Skip to content

improve map allocation check#19529

Merged
alalek merged 4 commits intoopencv:3.4from
WeiChungChang:3.4
Feb 23, 2021
Merged

improve map allocation check#19529
alalek merged 4 commits intoopencv:3.4from
WeiChungChang:3.4

Conversation

@WeiChungChang
Copy link
Copy Markdown
Contributor

@WeiChungChang WeiChungChang commented Feb 14, 2021 •
edited by alalek
Loading

Move map resize to the beginning of loop to avoid unnecessary check within inner loop.

Google benchmark test:
Vgg16 Faster RCNN model, detection output layer is used by

  1. proposal output; where we have 1 location class and 9 * 14 * 14 = 1764 prediction per class.
  2. Output layer: where we have 21 location classes and 100 prediction per class.

According to test on VGG16 faster RCNN, there are speedup of:
65.38% (17000ns / 10279 ns)
45.22% (30308 / 20870 ns)

the benchmark program is at the link and compiled by:
https://drive.google.com/file/d/1sBzXPuCLEekjoWh85k4ZUKlgqQrJcXsU/view?usp=sharing

g++ test.cpp -std=c++14 -O3 -I<benchmark>/benchmark/include -L<benchmark>/build/src -lbenchmark -lpthread -o 
benchmark

Run on (12 X 4600 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x6)
  L1 Instruction 32 KiB (x6)
  L2 Unified 256 KiB (x6)
  L3 Unified 12288 KiB (x1)
Load Average: 0.22, 0.33, 0.54
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
***WARNING*** Library was built as DEBUG. Timings may be affected.

1x1764
---------------------------------------------------------------------
Benchmark                           Time             CPU   Iterations
---------------------------------------------------------------------
BM_FasterRCNNProposl            10279 ns        10279 ns        70308
BM_FasterRCNNProposlOrigin      17001 ns        17000 ns        41550

21x100
---------------------------------------------------------------------
Benchmark                           Time             CPU   Iterations
---------------------------------------------------------------------
BM_FasterRCNNProposl            20871 ns        20870 ns        35424
BM_FasterRCNNProposlOrigin      30309 ns        30308 ns        23321

Pull Request Readiness Checklist

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

  • [V] I agree to contribute to the project under Apache 2 License.
  • [V] To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
  • [V] The PR is proposed to proper branch
  • There is reference to original bug report and related work
  • [V] 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
force_builders=linux,docs

alalek
alalek reviewed Feb 20, 2021
View reviewed changes
alalek
alalek approved these changes Feb 23, 2021
View reviewed changes
Copy link
Copy Markdown
Member

@alalek alalek left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you 👍

@alalek alalek merged commit d4d1216 into opencv:3.4 Feb 23, 2021
@alalek alalek mentioned this pull request Feb 27, 2021
Merged
@alalek alalek mentioned this pull request Apr 9, 2021
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@alalek alalek alalek approved these changes

Assignees

@alalek alalek

Labels

category: dnn optimization

Projects

None yet

Milestone

3.4.14

Development

Successfully merging this pull request may close these issues.

2 participants

@WeiChungChang @alalek