opencv 2.4.2 with gpu support crash in debug mode (Bug #2196)

Added by Jawad Rafid over 12 years ago. Updated over 12 years ago.

Status:Cancelled Start date:2012-07-25
Priority:Normal Due date:
Assignee:Anatoly Baksheev % Done:

0%
Category:gpu (cuda)
Target version:2.4.3
Affected version: Operating System:
Difficulty: HW Platform:
Pull request:

Description

Hi,

I am using opencv2.4.2 with visual studio 2010 build with cmake (CUDA enabled). The problem is i consistently get a crash when accessing gpu methods/initializations while being in debug mode. It works in release mode although that too needs to be run outside visual studio otherwise some crash occur. The exception that i receive is as follows:

First-chance exception at 0x76dbfbae in opencvGPUtest.exe: Microsoft C++ exception: cudaError_enum at memory location 0x00a3e258

The error is very consistent, and i have seen couple of examples over the internet with others (http://tech.groups.yahoo.com/group/OpenCV/message/77905) although nothing has helped solve the problem. It is very difficult to work without debug support while working with large projects, so kindly help me fix this.

Thanks,

CMakeCache.txt (148.7 kB) Jawad Rafid, 2012-07-25 01:34 pm

CMakeOutput.log - outputlog (894.7 kB) Jawad Rafid, 2012-07-25 06:46 pm

CMakeError.log - errorlog (310 kB) Jawad Rafid, 2012-07-25 06:46 pm

testGPUFlow.zip (882.1 kB) Jawad Rafid, 2012-07-25 09:21 pm

Associated revisions

Revision 810dc262
Added by Andrey Pavlenko about 11 years ago

Merge pull request #2196 from KonstantinMatskevich:ocl_bugcautching

History

#1
Updated by Marina Kolpakova over 12 years ago

We can't reproduce this bug. We are successfully build and execute our performance tests. Tested for Visual Studio 10 for both 32 and 64 architectures in debug mode.

  • Status changed from Open to Cancelled
  • Target version set to 2.4.3

#2
Updated by Jawad Rafid over 12 years ago

Could it be due to the version of opencv? Could you please check if there are any differences in cmakecache from the ones you are using to compile? It fails when for example i use feature detector with it such as :

cv::gpu::GpuMat prev_imgMatGPU;
cv::gpu::GoodFeaturesToTrackDetector_GPU gf(1000,0.01,0);
prev_imgMatGPU=cv::gpu::GpuMat(image); //This generate a crash , image is completely ok and can be visualized from imshow

Its not the type of detector which is causing the problem, instead if i use FAST_GPU instead i get error even earlier on the line where i declare gpu::GpuMat.

#3
Updated by Marina Kolpakova over 12 years ago

Could you provide CMake output log?

#4
Updated by Jawad Rafid over 12 years ago

Marina Kolpakova wrote:

Could you provide CMake output log?

please find cmakeoutput as well as error logs attached.

#5
Updated by Jawad Rafid over 12 years ago

Here i am attaching a project which can be used to replicate the error. Actually this is a sample from opencv for gpu optical flow tracker. It runs from command prompt while crashes in visual studio. Here the problem occurs when imshow/namedwindow is called. so i guess gpu is conflicting with imshow or maybe it is qt again.

#6
Updated by Jawad Rafid over 12 years ago

any reflections??

Also available in: Atom PDF