Skip to content

Impl RISC-V HAL for norm_hamming#26918

Merged
asmorkalov merged 1 commit intoopencv:4.xfrom
GenshinImpactStarts:norm_hamming
Feb 26, 2025
Merged

Impl RISC-V HAL for norm_hamming#26918
asmorkalov merged 1 commit intoopencv:4.xfrom
GenshinImpactStarts:norm_hamming

Conversation

@GenshinImpactStarts
Copy link
Copy Markdown
Contributor

@GenshinImpactStarts GenshinImpactStarts commented Feb 13, 2025

Implement through the existing cv_hal_normHamming8u and cv_hal_normHammingDiff8u interfaces.

Modified modules/core/src/norm.cpp:680 to ensure HAL calls are not bypassed.

Tested on

  • MUSE-PI
  • Compiler: gcc 14.2 (riscv-collab/riscv-gnu-toolchain Nightly: December 16, 2024)
$ opencv_test_core --gtest_filter="Core_Norm/ElemWiseTest*"
$ opencv_perf_core --gtest_filter="PerfHamming*" --perf_min_samples=300 --perf_force_samples=300

Compare with scalar:

Geometric mean (ms)

                    Name of Test                     scalar  hal      hal    
                                                                       vs    
                                                                     scalar  
                                                                   (x-factor)
norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480)   0.962  0.128     7.52   
norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080) 6.506  0.803     8.10   
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 640x480)    0.964  0.137     7.05   
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080)  6.426  0.660     9.74   
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480)    0.606  0.067     9.04   
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080)  4.122  0.427     9.65   
norm::PerfHamming::(NORM_HAMMING, 8UC1, 640x480)     0.610  0.049    12.55   
norm::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080)   4.135  0.333    12.43

Compare with ui:

Geometric mean (ms)

                    Name of Test                       ui    hal      hal    
                                                                       vs    
                                                                       ui    
                                                                   (x-factor)
norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480)   0.734  0.128     5.74   
norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080) 4.915  0.803     6.12   
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 640x480)    0.716  0.137     5.24   
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080)  4.771  0.660     7.23   
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480)    0.690  0.067    10.29   
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080)  4.670  0.427    10.93   
norm::PerfHamming::(NORM_HAMMING, 8UC1, 640x480)     0.652  0.049    13.40   
norm::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080)   4.433  0.333    13.33

While working on this, I noticed that the optimization logic in cv::norm could be improved:

  1. The recently introduced cv_hal_norm is too restrictive. I think optimization also can be performed after unfolding using map iteration.
  2. cv_hal_norm overlaps with cv_hal_normHamming8u, and due to its strictness, I implemented the optimization through the latter.
  3. cv::hal::normHamming has four overloads—two in norm.cpp and two in stat.dispatch.cpp. The latter should only be used by the former, as CALL_HAL_RET is present only in the two in norm.cpp. Since they have similar parameters, to avoid confusion, I suggest renaming the two in stat.dispatch.cpp.

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

@asmorkalov asmorkalov requested a review from mshabunin February 13, 2025 18:27
@asmorkalov asmorkalov added this to the 4.12.0 milestone Feb 13, 2025
asmorkalov
asmorkalov reviewed Feb 15, 2025
View reviewed changes
@asmorkalov
Copy link
Copy Markdown
Contributor

asmorkalov commented Feb 24, 2025

Please rebase and fix conflicts.

@GenshinImpactStarts @hanliutong
impl hal_rvv norm_hamming
33d632f 
Co-authored-by: Liutong HAN <liutong2020@iscas.ac.cn>
@asmorkalov
Copy link
Copy Markdown
Contributor

asmorkalov commented Feb 25, 2025

Perf results for Spacemit Muse Pi v30 (GCC 14.2):

norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480) 	0.740 	0.128 	5.80
norm2::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080) 	4.925 	0.805 	6.11
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 640x480) 	0.724 	0.125 	5.80
norm2::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080) 	4.842 	0.656 	7.38
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 640x480) 	0.698 	0.063 	11.01
norm::PerfHamming::(NORM_HAMMING2, 8UC1, 1920x1080) 	4.706 	0.429 	10.98
norm::PerfHamming::(NORM_HAMMING, 8UC1, 640x480) 	0.656 	0.047 	13.87
norm::PerfHamming::(NORM_HAMMING, 8UC1, 1920x1080) 	4.506 	0.343 	13.16

@asmorkalov asmorkalov self-assigned this Feb 25, 2025
@asmorkalov asmorkalov merged commit 3db1247 into opencv:4.x Feb 26, 2025
27 of 28 checks passed
@asmorkalov asmorkalov mentioned this pull request Mar 4, 2025
Merged
@GenshinImpactStarts GenshinImpactStarts deleted the norm_hamming branch March 12, 2025 07:32
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 Awaiting requested review from mshabunin

Assignees

@asmorkalov asmorkalov

Labels

optimization platform: riscv

Projects

None yet

Milestone

4.12.0

Development

Successfully merging this pull request may close these issues.

3 participants

@GenshinImpactStarts @asmorkalov @dkurt