System information (version)

• OpenCV => 4.5.2
• Operating System / Platform => Linux aarch64 / i.mx 8M board
• Compiler => gcc

Detailed description

[ FAILED ] , where GetParam() = (8UC3, 1280x720, SAME_TYPE, 0xaaaac93df890, MUL, false, 0.5, false)
[ FAILED ] MulTestGPU/MathOpTest.MatricesAccuracyTest/79, where GetParam() = (8UC3, 1280x720, 8UC1, 0xaaaac93df890, MUL, false, 0.5, false)
[ FAILED ] MulTestGPU/MathOpTest.MatricesAccuracyTest/97, where GetParam() = (8UC3, 640x480, SAME_TYPE, 0xaaaac93df890, MUL, false, 0.5, false)
[ FAILED ] MulTestGPU/MathOpTest.MatricesAccuracyTest/103, where GetParam() = (8UC3, 640x480, 8UC1, 0xaaaac93df890, MUL, false, 0.5, false)
[ FAILED ] MulTestGPU/MathOpTest.MatricesAccuracyTest/121, where GetParam() = (8UC3, 128x128, SAME_TYPE, 0xaaaac93df890, MUL, false, 0.5, false)
[ FAILED ] MulTestGPU/MathOpTest.MatricesAccuracyTest/127, where GetParam() = (8UC3, 128x128, 8UC1, 0xaaaac93df890, MUL, false, 0.5, false)

These issues are occured on gapi with OpenCL. I found these issues are caused by difference of round method between ocv and ocl.

• ocv always round down
• gapi use ocl method convert_uchar4_sat_rte. It will round down to nearest even and round up to nearest value.

I don't know if it's real a problem, the round to even method used by ocl seems more scientific, it can reduce systematic bias.

Please let me know how you will take care of these failures, Thankyou.

Steps to reproduce

bin/opencv_test_gapi --gtest_filter=MulTestGPU/MathOpTest.MatricesAccuracyTest/73
MulTestGPU_MathOpTest_MatricesAccuracyTest_73_data.zip

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues,
  forum.opencv.org, Stack Overflow, etc and have not found solution
• I updated to latest OpenCV version and the issue is still there
• There is reproducer code and related data files: videos, images, onnx, etc