0.088 0.056 1.58 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 7) 0.104 0.076 1.38 merge::Size_SrcDepth_DstChannels::(127x61, 16SC1, 8) 0.116 0.065 1.79 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 2) 0.421 0.176 2.39 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 3) 0.792 0.284 2.79 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 4) 1.090 0.370 2.95 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 5) 1.835 1.399 1.31 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 6) 2.389 1.776 1.35 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 7) 3.000 2.471 1.21 merge::Size_SrcDepth_DstChannels::(640x480, 8UC1, 8) 3.178 2.104 1.51 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 2) 0.490 0.377 1.30 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 3) 1.348 0.602 2.24 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 4) 1.827 0.813 2.25 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 5) 3.283 2.692 1.22 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 6) 4.922 3.334 1.48 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 7) 5.725 4.399 1.30 merge::Size_SrcDepth_DstChannels::(640x480, 16SC1, 8) 6.278 4.748 1.32 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 2) 1.267 0.603 2.10 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 3) 2.394 0.934 2.56 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 4) 3.236 1.434 2.26 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 5) 5.398 4.345 1.24 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 6) 7.127 5.459 1.31 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 7) 8.590 7.298 1.18 merge::Size_SrcDepth_DstChannels::(1280x720, 8UC1, 8) 9.360 6.152 1.52 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 2) 1.482 1.242 1.19 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 3) 4.008 1.817 2.21 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 4) 6.079 2.468 2.46 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 5) 11.300 8.644 1.31 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 6) 15.125 12.126 1.25 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 7) 17.555 14.804 1.19 merge::Size_SrcDepth_DstChannels::(1280x720, 16SC1, 8) 18.890 14.163 1.33 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 2) 2.910 1.326 2.19 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 3) 5.351 1.997 2.68 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 4) 7.290 2.629 2.77 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 5) 12.426 9.611 1.29 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 6) 16.453 12.162 1.35 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 7) 19.420 16.190 1.20 merge::Size_SrcDepth_DstChannels::(1920x1080, 8UC1, 8) 20.588 13.699 1.50 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 2) 3.400 2.640 1.29 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 3) 8.986 3.952 2.27 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 4) 11.972 5.273 2.27 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 5) 20.544 17.996 1.14 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 6) 28.677 22.086 1.30 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 7) 32.958 27.713 1.19 merge::Size_SrcDepth_DstChannels::(1920x1080, 16SC1, 8) 36.499 27.439 1.33 ``` See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch - [x] There is a reference to the original bug report and related work - [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable Patch to opencv_extra has the same branch name. - [x] The feature is well documented and sample code can be built with the project CMake
6 tasks
Merged
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
5 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.
Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
Patch to opencv_extra has the same branch name.