gitgat's profile - activity

Is there any set of recommanded flags to choose in order to get the most optimized code with the Intel c++ compiler?

for example I use /O3 /Arch:AVX /Oy /Ot /Qopt_matmul /fp:fast=2

However activating /Qparallel dramatically slows down the opencv functions!! What flags do you guys use to produce opencv release versions??

I got everything working including gpu. It seems that versions were compatible, the only problem was from openexr that I deactivated temporally.

I try to compile opencv last version under github with Intel c++ compiler under Visual Studio 8.

I generate the .sln with Cmake 2.8.11.2. when I compile with VS8 internal compiler, there is no error. However, when I compile using Intel Composer XE 2013 under VS8, I get many undefined errors in some .h files.

1. Am I doing something wrong or it is just some opencv bug that should be reported?

2. Is there any set of recommanded flags to choose in order to get the most optimized code with the Intel c++ compiler?

Thanks

Thanks for the answer.

I wonder why we dont have independent gpu functions such as gpu::resize etc. which really don't use objdetect or other modules. Maybe we can add a set of minimal gpu functions for GpuMat in core/gpumat.hpp??

It can be of great interest to have access to simple GpuMat functions everywhere in Opencv. what do you think?

I try to use gpu::resize in objdetect module. however, I dont know where to put

#include "opencv2/gpu/gpu.hpp"

When I include gpu.hpp in precomp.hpp, I get 157 link errors such as below. Can we actually use gpu functions in other modules???

Error 145 error LNK2019: unresolved external symbol "public: void __thiscall cv::gpu::ImagePyramid::build(class cv::gpu::GpuMat const &,int,class cv::gpu::Stream &)" (?build@ImagePyramid@gpu@cv@@QAEXABVGpuMat@23@HAAVStream@23@@Z) referenced in function "public: __thiscall cv::gpu::ImagePyramid::ImagePyramid(class cv::gpu::GpuMat const &,int,class cv::gpu::Stream &)" (??0ImagePyramid@gpu@cv@@QAE@ABVGpuMat@12@HAAVStream@12@@Z) opencv_objdetect_pch.obj Error 141 error LNK2019: unresolved external symbol "public: void __thiscall cv::gpu::CannyBuf::create(class cv::Size_<int> const &,int)" (?create@CannyBuf@gpu@cv@@QAEXABV?$Size_@H@3@H@Z) referenced in function "public: __thiscall cv::gpu::CannyBuf::CannyBuf(class cv::Size_<int> const &,int)" (??0CannyBuf@gpu@cv@@QAE@ABV?$Size_@H@2@H@Z) opencv_objdetect_pch.obj Error 117 error LNK2019: unresolved external symbol "public: virtual void __thiscall cv::SimpleBlobDetector::write(class cv::FileStorage &)const " (?write@SimpleBlobDetector@cv@@UBEXAAVFileStorage@2@@Z) referenced in function "[thunk]:public: virtual void __thiscall cv::SimpleBlobDetector::writevtordisp{4294967292,0}' (class cv::FileStorage &)const " (?write@SimpleBlobDetector@cv@@$4PPPPPPPM@A@BEXAAVFileStorage@2@@Z) opencv_objdetect_pch.obj Error 121 error LNK2019: unresolved external symbol "public: virtual void __thiscall cv::SimpleBlobDetector::read(class cv::FileNode const &)" (?read@SimpleBlobDetector@cv@@UAEXABVFileNode@2@@Z) referenced in function "[thunk]:public: virtual void __thiscall cv::SimpleBlobDetector::readvtordisp{4294967292,0}' (class cv::FileNode const &)" (?read@SimpleBlobDetector@cv@@$4PPPPPPPM@A@AEXABVFileNode@2@@Z) opencv_objdetect_pch.obj Error 139 error LNK2019: unresolved external symbol "public: virtual void __thiscall cv::OpponentColorDescriptorExtractor::write(class cv::FileStorage &)const " (?write@OpponentColorDescriptorExtractor@cv@@UBEXAAVFileStorage@2@@Z) referenced in function "[thunk]:public: virtual void __thiscall cv::OpponentColorDescriptorExtractor::write`vtordisp{4294967292,0}' (class cv::FileStorage &)const " (?write@OpponentColorDescriptorExtractor@cv@@$4PPPPPPPM@A@BEXAAVFileStorage@2@@Z) opencv_objdetect_pch.obj

Hi, I have many build errors for the lastest opencv under github: I use ipp 6, windows 64

I guess these variables are not part of ipp 6?

......\modules\core\src\dxt.cpp(1564) : error C2065: 'ippsDFTGetSize_R_32f' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1565) : error C2065: 'ippsDFTInit_R_32f' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1569) : error C2065: 'ippsDFTGetSize_R_64f' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1570) : error C2065: 'ippsDFTInit_R_64f' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1577) : error C2065: 'ippsDFTGetSize_C_32fc' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1578) : error C2065: 'ippsDFTInit_C_32fc' : undeclared identifier 7>tif_tile.c 5>......\modules\core\src\dxt.cpp(1582) : error C2065: 'ippsDFTGetSize_C_64fc' : undeclared identifier 5>......\modules\core\src\dxt.cpp(1583) : error C2065: 'ippsDFTInit_C_64fc' : undeclared identifier

When should I re-run cmake? for example do I need to run cmake each time I update my opencv code (pull) through Github? If not, so in what cases I should absolutely run cmake after updating opencv through Github?

I ask this, since running cmake and rebuilding all the debug and release libraries is very time consuming and I need to know when I should do this after updates!