New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
medianBlur slowness with grayscale images #7199
Comments
Actually I get ~200x faster processing in second case (with this sample):
Windows 10 - 64 bit, VS2015 - 64 bit, latest OpenCV version. Please try to reproduce the issue with latest version. Add more information about your OpenCV compilation environment (getBuildInformation function). |
Thank you for your reply @mshabunin |
@sbaush , yes, I'm using custom build. When I wrote "latest version" I meant latest commit from GitHub. Can you check it? I tried to reproduce with 3.1 - similar results:
And latest commit (release version, I used debug in previous post):
One more thought: try to disable OpenCL by setting environment variable |
@mshabunin I measure time using getTickCount I have another question: how is it possible that your third time is faster than the first? Here you can check my output:
The output with OPENCV_OPENCL_DEVICE=disabled is (in my opinion it is the same as before):
Obviously I tested also with other images, and the behavior is always the same. |
That's because first result was measured in Debug build. This is strange, please provide information about your hardware (CPU, RAM). Try to run with disabled IPPICV: it can be turned off by setting environment variable: |
With the environment variable: Here you can find the output of this test:
|
@sbaush , yes probably you should turn IPP off. I'd recommend you to build OpenCV and run performance test with IPP turned off and on and compare results to investigate the issue. There can be a problem with system configuration, conflict with some components etc.. I checked master version on OSX on Xeon E5426 with IPP enabled and it works well:
cc @lupustr3 |
@mshabunin I tested the executable with the median test in another 2 PC. PC with bug
PC without bug
|
@sbaush , thank you for investigation. What CPUs are installed in these machines? |
@mshabunin here you can find all the info about the 3 PC that I tested. PC with slow median blur |
@sbaush Thank you for your investigation! From the fist look it seems like IPP-ICV package flaw. I will escalate this with IPP team and we will try to find appropriate solution for such problems. I cannot check this right now, but here are my thoughts: |
System information (version)
Detailed description
src image: 2000x2000 bmp color image
median blur with ksize = 5
time elapsed ~70ms
src image: 2000x2000 bmp grayscale image
median blur with ksize = 5
time elapsed ~3700ms
Steps to reproduce
The text was updated successfully, but these errors were encountered: