Skip to content

cuda optflow TVL1 : run safely in async mode#17556

Merged
opencv-pushbot merged 2 commits intoopencv:3.4from
nglee:dev_optFlowTVL1Async
Jun 16, 2020
Merged

cuda optflow TVL1 : run safely in async mode#17556
opencv-pushbot merged 2 commits intoopencv:3.4from
nglee:dev_optFlowTVL1Async

Conversation

@nglee
Copy link
Copy Markdown
Contributor

@nglee nglee commented Jun 15, 2020 •
edited by alalek
Loading

Fixes #16013

We are going to use texture object, rather than texture reference.
It allows us to run the algorithm asynchronously.
This is similar to what's been done in #11483

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 OpenCV (BSD) License.
  • 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
  • The PR is proposed to proper branch
  • There is reference to 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
allow_multiple_commits=1
force_builders=Custom
buildworker:Custom=linux-4
build_image:Custom=ubuntu-cuda:18.04

@alalek
Copy link
Copy Markdown
Member

alalek commented Jun 15, 2020

@nglee Thank you for the contribution!

Could you please take a look on this error message (from Custom builder)?

/build/precommit_custom_linux/3.4/opencv/modules/cudaimgproc/src/cuda/canny.cu:93:8: error: 'struct canny::SrcTex' has virtual functions and accessible non-virtual destructor [-Werror=non-virtual-dtor]
     struct SrcTex
        ^~~~~~
/build/precommit_custom_linux/3.4/opencv/modules/cudaimgproc/src/cuda/canny.cu:104:8: error: base class 'struct canny::SrcTex' has accessible non-virtual destructor [-Werror=non-virtual-dtor]
     struct SrcTexRef : SrcTex
        ^~~~~~~~~
/build/precommit_custom_linux/3.4/opencv/modules/cudaimgproc/src/cuda/canny.cu:104:8: error: 'struct canny::SrcTexRef' has virtual functions and accessible non-virtual destructor [-Werror=non-virtual-dtor]
/build/precommit_custom_linux/3.4/opencv/modules/cudaimgproc/src/cuda/canny.cu:114:8: error: base class 'struct canny::SrcTex' has accessible non-virtual destructor [-Werror=non-virtual-dtor]
     struct SrcTexObj : SrcTex
        ^~~~~~~~~
/build/precommit_custom_linux/3.4/opencv/modules/cudaimgproc/src/cuda/canny.cu:114:8: error: 'struct canny::SrcTexObj' has virtual functions and accessible non-virtual destructor [-Werror=non-virtual-dtor]
cc1plus: some warnings being treated as errors

(perhaps it is enough to define destructor of the base class as virtual)

@nglee nglee force-pushed the dev_optFlowTVL1Async branch from 629d006 to 69a298b Compare June 15, 2020 18:47
@nglee
Copy link
Copy Markdown
Contributor Author

nglee commented Jun 15, 2020 •
edited
Loading

@alalek Build error is now fixed!

@cognex-nalee
cuda optflow tvl1 : async safety
2043e06 
also modify cuda canny to use createTextureObjectPitch2D, etc.
@nglee nglee force-pushed the dev_optFlowTVL1Async branch from 69a298b to 2043e06 Compare June 16, 2020 16:06
@nglee
Copy link
Copy Markdown
Contributor Author

nglee commented Jun 16, 2020

It is ready to be merged. I've also tested with OP's test code in #16013.

alalek
alalek approved these changes Jun 16, 2020
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.

Well done! Thank you for contribution 👍

@opencv-pushbot opencv-pushbot merged commit 9755ab1 into opencv:3.4 Jun 16, 2020
@nglee nglee deleted the dev_optFlowTVL1Async branch June 16, 2020 23:16
@tomoaki0705
Copy link
Copy Markdown
Contributor

tomoaki0705 commented Jun 17, 2020

Hi @nglee , I see build error on Jetson TX1 and TX2.

/opencv-fork/modules/cudaimgproc/src/cuda/canny.cu(131): error: namespace "std" has no member "enable_if_t"

/opencv-fork/modules/cudaimgproc/src/cuda/canny.cu(131): error: expected a "," or ">"

The direct cause come from this line

https://github.com/opencv/opencv/blame/9755ab160d94f4ad2a672653cf657ad421d46837/modules/cudaimgproc/src/cuda/canny.cu#L131

We only enable C++11 feature on master, not on 3.4, but it seems enable_if_t is a c++14 feature

template< bool B, class T = void > using enable_if_t = typename enable_if<B,T>::type; |   | (since C++14)

I appreciate if you could have a look on this.
Still it's unknown why I didn't face this on Jetson Xavier and Jetson Nano.

@nglee
Copy link
Copy Markdown
Contributor Author

nglee commented Jun 17, 2020

@tomoaki0705

Thank you for investigation. Would you please check your g++ version?

The default mode is C++98 for GCC versions prior to 6.1, and C++14 for GCC 6.1 and above. (Link)

To enable c++14 via cmake, maybe we can try this flag: -DCMAKE_CXX_STANDARD=14.

Just removing the lines with enable_if_t would also work.

@tomoaki0705
Copy link
Copy Markdown
Contributor

tomoaki0705 commented Jun 17, 2020

Jetson TX1 GCC 5.4.0
Jetson TX2 GCC 5.4.0
Jetson Nano GCC 7.5.0
Jetson Xavier GCC 7.5.0
That makes sense.

@tomoaki0705 tomoaki0705 mentioned this pull request Jun 18, 2020
Merged
6 tasks
This was referenced Jun 18, 2020
Merged
Merged
@alalek alalek mentioned this pull request Aug 21, 2020
Closed
4 tasks
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

@asmorkalov asmorkalov Awaiting requested review from asmorkalov

Assignees

@alalek alalek

Labels

bug category: gpu/cuda (contrib) OpenCV 4.0+: moved to opencv_contrib

Projects

None yet

Milestone

3.4.11

Development

Successfully merging this pull request may close these issues.

Corrupted optical flow using cuda::DenseOpticalFlow asynchronously in multithreaded environment

5 participants

@nglee @alalek @tomoaki0705 @opencv-pushbot @cognex-nalee