Skip to content

Add RISC-V HAL implementation for cv::pyrDown and cv::pyrUp#26958

Merged
asmorkalov merged 7 commits intoopencv:4.xfrom
amane-ame:pyramids_hal_rvv
Mar 4, 2025
Merged

Add RISC-V HAL implementation for cv::pyrDown and cv::pyrUp#26958
asmorkalov merged 7 commits intoopencv:4.xfrom
amane-ame:pyramids_hal_rvv

Conversation

@amane-ame
Copy link
Copy Markdown
Contributor

@amane-ame amane-ame commented Feb 22, 2025

This patch implements cv_hal_pyrdown/cv_hal_pyrup function in RVV_HAL using native intrinsics, optimizing the performance for cv::pyrDown, cv::pyrUp and cv::buildPyramids with data types {8U,16S,32F} x {C1,C2,C3,C4,Cn}.

Tested on MUSE-PI (Spacemit X60) for both gcc 14.2 and clang 20.0.

$ ./opencv_test_imgproc --gtest_filter="*pyr*:*Pyr*"
$ ./opencv_perf_imgproc --gtest_filter="*pyr*:*Pyr*" --perf_min_samples=300 --perf_force_samples=300
Untitled

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

  • I agree to contribute to the project under Apache 2 License.
  • 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
  • The PR is proposed to the proper branch
  • There is a reference to the original bug report and related work
  • There is accuracy test, performance test and test data in opencv_extra repository, if applicable
    Patch to opencv_extra has the same branch name.
  • The feature is well documented and sample code can be built with the project CMake

amane-ame and others added 5 commits February 16, 2025 16:11
@amane-ame @hanliutong
Add HAL RVV pyrDown and pyrUp.
8b0c870 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@amane-ame @hanliutong
Remove trailing space and newline.
1be9af6 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@amane-ame @hanliutong
Remove indent in namespaces.
c697499 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@amane-ame @hanliutong
Move functions into templates.
c5c8d7e 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@amane-ame @hanliutong
Remove AutoBuffer.
47f67bb 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@asmorkalov asmorkalov self-requested a review February 25, 2025 11:48
@asmorkalov asmorkalov self-assigned this Feb 25, 2025
@asmorkalov asmorkalov added this to the 4.12.0 milestone Feb 25, 2025
asmorkalov
asmorkalov requested changes Feb 25, 2025
View reviewed changes
@asmorkalov
Copy link
Copy Markdown
Contributor

asmorkalov commented Feb 25, 2025

Please rebase and fix conflicts.

@amane-ame amane-ame requested a review from asmorkalov February 25, 2025 13:11
@amane-ame amane-ame requested a review from mshabunin February 28, 2025 12:22
mshabunin
mshabunin reviewed Feb 28, 2025
View reviewed changes
mshabunin
mshabunin reviewed Feb 28, 2025
View reviewed changes
@amane-ame @hanliutong
Remove CV_ASSERT and copy borderInterpolate.
5ab47c2 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@amane-ame amane-ame requested a review from mshabunin March 1, 2025 03:18
@asmorkalov
Copy link
Copy Markdown
Contributor

asmorkalov commented Mar 1, 2025

Perf numbers for Spacemit Muse Pi v 30 (GCC 14.2):

pyrDown::Size_MatType::(127x61, 8UC1) 	0.018 	0.018 	1.01
pyrDown::Size_MatType::(127x61, 16SC1) 	0.022 	0.018 	1.17
pyrDown::Size_MatType::(127x61, 32FC1) 	0.019 	0.019 	1.00
pyrDown::Size_MatType::(127x61, 8UC3) 	0.065 	0.031 	2.14
pyrDown::Size_MatType::(127x61, 16SC3) 	0.070 	0.032 	2.16
pyrDown::Size_MatType::(127x61, 32FC3) 	0.065 	0.035 	1.83
pyrDown::Size_MatType::(127x61, 8UC4) 	0.042 	0.036 	1.17
pyrDown::Size_MatType::(127x61, 16SC4) 	0.069 	0.040 	1.73
pyrDown::Size_MatType::(127x61, 32FC4) 	0.065 	0.042 	1.56
pyrDown::Size_MatType::(320x240, 8UC1) 	0.058 	0.051 	1.14
pyrDown::Size_MatType::(320x240, 16SC1) 	0.063 	0.055 	1.14
pyrDown::Size_MatType::(320x240, 32FC1) 	0.064 	0.064 	0.99
pyrDown::Size_MatType::(320x240, 8UC3) 	0.456 	0.138 	3.31
pyrDown::Size_MatType::(320x240, 16SC3) 	0.440 	0.156 	2.82
pyrDown::Size_MatType::(320x240, 32FC3) 	0.382 	0.166 	2.29
pyrDown::Size_MatType::(320x240, 8UC4) 	0.237 	0.185 	1.28
pyrDown::Size_MatType::(320x240, 16SC4) 	0.438 	0.202 	2.16
pyrDown::Size_MatType::(320x240, 32FC4) 	0.400 	0.265 	1.51
pyrDown::Size_MatType::(640x480, 8UC1) 	0.180 	0.156 	1.16
pyrDown::Size_MatType::(640x480, 16SC1) 	0.181 	0.162 	1.12
pyrDown::Size_MatType::(640x480, 32FC1) 	0.222 	0.225 	0.98
pyrDown::Size_MatType::(640x480, 8UC3) 	1.770 	0.558 	3.17
pyrDown::Size_MatType::(640x480, 16SC3) 	1.762 	0.721 	2.44
pyrDown::Size_MatType::(640x480, 32FC3) 	1.631 	1.021 	1.60
pyrDown::Size_MatType::(640x480, 8UC4) 	0.948 	0.824 	1.15
pyrDown::Size_MatType::(640x480, 16SC4) 	1.830 	0.996 	1.84
pyrDown::Size_MatType::(640x480, 32FC4) 	1.833 	1.413 	1.30
pyrDown::Size_MatType::(1280x720, 8UC1) 	0.495 	0.437 	1.13
pyrDown::Size_MatType::(1280x720, 16SC1) 	0.563 	0.497 	1.13
pyrDown::Size_MatType::(1280x720, 32FC1) 	0.829 	0.819 	1.01
pyrDown::Size_MatType::(1280x720, 8UC3) 	5.462 	1.609 	3.40
pyrDown::Size_MatType::(1280x720, 16SC3) 	5.252 	2.294 	2.29
pyrDown::Size_MatType::(1280x720, 32FC3) 	4.644 	3.274 	1.42
pyrDown::Size_MatType::(1280x720, 8UC4) 	2.970 	2.415 	1.23
pyrDown::Size_MatType::(1280x720, 16SC4) 	5.621 	3.238 	1.74
pyrDown::Size_MatType::(1280x720, 32FC4) 	6.236 	5.073 	1.23
pyrDown::Size_MatType::(1920x1080, 8UC1) 	2.271 	1.961 	1.16
pyrDown::Size_MatType::(1920x1080, 16SC1) 	1.466 	1.352 	1.08
pyrDown::Size_MatType::(1920x1080, 32FC1) 	2.026 	2.068 	0.98
pyrDown::Size_MatType::(1920x1080, 8UC3) 	13.350 	3.836 	3.48
pyrDown::Size_MatType::(1920x1080, 16SC3) 	12.742 	5.957 	2.14
pyrDown::Size_MatType::(1920x1080, 32FC3) 	10.460 	8.651 	1.21
pyrDown::Size_MatType::(1920x1080, 8UC4) 	8.920 	6.165 	1.45
pyrDown::Size_MatType::(1920x1080, 16SC4) 	14.579 	9.148 	1.59
pyrDown::Size_MatType::(1920x1080, 32FC4) 	16.442 	13.892 	1.18
pyrUp::Size_MatType::(127x61, 8UC1) 	0.124 	0.074 	1.68
pyrUp::Size_MatType::(127x61, 16SC1) 	0.119 	0.077 	1.55
pyrUp::Size_MatType::(127x61, 32FC1) 	0.134 	0.141 	0.95
pyrUp::Size_MatType::(127x61, 8UC3) 	0.363 	0.221 	1.64
pyrUp::Size_MatType::(127x61, 16SC3) 	0.362 	0.243 	1.49
pyrUp::Size_MatType::(127x61, 32FC3) 	0.519 	0.473 	1.10
pyrUp::Size_MatType::(127x61, 8UC4) 	0.493 	0.288 	1.71
pyrUp::Size_MatType::(127x61, 16SC4) 	0.506 	0.346 	1.46
pyrUp::Size_MatType::(127x61, 32FC4) 	0.934 	0.511 	1.83
pyrUp::Size_MatType::(320x240, 8UC1) 	1.109 	0.807 	1.37
pyrUp::Size_MatType::(320x240, 16SC1) 	1.013 	0.648 	1.56
pyrUp::Size_MatType::(320x240, 32FC1) 	1.093 	0.559 	1.96
pyrUp::Size_MatType::(320x240, 8UC3) 	3.553 	3.017 	1.18
pyrUp::Size_MatType::(320x240, 16SC3) 	3.342 	2.397 	1.39
pyrUp::Size_MatType::(320x240, 32FC3) 	3.911 	2.302 	1.70
pyrUp::Size_MatType::(320x240, 8UC4) 	5.381 	3.562 	1.51
pyrUp::Size_MatType::(320x240, 16SC4) 	4.493 	3.174 	1.42
pyrUp::Size_MatType::(320x240, 32FC4) 	5.149 	3.185 	1.62
pyrUp::Size_MatType::(640x480, 8UC1) 	4.442 	3.892 	1.14
pyrUp::Size_MatType::(640x480, 16SC1) 	4.107 	2.861 	1.44
pyrUp::Size_MatType::(640x480, 32FC1) 	4.654 	2.863 	1.63
pyrUp::Size_MatType::(640x480, 8UC3) 	15.506 	11.783 	1.32
pyrUp::Size_MatType::(640x480, 16SC3) 	13.772 	9.141 	1.51
pyrUp::Size_MatType::(640x480, 32FC3) 	15.384 	9.701 	1.59
pyrUp::Size_MatType::(640x480, 8UC4) 	20.438 	13.956 	1.46
pyrUp::Size_MatType::(640x480, 16SC4) 	19.403 	13.898 	1.40
pyrUp::Size_MatType::(640x480, 32FC4) 	22.509 	14.574 	1.54
pyrUp::Size_MatType::(1280x720, 8UC1) 	15.634 	11.194 	1.40
pyrUp::Size_MatType::(1280x720, 16SC1) 	13.798 	9.824 	1.40
pyrUp::Size_MatType::(1280x720, 32FC1) 	16.698 	10.182 	1.64
pyrUp::Size_MatType::(1280x720, 8UC3) 	45.080 	32.759 	1.38
pyrUp::Size_MatType::(1280x720, 16SC3) 	40.116 	29.394 	1.36
pyrUp::Size_MatType::(1280x720, 32FC3) 	45.339 	31.277 	1.45
pyrUp::Size_MatType::(1280x720, 8UC4) 	58.486 	41.172 	1.42
pyrUp::Size_MatType::(1280x720, 16SC4) 	58.284 	40.594 	1.44
pyrUp::Size_MatType::(1280x720, 32FC4) 	67.617 	41.300 	1.64

@fengyuentau
Copy link
Copy Markdown
Member

fengyuentau commented Mar 4, 2025

My perf on K1 (clang) and RK3568:
pyrUp-pyrDown-perf.zip

@asmorkalov asmorkalov merged commit cbcfd77 into opencv:4.x Mar 4, 2025
27 of 28 checks passed
@GenshinImpactStarts GenshinImpactStarts mentioned this pull request Mar 7, 2025
Merged
6 tasks
@asmorkalov asmorkalov mentioned this pull request Mar 11, 2025
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@asmorkalov asmorkalov asmorkalov approved these changes

@mshabunin mshabunin mshabunin approved these changes

@fengyuentau fengyuentau fengyuentau approved these changes

Assignees

@asmorkalov asmorkalov

Labels

optimization platform: riscv

Projects

None yet

Milestone

4.12.0

Development

Successfully merging this pull request may close these issues.

4 participants

@amane-ame @asmorkalov @fengyuentau @mshabunin