gpucv

- global: - remove lot of old code ...

allusse — Tue, 02 Oct 2007 16:45:44 +0100

Rev 126 - allusse (42 file(s) modified)

- global:
- remove lot of old code that was not used any more.
- add options: GPUCV_SETTINGS_CHECK_SHADER_UPDATE and GPUCV_SETTINGS_GLOBAL_NOTICE to GpuCVSettings => shader file control is now dynamic at runtime
- operators:
- move cvgSetCpuReturn/cvgUnsetCpuReturn to misc
- move some beta operators to CvgOperators.cpp
- GPUCVConsole:
- correct minor bugs in the benchloop macro
- rename/remove:
- remove cvgBaseOp cpp/h from SVN, it was empty.
- remove cvgImage cpp/h from SVN, it was replaced by CvgArr.h/cpp.
- remove GpuImageManager cpp/h from SVN, it was replaced by GpuTextureManager.h/cpp.

- operators: - cvg.h: reformat file to fit ...

allusse — Tue, 02 Oct 2007 11:38:56 +0100

Rev 124 - allusse (14 file(s) modified)

- operators:
- cvg.h: reformat file to fit OpenCV documentation order.
- misc.h: update documentation.
- cvgMaxS/cvgMinS: use now the TemplateOperator function.
- global:
- Build_release.bat: add script to update application revision from the SVN.
- GpuCVSettings: add fields for SVN version/data/URL.

- GPUCVConsole:

- rename/move/remove:
- cvgSynchronize/cvgSetLocation/cvgResizeGLSLFct functions moved to GPUCV/misc
- TemplateOperator functions moved to GPUCV/misc
- remove obsolete functions: Fct1Img1ArgTemp2/Fct1Img1ArgTemp/Fct2ImgTemp
- remove obsolete functions: ImgSizeComp2/ImgSizeComp3, ImgnChannelsComp2/ImgnChannelsComp3, ImgDepthComp2/ImgDepthComp3

- operators: - update cvgCop() to use only one ...

allusse — Thu, 27 Sep 2007 17:06:32 +0100

Rev 123 - allusse (70 file(s) modified)

- operators:
- update cvgCop() to use only one shader instead of 2.
- update all operators to use GPUCV_START_OP() and GPUCV_STOP_OP() macros.
- update GPUCV_START_OP() and GPUCV_STOP_OP() macros to allow dynamic switching between OPENCV and GPUCV, and dynamic benchmarking of active library.
- add macros IS_OPENCV_FORCED().
- all destination images are now called 'dst' instead of 'dest', to match OpenCV.
- global:
- automatic project generation using premake is now creating makefile for GCC.
- add GpuCVSettings class to store global settings for debugging or switching between GPUCV and OPENCV.
- TextureManager class: add Find() that return the object pointer and does not create a new object if not found.
- add cvgTerminate() to help destroying all objects and save benchmarks.
- _GPUCV_GL_ERROR_TEST is now using a dynamic switch with option GpuCVSettings::GPUCV_SETTINGS_GL_ERROR_CHECK.
- OpenGL errors can rise exceptions with flag GpuCVSettings::GPUCV_SETTINGS_GL_ERROR_RISE_EXCEPTION
- HardwareProfile: add The Profiler object from SugoiTracer as a member.
- add 2 macros for profiling GPUCV_PROFILE_CURRENT_FCT and GPUCV_PROFILE_FCT.
- shader path is now store into GpuCVSettings.
- GPUCVConsole:
- add commands to reach GpuCVSettings options.
- make it compatible with new GpuCVSettings parameters, specify default options at runtime.
- rename/remove:
- move Beep definition for linux to Tools.h.
- remove file cxcoreg.cpp that was empty, now all cxcore operators are into sub cpp files.
- move cvgCreateHistImage() to misc.h
- move cvgResizeGLSLFct to misc
- move ftostr to GPUCVHardware\tools.h

- move some functions to misc.h - correct minor bugs. - start ...

allusse — Tue, 25 Sep 2007 15:49:25 +0100

Rev 121 - allusse (22 file(s) modified)

- move some functions to misc.h
- correct minor bugs.
- start adding argument processing for console application, to call from the command prompt.

[changed made by Richard Viney] => port GPUCV on LINUX, ...

allusse — Mon, 24 Sep 2007 09:10:31 +0100

Rev 119 - allusse (92 file(s) modified)

[changed made by Richard Viney] => port GPUCV on LINUX, thanks.
- change all back slashes in #include directives to forward slashes
- fix capitalization in #include directives
- fixed non-standard use of the typename keyword
- added missing typename keyword where needed
- new lines at end of header and source files
- made sure class member variables are initialized in the order they
- are specified when doing direct initialization in the constructor
- in classes with multiple inheritance, when using a method defined in a parent class code should explicitly specify which parent class is being called
- define all xxxx_EXPORT macros on linux
- add cvAddWeighted()
- removed non used variables.
- other minor gcc compatibility fixes

- operators: - add cvgSetData. - ...

allusse — Mon, 24 Sep 2007 08:32:28 +0100

Rev 118 - allusse (11 file(s) modified)

- operators:
- add cvgSetData.
- update cvgSetZero to take care of actual image location.
- update cvgConvertScale to be processed on GPU.
- global:
- ShaderObject => meta shader parameters must now always start by $ => "$DEF_SCALAR=1//"
- add function SetLoggingOutput() to select logging ouput(console/file) dynamically.
- rename/remove:
- rename cvgSetRuntimePath into cvgSetShaderPath

operators: - add a few function that are ...

allusse — Tue, 18 Sep 2007 16:58:26 +0100

Rev 115 - allusse (12 file(s) modified)

operators:
- add a few function that are just here for compatibility (cvgSaveImage/cvgSetZero/cvgConvertScale), they should be ported on GPU later.
global:
- add a global variable Global_Gpucv_Runtime_Path(cvgSetRuntimePath()) to set default directory of shader files.

- Global changes - correct doxygen comments, \param ...

allusse — Mon, 17 Sep 2007 11:02:21 +0100

Rev 114 - allusse (107 file(s) modified)

- Global changes
- correct doxygen comments, \param and \notice tags. Move some comments to .h files.
- replace all IplImage by CvArr when needed, remove most of the _GPUCV_CORE_USE_CVGARR flags.
- add/reformat some documentation to all class/function.
- premake script: add include files for cvaux.

- cvg operators:
- cvgGEMM/matmul_add.frag: matrix multiplication, optimize multiplication to use a maximum of channels.
- cxcoreg.h => reorganize all function definition to looks like opencv documentation, please KEEP SAME ORGANIZATION.
- cvgShowImage() : add options to add image reduction to show image.
- move cvgOp2ImgTemp(), ImgSizeComp*(), ImgDepthComp*(), ImgnChannelsComp*() to gpucv/misc.h, theses function will be depracated soon.
- CvgArr : add more details to documentation.
- change cvgAvg to use mipmapping => faster.
- add cvgSum based on cvgAvg
- add cvgPow

- add/rename:
- add GPUCV_PERF_WARNING macro.
- add Beep when exception is rise in cvg operators.
- add functions GetWidth/GetHeight/getSize/GetDepth/GetnChannels to get properties of CvArr*
- add class TextSize to contain width/height of object...equivalent to CvSize.
- class GlContext : add field to store pointers to current OpenGL context, and a vector to push/pop contexts....
- add Auto_MIMAP option to CL_Texture.

- remove:
- TextureIOGrp

- GpuFilter/GpuFilterManager:
- SetShadersFiles() => string one for fragment shader, string 2 for vertex shader.
- All apply function now accept a pointer to a custom draw function
- ShaderObject :
- add field vertex_name/vertex_meta_tag/fragment_name/fragment_meta_tag.

Rev 110

allusse — Thu, 30 Aug 2007 10:19:46 +0100

Rev 110 - allusse (4 file(s) modified)

- Global changes - ...

allusse — Thu, 30 Aug 2007 10:17:31 +0100

Rev 109 - allusse (82 file(s) modified)

- Global changes
- remove all const qualifier from the cvgOperators, cause GpuCV can modify input CvArr* and IplImage*
- correct some mistakes in documentation.
- flag _GPUCV_CORE_USE_CVGARR is used to add CvArr support the GPUCV, we were only supporting IplImage. Allow support of CvMat now.
- class CvgImage has been replaced by CvgArr, CvgArr is equivalent of CvArr and manage IplImage and CvMat.
- GpuFilterManager/GpuImageManager/ShaderManager managers now unherit from CL_TEMPLATE_MANAGER. Cleaning object access method.
- GpuFilter/Cl_Texture/ShaderObject now unherits from CL_BASE_OBJ and can be managed by CL_TEMPLATE_MANAGER.
- add GPUCV_START_OP/GPUCV_STOP_OP macro used to begin/end operator function code, all existing operators should be updated
- all add/sub functions are now based on TemplateOperator() and are using only two shaders(add_all/sub_all.frag) with meta options.
- most of blocking errors are now managed by exception mechanism, GPUCV operators should now use GPUCV_START_OP/GPUCV_STOP_OP to handle exceptions. If errors occurs while processing an operator, all local images are read back to CPU, and original CV operator is call instead.

- add/rename:
- cvgSynchronize that copy back GPU data to CPU.
- TemplateOperator generic function used to call simple shader operators, can be use in most cases.
- cvgScaleAdd and cvGEMM function, still in test.
- add cvgConvertCVMatrixFormatToGL/cvgConvertCVMatrixFormatToGL to manage matrix format conversion to GL.
- add cvgReleaseMat/ cvgCreateMat
- rename cvgSetImageLabel/cvggetImageLabel to cvgSetLabel/cvgGetLabel

- GpuFilter:
- move some documentation from some .cpp files to .h files.
- update the Parameter functions from GpuFilter, UpdateFilterParams became => SetParams.
- rename CvgApply to Apply.
- Apply(TextureGrp*) is now working, but still a bit slower that Apply(IplImage*)

- ShaderObject :
- add SplitFileName() and PrintInfo() functions. printInfoLog() show debug info about meta parameters of shaders.

- CL_Options has now virtual members functions.
- Cl_Texture:
- add number of channel member and Access function, useful for matrix.
- affecting image size > 2048 now raise exceptions.

- TextureGrp:
- add flag TEXTUREGRP_USE_VECTOR for using std::vectors or [] of textures. Array of texture is faster!
- unherits from CL_Options, so it can manipulate options of all contained textures.
- add macros TEXTURE_GRP_INTERNE_DO_FOR_ALL/TEXTURE_GRP_EXTERNE_DO_FOR_ALL to loop into each textures.

- mainSampletest application
- runAdd function corrected to manage mask, and add support to call cvgScaleAdd.
- add a global image called GlbMask, so operators using mask don't need to create their own mask.
- replace all functions cvCreateImage by cvgCreateImage()
- add operators to bench (matmul, matmuladd, addscale,
- move some cxcoreg operators to cxcoreg_test.cpp
- optimize image resizing in the benchmark loop.

add mat(matrix) sub dir

allusse — Thu, 30 Aug 2007 09:31:03 +0100

Rev 108 - allusse (2 file(s) modified)

add mat(matrix) sub dir

- add add_all/sub_all shader that contains all add/sub operators

allusse — Thu, 30 Aug 2007 09:27:40 +0100

Rev 107 - allusse (2 file(s) modified)

- add add_all/sub_all shader that contains all add/sub operators

- correct threshold operators

allusse — Thu, 30 Aug 2007 09:26:57 +0100

Rev 106 - allusse (5 file(s) modified)

- correct threshold operators

Rev 105

allusse — Thu, 30 Aug 2007 09:26:25 +0100

Rev 105 - allusse (1 file(s) modified)

- add some matrix shader operation. - operator cvgGEMM started, matrix ...

allusse — Wed, 22 Aug 2007 09:29:10 +0100

Rev 104 - allusse (7 file(s) modified)

- add some matrix shader operation.
- operator cvgGEMM started, matrix multiplication almost finished.

Rev 103

allusse — Mon, 20 Aug 2007 12:35:32 +0100

Rev 103 - allusse (18 file(s) modified)

- commit before merge

allusse — Mon, 20 Aug 2007 11:14:36 +0100

Rev 100 - allusse (4 file(s) modified)

- commit before merge

- add copy constructor to CL_Texture and CvgImage. - update cvgCloneImage ...

allusse — Thu, 09 Aug 2007 18:02:23 +0100

Rev 95 - allusse (18 file(s) modified)

- add copy constructor to CL_Texture and CvgImage.
- update cvgCloneImage and cvgCopy functions
- make some correction into mainSampletest.cpp

- update to new changes in GPUCVTextures. - clean and reformat ...

allusse — Thu, 09 Aug 2007 09:34:15 +0100

Rev 94 - allusse (43 file(s) modified)

- update to new changes in GPUCVTextures.
- clean and reformat some function in mainSampletest.cpp

- FBO: add m_textureType field so we use real texture type ...

allusse — Thu, 09 Aug 2007 09:29:21 +0100

Rev 93 - allusse (14 file(s) modified)

- FBO:
add m_textureType field so we use real texture type and not default one.
remove Push/Pop oengl attrib from FBO, they are done in TextureRenderBufferXXXX
- CL_Texture:
add more texture location options(LOC_NO_LOCATION, LOC_CUDA, LOC_LAST, LOC_ALL).
add new texture option (UBIQUITY, CPU_RETURN, DBG_IMG_TRANSFERT, DBG_IMG_FORMAT, DBG_IMG_LOCATION).
add unheritance from CL_Option(push/pop).
all function directly related to opengl and/or that are not SMART function and will not perform smart GPU <=> CPU transfert are renamed with "_" prefixe. They should not be called by normal user unless you know what you are doing.
function removed : ReadFromFrameBuffer/WriteToFrameBuffer and replaced by FrameBufferToCpu/FrameBufferToGpu/GpuToFrameBuffer/CpuToFrameBuffer
add functions for location managment : _AddLocation/_CreateLocation/_ForceLocation/_RemoveLocation/_GetLocation/_IsLocation/_SetUbiquity

- TextureGrp::ProcessCPUReturn() => updated for the new SetLocation mechanism