OpenCV recompile with CuDNN for Win10 undeclared identifier error

cudawarped · February 22, 2023, 6:03am

I wouldn’t as it looks like you are mixing up versions of the OpenCV main and contrib repositoriess.

DIFFERENT_SIZES_EXTRA was added to the main repository in

github.com/opencv/opencv

core: add cuda `GpuMat` single dimension testing define

opencv:4.x ← cudawarped:fix_issue_3412

opened 04:28PM - 10 Jan 23 UTC

cudawarped

+2 -0

Add single dimension `GpuMat` testing define because single dimension `GpuMat`'s… memory is allocated differently. ### Pull Request Readiness Checklist 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. - [ ] The feature is well documented and sample code can be built with the project CMake

so if you have the test cases from

github.com/opencv/opencv_contrib

cudev: fix 1D error introduced in PR 3378

opencv:4.x ← cudawarped:fix_issue_3412

opened 04:18PM - 10 Jan 23 UTC

cudawarped

+57 -32

This fixes issue https://github.com/opencv/opencv_contrib/issues/3412 returning …incorrect results for single `GpuMat`'s with a single row, which was a result of an error in https://github.com/opencv/opencv_contrib/pull/3378. Unfortunatley because `GpuMat`'s with a single row or column are allocated with `cudaMalloc` and not `cudaMallocPitch` and 2D texture objects require pitched memory this PR introduces extra staging memory, when the row/col dimension is 1. This is a quick fix for the issue which only introduces extra overhead for single dimension `GpuMat`'s who's use should not be that common. A "better" fix would be to rework the `TexturePtr` class so that ``` __device__ __forceinline__ R operator ()(index_type y, index_type x) const; ``` staticly dispatches a call to `tex1Dfetch` when either x or y is 1, but this will be a more significant change. Dependant on https://github.com/opencv/opencv/pull/23126 ### Pull Request Readiness Checklist 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. - [ ] The feature is well documented and sample code can be built with the project CMake ``` force_builders=Custom buildworker:Custom=linux-1 docker_image:Custom=ubuntu-cuda:16.04 ```

but not the defines you are using a more recent commit from the contrib than the main repo.

I would rebuild from the tip of the master of both the main and contrib repositories.

Topic		Replies	Views
DNN GPU Broken CUDA Issues - Pls help Python dnn , build , cuda	14	2983	February 4, 2021
[ WARN:0] global ../modules/dnn/src/dnn.cpp (1363) setUpNet DNN module was not built with CUDA backend; switching to CPU After 5 OS installs and over 10 OpenCV builds C++ dnn , build	1	1097	October 28, 2021
setUpNet DNN module was not built with CUDA backend; switching to CPU C++ dnn , build , cuda	2	1286	November 21, 2022
OpenCV not build correctly for CUDA_ARCH C++ dnn , build , cuda	1	1189	December 1, 2021
OpenCV DNN working with CPU but not with CUDA C++ dnn , cuda	0	746	February 8, 2023

OpenCV recompile with CuDNN for Win10 undeclared identifier error

Related Topics