Troubleshooting#

Can’t reconstruct with FBP_CUDA and SIRT_CUDA#

These algorithms have 2 requirements:

  • Having a CUDA-compatible graphics card (GPU)

  • Having the CUDA Runtime 10.2 libraries installed

Having a CUDA-compatible graphics card#

Please check that your GPU is on the list of compatible GPUs: https://developer.nvidia.com/cuda-gpus

Having the CUDA Runtime 10.2 libraries installed#

Please install the CUDA Runtime version 10.2 binaries from https://developer.nvidia.com/cuda-10.2-download-archive

Reinstalling CUDA on Linux can be quickly done with this script:

# uninstall nvidia driver and cuda driver
nvidia-uninstall && /usr/local/cuda/bin/cuda-uninstaller

# download the CUDA installer
wget http://developer.download.nvidia.com/compute/cuda/10.2/Prod/local_installers/cuda_10.2.89_440.33.01_linux.run ~/Downloads/cuda_10.2.89_440.33.01_linux.run

# install GPU driver and CUDA Toolkit from CLI silently
sudo bash ~/Downloads/cuda_10.2.89_440.33.01_linux.run --silent --driver --toolkit

Specific Errors#

SystemError: <built-in function connectSlotsByName> returned a result with an error set#

This means that the PyQt package is missing or an old version (it must be newer than PyQt5==5.13.2).

Suggested Fix#

Install PyQt5 and pyqtgraph with pip install pyqt5==5.15 pyqtgraph==0.12.

qt.qpa.plugin: Could not load the Qt platform plugin “xcb” in “” even though it was found#

This might mean that some system libraries are missing.

Suggested Fix#

Try installing apt install libxcb-xinerama0.

qt.qpa.xcb: could not connect to display…#

Missing DISPLAY variable. This means the application cannot find the display to show itself on.

Suggested Fix#

You can set this via export DISPLAY=:N where :N should be the number of your display.

IDAaaS Terminal Error Messages#

Errors from the terminal when Mantid Imaging is launched, such as:

ERROR: ld.so: object 'libdlfaker.so' from LD_PRELOAD cannot be preloaded: ignored.

These are harmless and can be ignored.

ImportError: DLL load failed while importing astra_c: The specified module could not be found#

This error can arise when the CUDA-version of astra-toolbox is installed on a machine that doesn’t have a GPU. By default, the CUDA-version of astra-toolbox is installed during environment setup.

Suggested Fix#

Install the Python-version of astra-toolbox with mamba install astra-toolbox=*=py*.

On this page
Show Source