| Log message: | * Applications:
* GpuCVConsole:
* add cxcoreg_test.h and cvg_test.h to be headers for cxcore_test.cpp and cvg_test.cpp
* Add cxcore_processCommand() and cv_processCommand() to cxcore_test.cpp and cvg_test.cpp to allow local command parsing into each part of the library.
* start cleaning GPUCVConsole/mainSampleTest.h
* Addd _GCU_BENCH_PRELOAD_INTO_ARRAY and _GCU_BENCH_PRELOAD_INTO_BUFFER that can be used to define CUDA data preloading when benchmarking, should be 1 during normal use.
* Improve TestClone()/runCloneTest() and TestTransfer()/RunDataTransferTest() to be compliant with GpuCVSelectionMask mechanism and give more debug informations.
* Correct __CreateImages__ depth in runMaxs/runMins/runMinMaxLoc/runDilate/runErode/runMorpho/runSobel/runLaplace/runSmooth/runThreshold to allow any image format test.
* GpuCVSimpleApp: add a simple example to test GpuCV.
* GPUCVCuda:
* create a dedicated struct StAffectFilterTex for KERNEL_ARITHM_OPER_AFFECT.
* Add some function to reshape data to improve efficencie
* Add template kernels GCUDA_KRNL_ARITHM_1BUFF_MASK/GCUDA_KRNL_ARITHM_2BUFF_MASK
* Renamed function cvgCudaInit into cvgcuInit.
* CudaArithm_SWITCHALL can be used with device buffer or CUDA array.
* GPUCVTexture:
* DataContainer:
* correct use of m_textureLastLocations
* remove some deprecated functions.
* DataContainer dont inherit from TextSize<> any more, TextSize is moved to the DataDsc_Base level to have a closer control to it into each objects.
* DataDsc_*:
* m_nChannelsBackup is introduced to allow data repacking, adapt size to have 4 channels when it is possible, see SetReshape()/GetReshape()
* All Call the GetParent() are now secure so DataDsc_* can be used even if they have no parent.
* DataDsc_GLTex:
* Add auto mipmap flag.
* Correct cloning.
- GpuCVCore:
* ShaderObject::ParseShaderCode() do not raise exceptions when meta-tag are not found, only a Warning message.
- Global:
* Major Bug corrected: In conversion functions, format short/GL_SHORT/IPL_DEPTH_16* where not manage correctly and mixed with format GL_INT/IPL_DEPTH_32*.
- operators:
* cvgCreateImage: add internal macro to be benchmarked correctly
|
