gpucv

doc: * add draft tutorial for NVIDIA Parallel ...

allusse — Sun, 04 Jul 2010 12:33:18 +0100

Rev 567 - allusse (107 file(s) modified)

doc:
* add draft tutorial for NVIDIA Parallel Nsight
* improve switch reference manual
CUDA:
* custom build rules, add option to choose one or several SM targets
* add function cvgCudaSetProcessingDevice() to choose CUDA device before init. (experimental)
* change opengl interoperability and mapping with gl buffers.
* use pragma unroll on all arithmetic and logic operators, gain of 20%
SWITCH:
* add function to register profiler singleton into plugin DLL (to be applied to other lib), see cvg_CXCORESWITCH_RegisterTracerSingletons

GLOBAL:
* add _GPUCV_GL_USE_GLUT cause initializing GpuCV with glut does not work with Parallel Nsight
* add function cvgFlush() to Synchronize OpenGL / CUDA context based on one image location.
* add class DataDsc_GLBasse to be the base class of all opengl descriptors
Switch:
* fix switch, add support for multiple dll versions
Demo:
* fix and improve CamDemo
* add some switch instructions to console app, still experimental
Bug:
* cvgCudaFlip() fail on Windows 7 64 on macbook pro with geforce 9600

Global: * remove some obsolete scripts from ./bin/ ...

allusse — Mon, 21 Jun 2010 08:47:48 +0100

Rev 564 - allusse (338 file(s) modified)

Global:
* remove some obsolete scripts from ./bin/
* Benchmarks are now saved by default in ./data/Benchmarks
* move plugin GL/CUDA into the plugin folder
Switch:
* Fix switch mechanism under windows. Switch config files are now in ./data/
* compatible with opencv 2.1
* export the switch generator into a new binary project: genswitch
* Add DllDescriptor class to have several DLL file depending on versions/architecture/debug...etc...
* Add functions cvQueryFrame/cvRetrieveFrame in the forced list that use OpenGL. This enable data synchro on video streams.
CamDemo:
* small improve/cleaning/doc update. Now shows OpenCV/OpenGL/CUDA logos
Console:
* Disable switch generation.
Doc:
* Update setup page.
Lib:
* add cvgRetrieveShaderPath() to retrieve full path to data folder
* fix cvgInitGLView(), data were not transfered when images was had DESTINATION flag on.

- update Parabuild tutorial - clean some warning in code.

allusse — Fri, 11 Jun 2010 12:52:34 +0100

Rev 555 - allusse (15 file(s) modified)

- update Parabuild tutorial
- clean some warning in code.

CUDA: * move custom build rules for VS into etc/. They ...

allusse — Wed, 09 Jun 2010 15:47:05 +0100

Rev 553 - allusse (69 file(s) modified)

CUDA:
* move custom build rules for VS into etc/. They are now detected automatically when generating solutions
* Fixed bug in GetCvgArr(), the local cache pointer could not be updated if the cache image is deleted. error occured when a new image had the same pointer. -> local cache disabled.
DEPENDENCY:
* update denpendency lib/dlls for x32 and x64
DOC:
* add FAQ: how to compile OpenCV2.1 for Win64
* Add tutorial about ParaBuild
PREMAKE:
* move some internal function to dependencies/sugoitools/etc/premake/functions.premake4.lua
DEMO:
*CamDemo: add command line parameters to select video file and camera. Cam is not working under Win7-64, what about other OS?
*Console:
* add enable/disable flag for cv/cxcore/transfertest, it enable/disable the corresponding commands, useful to perform partial benchmarks.
* fixed the transfer test function.
* [B] memory leak when transfering data to/from DataDsc_GLBuff

GLOBAL:
* fix some logging indentation issues.

CUDA: * [U] update cuda makefiles to compile ...

allusse — Thu, 03 Jun 2010 14:00:47 +0100

Rev 550 - allusse (46 file(s) modified)

CUDA:
* [U] update cuda makefiles to compile with CUDPP/CUFFT
* [F] fixed cvgCudaDFT().
PREMAKE:
* [A] premake command 'getpath' added, it return path/LD_LIBRARY_path to set.
* [A] env variable are controlled and return error messages if not set correctly
GLOBAL:
* [U] update' operators header files to be C compliant. More work to do here.
* [A] add new example to test C integration.
bench:
* [A] add OpenCV version to benchmark parameters

moved back branch allusse/cuda-3.0 to experimental/trunk

allusse — Mon, 31 May 2010 16:14:12 +0100

Rev 547 - allusse (2 file(s) modified)

moved back branch allusse/cuda-3.0 to experimental/trunk

* [U] update and clean allusse experimental ...

allusse — Mon, 31 May 2010 14:23:28 +0100

Rev 545 - allusse (4 file(s) modified)

* [U] update and clean allusse experimental branch to keep only CUDA version 3.0

CUDA: * remove obsolete CUDA DLLs. ...

allusse — Wed, 26 May 2010 16:31:02 +0100

Rev 544 - allusse (77 file(s) modified)

CUDA:
* remove obsolete CUDA DLLs.
* Update GpuCV-CUDA documentation
GLOBAL:
* remove all version information from reference manual, just kept it in documentation and tutorials

GPUVCamDemo: *[F] Fixed issue with missing library ...

allusse — Tue, 27 Apr 2010 16:17:43 +0100

Rev 532 - allusse (197 file(s) modified)

GPUVCamDemo:
*[F] Fixed issue with missing library
*[I] Demo is based on the switch mechanism that is currently not fully operational
GPUCVConsole:
*[U] improve operator self test procedure, better log with color(under windows).
*[U] using new version of SugoiTracer, benchmarks SVG files are much better and interactive.
GLSL:
* [F] Fixed operators cvgMerge()/...?
* [A] Allow filter to define their own texture name in shader files.
CUDA:
* [A] link with CUBLAS from CUDA.
* [F] Fixed operators: cvgCudaSobel()/cvgCudaAddWeighted()/cvgCudaScaleAdd()
* [A] Add operators: cvgCudaSet()/cvgCudaSetZero()
* [U] Update arithm & logics operators.
* [A] add macro GCU_OPER_CHECK_ERROR to check kernels errors.
* [U] set deprecated: gcuSetReshapeObj()/gcuUnsetReshapeObj(), reshape is done locally.
GLOBAL:
* [U] clean old header that were not included any more
* [U] clean all ifndef _LINUX that were not necessary
* [U] remove all local inline definitions to have only a global one: __GPUCV_INLINE
* [A] add several exception class to manage error/compatibility/todo exceptions
* [U] update all opeartors to match new exception handling

Global: * [M] Moved some data files from ...

allusse — Tue, 30 Mar 2010 12:18:03 +0100

Rev 528 - allusse (222 file(s) modified)

Global:
* [M] Moved some data files from ./bin to ./data such as: GLSL shader files, test images...
* [D] delete old project files (premake3)
* [U] replace all Namespace macro by real namespace tags, namespace name is GCV.
* [A] add new Assert Macros: GCV_OPER_ASSERT/GCV_OPER_COMPAT_ASSERT to be used in operators to make difference between execution errors(empty pointers...) and compatibility issues with hardware/libraries
CUDA:
* [U] reorganize some operators into the corresponding files

CUDA: * [F] fixed cvgCudaThreshold()/cvgCudaCalcHist()/cvgCudaDFT()/cvgCudaNot() and ...

allusse — Fri, 19 Mar 2010 23:38:09 +0100

Rev 524 - allusse (98 file(s) modified)

CUDA:
* [F] fixed cvgCudaThreshold()/cvgCudaCalcHist()/cvgCudaDFT()/cvgCudaNot() and cvgCudaScaleAdd().
* [U] move some cuda operators to the right files and clean some code.
* [U] re-enable CUFFT support.
* [I] Deriche()/Erode()/Dilate()/Laplace()/Sobel are still not working.
GLSL:
* [F] fixed cvgFlip()

BUILD: * [U] the internal preprocessing macros used ...

allusse — Mon, 15 Mar 2010 17:25:31 +0100

Rev 521 - allusse (136 file(s) modified)

BUILD:
* [U] the internal preprocessing macros used for Windows/Linux and MacOS are set to _WINDOWS/_LINUX/_MACOS.
* [F]fix some 32/64 bit compiling issues in the premake4.lua file.

cxcoregcu:
* [F] fix operators: cvgCudaSplit()/cvgCudaMerge()/cvgCudaAddWeighted()/cvgCudaLut() and cvgCudaTranspose()
* [U] major cleaning of arithm_logic.h, reshaping works fine when no mask is required
* [B] cvgCudaAddWeighted() does not work under linux
* [B] cvgCudaPow() && cvgCudaFlip() fail when running cxcore_all command but not when running as stand alone..??
* [B] cvgCudaAvg()/cvgCudaScaleAdd are not working
cxcoreg:
* [B] cvgMerge() is not working

cxcore_gcu: * [F] Fix issue with GCU_NO_RESHAPE option ...

allusse — Wed, 03 Mar 2010 17:42:57 +0100

Rev 519 - allusse (44 file(s) modified)

cxcore_gcu:
* [F] Fix issue with GCU_NO_RESHAPE option that was not use correctly in CudaArithm_SwitchCHANNELS(). Now arithmetic and logic operators are optimized for images of w*h*channels = multiple of 4 or 2. (Speed factor x9).
switch:
* [A] add a filename field for all XML files involved in DllManager and CL_FctSw_Mngr classes.
* [F] switch function were crashing when no DLL were loaded. Fix with an error msg and a clean exit().
DOC:
* [F] fix some synthaxe errors in main doxygen documentation
* [A] add new sections: "Tools" and "OpenCV programming tips"
BUILD:
* [F] include sugoitools folder before the resources folder. In case we have a local version of Sugoitools locally.

CUDA: => port to CUDA 2.3 [F]: ...

allusse — Wed, 24 Feb 2010 16:28:11 +0100

Rev 514 - allusse (104 file(s) modified)

CUDA*: => port to CUDA 2.3
* [F]: *.cu used to include opencv headers, it was generating compiling issue related to SSE and MMX instrucitons that are not managed by NVCC
* [F] define new 8Clamp() function to be called in kernel. NVCC was complaining about previous one.
* [F] re-enable some dead kernels that were not working with CUDA 2.1
* [U] Update CUDA makefiles, add option NO_LINKING=[0;1;2] to tell CUDA common makefile to link or not the libraries.
BUILD:
* [A]: Add new Premake4 files that will soon support MacOS target with Xcode. Almost ready to compile external plugins from contributors.

=> CUDA 2.3 tested under linux x64, most opertors working. Further tests must be done.
=> CUDA 2.3 must be tested under windows

* GpuCVCuda: * [A] Setup informations for CUDPP. ...

allusse — Sat, 21 Mar 2009 14:28:00 +0100

Rev 509 - allusse (14 file(s) modified)

* GpuCVCuda:
* [A] Setup informations for CUDPP.
* [F] fix some DLL exports macro for windows.
* [F] vscuda.rules, option -D _WINDOWS was missing
* [I] reduction kernels disabled.
* [I] Convolutions kernels disabled.
* [U] CUDPP dependencies is only used in CUDA plugins, not in GpuCVCuda

* GPUCVCuda: * [F] Fix some compilation error ...

allusse — Wed, 11 Feb 2009 13:20:33 +0100

Rev 502 - allusse (29 file(s) modified)

* GPUCVCuda:
* [F] Fix some compilation error under linux on AMD64 platforms. Specially about having CUDA shared libraries working.

* GpuCV-Cuda: ...

allusse — Tue, 10 Feb 2009 14:46:32 +0100

Rev 501 - allusse (22 file(s) modified)

* GpuCV-Cuda:
* [U] global MakeFile and ./etc/*.mk that are included into each locl MakeFileCUDA in CUDA plugins. Updated from CUDA 2.1 makefile.
* [A] *gcu/config.h: add some preprocessor directive to (#undef __SSE2__, #undef _MM_SHUFFLE2, #undef CV_ICC) to avoid SSE and MMX error while compiling plugins. See file documentation.
* [UI] *.cu: Most cuda kernel and operators are disabled under LINUX=> not compatible with CUDA 2.1 and may be linux C style...
* Global:
* [U] Premake.lua: add support for Linux 32b/64b compilation using $(TARGET_ARCH_SUFFIXE) and $(TARGET_ARCH) in generated MakeFiles.

* GpuCV CUda: * [A] Add a test ...

allusse — Wed, 04 Feb 2009 16:44:13 +0100

Rev 500 - allusse (11 file(s) modified)

* GpuCV CUda:
* [A] Add a test to disable operators that are not compatible with CUDA 2.1
* [U] Improve 64 bit support

* Global: * Update PREMAKE/LUA projects creation scripts: ...

allusse — Mon, 02 Feb 2009 16:51:52 +0100

Rev 499 - allusse (64 file(s) modified)

* Global:
* Update PREMAKE/LUA projects creation scripts:
* [U] use as much system paths and env. variables paths.
* [A] Start support for AMD64 arch.
* GPUCVCuda:
* [F] some kernel template definitions to be compatible with CUDA 2.1
* [I] All kernels might be tested...
* |U] Cuda rules to use system path and default directories.

* GPUCVCuda: * [AU] local make files to ...

allusse — Tue, 13 Jan 2009 11:55:23 +0100

Rev 492 - allusse (60 file(s) modified)

* GPUCVCuda:
* [AU] local make files to parse all subdirs and compile all .cu files
* [R] rename "cxcoregcu/oper_array/Discrete transforms" into "cxcoregcu/oper_array/discrete_transforms/"
* Resources:
* [U] SugoiTools
* Global:
* [F] fix some issue about OpenGL compatiblity detection. Now GLSL and CUDA are disabled into GPUCVConsole when the graphics card is not compatible.

gpucv

doc: * add draft tutorial for NVIDIA Parallel ...

Global: * remove some obsolete scripts from ./bin/ ...

- update Parabuild tutorial - clean some warning in code.

CUDA: * move custom build rules for VS into etc/. They ...

CUDA: * [U] update cuda makefiles to compile ...

moved back branch allusse/cuda-3.0 to experimental/trunk

* [U] update and clean allusse experimental ...

CUDA: * remove obsolete CUDA DLLs. ...

GPUVCamDemo: *[F] Fixed issue with missing library ...

Global: * [M] Moved some data files from ...

CUDA: * [F] fixed cvgCudaThreshold()/cvgCudaCalcHist()/cvgCudaDFT()/cvgCudaNot() and ...

BUILD: * [U] the internal preprocessing macros used ...

cxcore_gcu: * [F] Fix issue with GCU_NO_RESHAPE option ...

CUDA*: => port to CUDA 2.3 * [F]: ...

* GpuCVCuda: * [A] Setup informations for CUDPP. ...

* GPUCVCuda: * [F] Fix some compilation error ...

* GpuCV-Cuda: ...

* GpuCV CUda: * [A] Add a test ...

* Global: * Update PREMAKE/LUA projects creation scripts: ...

* GPUCVCuda: * [AU] local make files to ...

CUDA: => port to CUDA 2.3 [F]: ...