Phriction Projects Wikis Bioimaging And Optics Platform Image Processing Deconvolution CUDA Deconvolution History Version 2 vs 5
Version 2 vs 5
Version 2 vs 5
Content Changes
Content Changes
= Quoi La Patate? =
Stephan Preibish and the teams at MPI-CBG created a CUDA Deconvolution Library that their MultiView Reconstruction tool can use in order to Fuse/Deconvolve MultiView Lightsheet Images with beads.
But this involves compiling this on the machine that will run the deconvolution and is not as simple as just downloading the library because
1. It is dependent on the system architecture
2. It is dependend on the GPU that is used.
It would be too tedious to offer libraries for each GPU so you'll have to suffer through this thing at least once (per machine)
= Installation =
You need to install a lot of stuff to build these CUDA libraries
1. **[[https://www.visualstudio.com/downloads/| Visual Studio 2013 Community Edition]]** that contains the compiler
2. **[[https://developer.nvidia.com/cuda-toolkit| NVIDIA CUDA Toolkit]]** that contains the libraries we need
3. **[[https://cmake.org/download/|CMAKE]]** that simplifies (and I am not even joking) the building of these libraries
4. **[[https://github.com/StephanPreibisch/FourierConvolutionCUDALib| The Source Code from GitHub we will be building]]**
= Build =
Install all of the 3 tools first, then clone or copy the GitHub project somewhere.
Suppose the project was copied to `C:\cuda-fft-lib\`
Moreover, suppose we want the final results to be under `C:/compiled-cuda-fft-lib\`
Start a command prompt {key Win } + `cmd` + {key Enter}
```
cd C:\cuda-fft-lib
mkdir build
cd build
cmake.exe -G "Visual Studio 14 2015 Win64" -DCMAKE_INSTALL_PREFIX=C:/compiled-cuda-fft-lib\ ..
cmake.exe --build . --target install --config Release
```
That should give you the files in the `C:/compiled-cuda-fft-lib\` that you can now move into the Fiji folder
= Quoi La Patate? =
Stephan Preibish and the teams at MPI-CBG created a CUDA Deconvolution Library that their MultiView Reconstruction tool can use in order to Fuse/Deconvolve MultiView Lightsheet Images with beads.
But this involves compiling this on the machine that will run the deconvolution and is not as simple as just downloading the library because
1. It is dependent on the system architecture
2. It is dependend on the GPU that is used.
It would be too tedious to offer libraries for each GPU so you'll have to suffer through this thing at least once (per machine)
= Installation =
You need to install a lot of stuff to build these CUDA libraries
1. **[[https://www.visualstudio.com/downloads/| Visual Studio 2013 Community Edition]]** that contains the compiler
2. **[[https://developer.nvidia.com/cuda-toolkit| NVIDIA CUDA Toolkit]]** that contains the libraries we need
3. **[[https://cmake.org/download/|CMAKE]]** that simplifies (and I am not even joking) the building of these libraries
4. **[[https://github.com/StephanPreibisch/FourierConvolutionCUDALib| The Source Code from GitHub we will be building]]**
= Build =
Install all of the 3 tools first, then clone or copy the GitHub project somewhere.
Suppose the project was copied to `C:\cuda-fft-lib\`
Moreover, suppose we want the final results to be under `C:/compiled-cuda-fft-lib\`
Start a command prompt {key Win } + `cmd` + {key Enter}
```
cd C:\cuda-fft-lib
mkdir build
cd build
cmake.exe -G "Visual Studio 14 2015 Win64" -DCMAKE_INSTALL_PREFIX=C:/compiled-cuda-fft-lib\ ..
cmake.exe --build . --target install --config Release
```
That should give you the files in the `C:/compiled-cuda-fft-lib\` that you can now move into the Fiji folder
c4science · Help